diff --git a/migrations/00000000000000_diesel_initial_setup/down.sql b/migrations/00000000000000_diesel_initial_setup/down.sql
new file mode 100644
index 00000000..a9f52609
--- /dev/null
+++ b/migrations/00000000000000_diesel_initial_setup/down.sql
@@ -0,0 +1,6 @@
+-- This file was automatically created by Diesel to setup helper functions
+-- and other internal bookkeeping. This file is safe to edit, any future
+-- changes will be added to existing projects as new migrations.
+
+DROP FUNCTION IF EXISTS diesel_manage_updated_at(_tbl regclass);
+DROP FUNCTION IF EXISTS diesel_set_updated_at();
diff --git a/migrations/00000000000000_diesel_initial_setup/up.sql b/migrations/00000000000000_diesel_initial_setup/up.sql
new file mode 100644
index 00000000..d68895b1
--- /dev/null
+++ b/migrations/00000000000000_diesel_initial_setup/up.sql
@@ -0,0 +1,36 @@
+-- This file was automatically created by Diesel to setup helper functions
+-- and other internal bookkeeping. This file is safe to edit, any future
+-- changes will be added to existing projects as new migrations.
+
+
+
+
+-- Sets up a trigger for the given table to automatically set a column called
+-- `updated_at` whenever the row is modified (unless `updated_at` was included
+-- in the modified columns)
+--
+-- # Example
+--
+-- ```sql
+-- CREATE TABLE users (id SERIAL PRIMARY KEY, updated_at TIMESTAMP NOT NULL DEFAULT NOW());
+--
+-- SELECT diesel_manage_updated_at('users');
+-- ```
+CREATE OR REPLACE FUNCTION diesel_manage_updated_at(_tbl regclass) RETURNS VOID AS $$
+BEGIN
+    EXECUTE format('CREATE TRIGGER set_updated_at BEFORE UPDATE ON %s
+                    FOR EACH ROW EXECUTE PROCEDURE diesel_set_updated_at()', _tbl);
+END;
+$$ LANGUAGE plpgsql;
+
+CREATE OR REPLACE FUNCTION diesel_set_updated_at() RETURNS trigger AS $$
+BEGIN
+    IF (
+        NEW IS DISTINCT FROM OLD AND
+        NEW.updated_at IS NOT DISTINCT FROM OLD.updated_at
+    ) THEN
+        NEW.updated_at := current_timestamp;
+    END IF;
+    RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
diff --git a/migrations/2025-11-01-114838_actors/up.sql b/migrations/2025-11-01-114838_actors/up.sql
index 2061e3e4..29d00559 100644
--- a/migrations/2025-11-01-114838_actors/up.sql
+++ b/migrations/2025-11-01-114838_actors/up.sql
@@ -14,6 +14,9 @@
 CREATE TYPE actor_status AS ENUM ('active', 'takendown', 'suspended', 'deleted', 'deactivated');
 CREATE TYPE actor_sync_state AS ENUM ('synced', 'dirty', 'partial', 'processing');
 
+COMMENT ON TYPE actor_status IS 'Actor account status: active (normal), takendown (moderation), suspended (temporary), deleted (user-initiated), deactivated (user-initiated temporary)';
+COMMENT ON TYPE actor_sync_state IS 'Repository sync state: synced (up-to-date), dirty (needs sync), partial (incomplete data), processing (sync in progress)';
+
 -- =============================================================================
 -- PREFERENCE TYPES
 -- =============================================================================
@@ -21,12 +24,17 @@ CREATE TYPE actor_sync_state AS ENUM ('synced', 'dirty', 'partial', 'processing'
 CREATE TYPE chat_allow_incoming AS ENUM ('all', 'following', 'none');
 CREATE TYPE status_type AS ENUM ('app.bsky.actor.status#live');
 
+COMMENT ON TYPE chat_allow_incoming IS 'Chat preference: who can send direct messages to this actor';
+COMMENT ON TYPE status_type IS 'Actor status type (currently only live streaming supported)';
+
 -- MIME Types (used in status records)
 CREATE TYPE image_mime_type AS ENUM (
     'image/jpeg', 'image/png', 'image/webp', 'image/gif', 'image/avif',
     'image/svg+xml', 'image/bmp', 'image/tiff', 'image/heif', 'image/heic', 'image/jxl'
 );
 
+COMMENT ON TYPE image_mime_type IS 'Supported image MIME types for avatars, banners, and thumbnails';
+
 -- =============================================================================
 -- CORE IDENTITY TABLES
 -- =============================================================================
@@ -43,12 +51,29 @@ CREATE TABLE actors (
     account_created_at TIMESTAMPTZ
 );
 
+COMMENT ON TABLE actors IS 'Core identity table: one row per ATProto DID. Tracks handle, status, and repository sync state.';
+COMMENT ON COLUMN actors.id IS 'Synthetic primary key for internal references';
+COMMENT ON COLUMN actors.did IS 'Decentralized identifier (did:plc:* or did:web:*) - globally unique';
+COMMENT ON COLUMN actors.handle IS 'Human-readable handle (e.g., alice.bsky.social). Nullable for actors without handles.';
+COMMENT ON COLUMN actors.status IS 'Account status (active, takendown, suspended, deleted, deactivated)';
+COMMENT ON COLUMN actors.sync_state IS 'Repository sync state (synced, dirty, partial, processing)';
+COMMENT ON COLUMN actors.repo_rev IS 'Repository revision string (opaque token from PDS)';
+COMMENT ON COLUMN actors.repo_cid IS 'Repository commit CID (content identifier)';
+COMMENT ON COLUMN actors.last_indexed IS 'Timestamp of last successful indexing operation';
+COMMENT ON COLUMN actors.account_created_at IS 'Account creation timestamp from PDS';
+
 CREATE INDEX idx_actors_handle ON actors(handle) WHERE handle IS NOT NULL;
 CREATE INDEX idx_actors_sync_state ON actors(sync_state);
 CREATE INDEX idx_actors_status ON actors(status);
 CREATE INDEX idx_actors_last_indexed ON actors(last_indexed) WHERE last_indexed IS NOT NULL;
 CREATE INDEX idx_actors_account_created_at ON actors(account_created_at) WHERE account_created_at IS NOT NULL;
 
+COMMENT ON INDEX idx_actors_handle IS 'Lookup actors by handle (partial index excluding NULL handles)';
+COMMENT ON INDEX idx_actors_sync_state IS 'Find actors needing sync (dirty/partial states)';
+COMMENT ON INDEX idx_actors_status IS 'Filter actors by status (e.g., exclude takendown)';
+COMMENT ON INDEX idx_actors_last_indexed IS 'Find stale actors for re-indexing';
+COMMENT ON INDEX idx_actors_account_created_at IS 'Query actors by signup date';
+
 
 -- =============================================================================
 -- PROFILES
@@ -62,12 +87,25 @@ CREATE TABLE profiles (
     banner_cid BYTEA,
     display_name TEXT,
     description TEXT,
-    pinned_post_id BIGINT, -- FK added in posts migration
+    pinned_post_rkey BIGINT,  -- FK added after posts migration (natural key)
     joined_sp_id BIGINT,   -- FK added in feeds migration
     pronouns TEXT,
     website TEXT
 );
 
+COMMENT ON TABLE profiles IS 'User profiles (app.bsky.actor.profile): display name, bio, avatar, banner, etc. One profile per actor.';
+COMMENT ON COLUMN profiles.actor_id IS 'Reference to actors table (also primary key)';
+COMMENT ON COLUMN profiles.cid IS 'Content identifier (CID) of profile record';
+COMMENT ON COLUMN profiles.created_at IS 'Profile record creation timestamp';
+COMMENT ON COLUMN profiles.avatar_cid IS 'Avatar image CID (blob reference)';
+COMMENT ON COLUMN profiles.banner_cid IS 'Banner image CID (blob reference)';
+COMMENT ON COLUMN profiles.display_name IS 'User-facing display name (e.g., "Alice Johnson")';
+COMMENT ON COLUMN profiles.description IS 'Profile bio/description text';
+COMMENT ON COLUMN profiles.pinned_post_rkey IS 'Pinned post rkey (natural key). No FK constraint due to hypertable limitation.';
+COMMENT ON COLUMN profiles.joined_sp_id IS 'Joined via starter pack ID (FK added in feeds migration)';
+COMMENT ON COLUMN profiles.pronouns IS 'User pronouns (e.g., "she/her", "they/them")';
+COMMENT ON COLUMN profiles.website IS 'Personal website URL';
+
 -- Full-text search on profiles
 ALTER TABLE profiles ADD COLUMN search_vector tsvector
     GENERATED ALWAYS AS (
@@ -75,8 +113,12 @@ ALTER TABLE profiles ADD COLUMN search_vector tsvector
         setweight(to_tsvector('english', coalesce(description, '')), 'B')
     ) STORED;
 
+COMMENT ON COLUMN profiles.search_vector IS 'Full-text search vector: display_name (weight A) + description (weight B)';
+
 CREATE INDEX idx_profiles_search ON profiles USING GIN(search_vector);
 
+COMMENT ON INDEX idx_profiles_search IS 'Full-text search on profile display names and bios';
+
 -- =============================================================================
 -- USER PREFERENCES
 -- =============================================================================
@@ -88,6 +130,11 @@ CREATE TABLE chat_decls (
     created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
 );
 
+COMMENT ON TABLE chat_decls IS 'Chat preferences (app.bsky.actor.defs#chatDeclaration): who can send direct messages';
+COMMENT ON COLUMN chat_decls.actor_id IS 'Reference to actors table (also primary key)';
+COMMENT ON COLUMN chat_decls.allow_incoming IS 'Who can message this actor: all, following, or none';
+COMMENT ON COLUMN chat_decls.created_at IS 'Timestamp when chat preference was set';
+
 -- Status Records (Live streaming status)
 CREATE TABLE statuses (
     actor_id INTEGER PRIMARY KEY REFERENCES actors(id) ON DELETE CASCADE,
@@ -95,7 +142,19 @@ CREATE TABLE statuses (
     created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
     status status_type NOT NULL,
     duration INTEGER,
-    embed_post_id BIGINT, -- FK added in posts migration
+    embed_post_actor_id INTEGER,  -- Natural key references to posts (no FK - hypertable limitation)
+    embed_post_rkey BIGINT,
     thumb_mime_type image_mime_type,
     thumb_cid BYTEA
 );
+
+COMMENT ON TABLE statuses IS 'Actor status records (app.bsky.actor.status): live streaming status and metadata';
+COMMENT ON COLUMN statuses.actor_id IS 'Reference to actors table (also primary key)';
+COMMENT ON COLUMN statuses.cid IS 'Content identifier (CID) of status record';
+COMMENT ON COLUMN statuses.created_at IS 'Status record creation timestamp';
+COMMENT ON COLUMN statuses.status IS 'Status type (currently only app.bsky.actor.status#live)';
+COMMENT ON COLUMN statuses.duration IS 'Stream duration in seconds (nullable for ongoing streams)';
+COMMENT ON COLUMN statuses.embed_post_actor_id IS 'Referenced post actor_id (no FK due to hypertable limitation)';
+COMMENT ON COLUMN statuses.embed_post_rkey IS 'Referenced post rkey (natural key part 2)';
+COMMENT ON COLUMN statuses.thumb_mime_type IS 'Thumbnail image MIME type';
+COMMENT ON COLUMN statuses.thumb_cid IS 'Thumbnail image CID (blob reference)';
diff --git a/migrations/2025-11-01-114839_posts/up.sql b/migrations/2025-11-01-114839_posts/up.sql
index ff73017d..6f4d4692 100644
--- a/migrations/2025-11-01-114839_posts/up.sql
+++ b/migrations/2025-11-01-114839_posts/up.sql
@@ -2,19 +2,35 @@
 -- POSTS & CONTENT
 -- =============================================================================
 --
--- Posts, embeds (images, video, external links, record quotes), facets
--- (mentions, links, tags), and URI deduplication
+-- Posts, denormalized embeds/facets, and URI deduplication
 --
 -- =============================================================================
 
+-- =============================================================================
+-- ENABLE TIMESCALEDB
+-- =============================================================================
+-- TimescaleDB must be enabled before creating hypertables
+CREATE EXTENSION IF NOT EXISTS timescaledb CASCADE;
+
 -- =============================================================================
 -- POST & EMBED TYPES
 -- =============================================================================
 
 CREATE TYPE post_status AS ENUM ('complete', 'stub', 'missing', 'deleted', 'forbidden', 'pruned');
 CREATE TYPE embed_type AS ENUM ('images', 'video', 'external', 'record', 'record_with_media');
-CREATE TYPE video_mime_type AS ENUM ('video/mp4', 'video/webm', 'video/quicktime', 'video/mpeg');
-CREATE TYPE caption_mime_type AS ENUM ('text/vtt');
+
+-- Video mime types (GIFs are treated as video embeds in Bluesky)
+CREATE TYPE video_mime_type AS ENUM (
+    'video/mp4', 'video/webm', 'video/quicktime', 'video/mpeg',
+    'image/gif', 'video/x-m4v', 'video/3gpp',
+    'video/x-msvideo', 'video/x-matroska', 'video/ogg'
+);
+
+-- Caption mime types (subtitles)
+CREATE TYPE caption_mime_type AS ENUM (
+    'text/vtt', 'text/srt', 'application/x-subrip',
+    'text/x-ssa', 'text/x-ass'
+);
 
 -- Language Codes (ISO 639-1 + regional variants)
 -- Includes all language codes from historical database (4.6M posts)
@@ -47,9 +63,8 @@ CREATE TYPE language_code AS ENUM (
 
 CREATE TYPE facet_type AS ENUM ('mention', 'link', 'tag');
 
--- =============================================================================
--- POSTS
--- =============================================================================
+COMMENT ON TYPE language_code IS 'ISO 639-1 language codes (+ regional variants) for post content. Covers 99.99%+ of real-world usage on Bluesky.';
+COMMENT ON TYPE facet_type IS 'Post facet types: mention (@handle), link (URL), tag (#hashtag)';
 
 -- =============================================================================
 -- TID HELPER FUNCTIONS
@@ -113,143 +128,210 @@ $$ LANGUAGE plpgsql IMMUTABLE STRICT;
 
 COMMENT ON FUNCTION i64_to_tid(BIGINT) IS 'Encodes a BIGINT to a base32-sortable TID string (13 characters). Uses AT Protocol TID encoding: https://atproto.com/specs/tid';
 
+-- =============================================================================
+-- DENORMALIZED EMBED COMPOSITE TYPES
+-- =============================================================================
+
+-- External link embed (used by 0.8% of posts)
+CREATE TYPE post_ext_embed AS (
+    uri text,
+    title text,
+    description text,
+    thumb_mime_type image_mime_type,
+    thumb_cid bytea
+);
+
+COMMENT ON TYPE post_ext_embed IS 'External link embed (0.8% of posts): URL, title, description, thumbnail';
+
+-- Video caption (subtitles)
+CREATE TYPE post_video_caption AS (
+    lang language_code,
+    mime_type caption_mime_type,
+    cid bytea
+);
+
+COMMENT ON TYPE post_video_caption IS 'Video caption track: language, MIME type, blob CID';
+
+-- Video embed (used by 0.16% of posts)
+CREATE TYPE post_video_embed AS (
+    mime_type video_mime_type,
+    cid bytea,
+    alt text,
+    width integer,
+    height integer,
+    caption_1 post_video_caption,  -- First caption track (e.g., English)
+    caption_2 post_video_caption,  -- Second caption track (e.g., Spanish)
+    caption_3 post_video_caption   -- Third caption track (e.g., French)
+);
+
+COMMENT ON TYPE post_video_embed IS 'Video embed (0.16% of posts): metadata + up to 3 caption tracks';
+
+-- Image embed (used by 1.8% of posts, max 4 per Bluesky protocol)
+CREATE TYPE post_image_embed AS (
+    mime_type image_mime_type,
+    cid bytea,
+    alt text,
+    width integer,
+    height integer
+);
+
+COMMENT ON TYPE post_image_embed IS 'Image embed metadata: MIME type, CID, alt text, dimensions. Max 4 per post.';
+
+-- Facet embed (facet_type, index_start, index_end, link_uri, mention_actor_id, tag)
+-- Used by 12.5% of posts, 99.3% have ≤8 facets
+CREATE TYPE post_facet_embed AS (
+    facet_type facet_type,
+    index_start integer,
+    index_end integer,
+    link_uri text,
+    mention_actor_id integer,
+    tag text
+);
+
+COMMENT ON TYPE post_facet_embed IS 'Rich text facet: type (mention/link/tag) + position + metadata. 12.5% of posts use facets, 99.3% have ≤8.';
+
+-- =============================================================================
+-- POSTS TABLE
+-- =============================================================================
+
 CREATE TABLE posts (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
     cid BYTEA NOT NULL,
     content BYTEA,
     langs language_code[] NOT NULL DEFAULT '{}',
     tags TEXT[] NOT NULL DEFAULT '{}',
-    parent_post_id BIGINT REFERENCES posts(id) ON DELETE SET NULL,
-    root_post_id BIGINT REFERENCES posts(id) ON DELETE SET NULL,
+    tokens TEXT[],  -- Token-based search (replaces search_vector)
+    parent_post_actor_id INTEGER,
+    parent_post_rkey BIGINT,
+    root_post_actor_id INTEGER,
+    root_post_rkey BIGINT,
     embed_type embed_type,
     embed_subtype embed_type,
     violates_threadgate BOOLEAN NOT NULL DEFAULT FALSE,
     status post_status NOT NULL DEFAULT 'complete',
-    search_vector tsvector,
-    UNIQUE (actor_id, rkey),
+    -- Denormalized embeds (NULL columns compress to ~0 bytes in TimescaleDB)
+    ext_embed post_ext_embed,
+    video_embed post_video_embed,
+    embedded_post_actor_id INTEGER,
+    embedded_post_rkey BIGINT,
+    record_detached BOOLEAN,
+    image_1 post_image_embed,
+    image_2 post_image_embed,
+    image_3 post_image_embed,
+    image_4 post_image_embed,
+    facet_1 post_facet_embed,
+    facet_2 post_facet_embed,
+    facet_3 post_facet_embed,
+    facet_4 post_facet_embed,
+    facet_5 post_facet_embed,
+    facet_6 post_facet_embed,
+    facet_7 post_facet_embed,
+    facet_8 post_facet_embed,
+    mentions INTEGER[],  -- Array of actor_ids (denormalized, max 8)
+    PRIMARY KEY (actor_id, rkey),
     CHECK (array_length(langs, 1) IS NULL OR array_length(langs, 1) <= 3)
 );
 
 COMMENT ON COLUMN posts.content IS 'Zstd-compressed post content (BYTEA). NULL for stub/deleted posts. Dictionary version stored as first byte prefix.';
 COMMENT ON COLUMN posts.langs IS 'Languages of post content (ISO 639-1 codes). Max 3 per AT Protocol spec. Empty array means no language specified.';
-COMMENT ON COLUMN posts.search_vector IS 'Full-text search vector (NULL except for allowlisted users)';
+COMMENT ON COLUMN posts.tokens IS 'Token-based search tokens (replaces tsvector). NULL for non-indexed posts.';
 COMMENT ON COLUMN posts.status IS 'Post lifecycle state: complete (normal), stub (placeholder, needs fetch), missing (permanently unfetchable), deleted (soft-deleted), forbidden (access denied)';
+COMMENT ON COLUMN posts.embedded_post_actor_id IS 'Quote post reference - actor part';
+COMMENT ON COLUMN posts.embedded_post_rkey IS 'Quote post reference - rkey part';
+COMMENT ON COLUMN posts.record_detached IS 'Whether the quoted post is detached (postgate)';
 
 CREATE INDEX idx_posts_actor ON posts(actor_id);
-CREATE INDEX idx_posts_parent ON posts(parent_post_id) WHERE parent_post_id IS NOT NULL;
-CREATE INDEX idx_posts_root ON posts(root_post_id) WHERE root_post_id IS NOT NULL;
+CREATE INDEX idx_posts_parent
+    ON posts(parent_post_actor_id, parent_post_rkey)
+    WHERE parent_post_actor_id IS NOT NULL;
+CREATE INDEX idx_posts_root
+    ON posts(root_post_actor_id, root_post_rkey)
+    WHERE root_post_actor_id IS NOT NULL;
 CREATE INDEX idx_posts_langs ON posts USING GIN(langs);
 CREATE INDEX idx_posts_tags ON posts USING GIN(tags);
 CREATE INDEX idx_posts_status ON posts(status) WHERE status != 'complete';
-CREATE INDEX idx_posts_search_vector ON posts USING GIN (search_vector) WHERE search_vector IS NOT NULL;
 CREATE INDEX idx_posts_rkey_prunable ON posts (rkey) WHERE status NOT IN ('forbidden', 'pruned');
-
--- Add foreign key constraints to profiles and statuses
-ALTER TABLE profiles ADD CONSTRAINT fk_profiles_pinned_post
-    FOREIGN KEY (pinned_post_id) REFERENCES posts(id) ON DELETE SET NULL;
-
-ALTER TABLE statuses ADD CONSTRAINT fk_statuses_embed_post
-    FOREIGN KEY (embed_post_id) REFERENCES posts(id) ON DELETE SET NULL;
-
--- Add indexes on foreign keys for improved DELETE performance
-CREATE INDEX idx_profiles_pinned_post ON profiles(pinned_post_id) WHERE pinned_post_id IS NOT NULL;
+CREATE INDEX idx_posts_tokens_gin ON posts USING GIN (tokens) WHERE tokens IS NOT NULL;
+CREATE INDEX idx_posts_embedded_post
+    ON posts(embedded_post_actor_id, embedded_post_rkey)
+    WHERE embedded_post_actor_id IS NOT NULL;
+CREATE INDEX idx_posts_mentions ON posts USING GIN(mentions) WHERE mentions IS NOT NULL;
+
+COMMENT ON TABLE posts IS 'Posts (app.bsky.feed.post): text, embeds, facets, reply metadata. TimescaleDB hypertable with 1-day chunks.';
+COMMENT ON INDEX idx_posts_actor IS 'Lookup all posts by actor (author timeline)';
+COMMENT ON INDEX idx_posts_parent IS 'Find replies to a specific post (partial: only WHERE parent exists)';
+COMMENT ON INDEX idx_posts_root IS 'Find all posts in a thread (partial: only WHERE root exists)';
+COMMENT ON INDEX idx_posts_langs IS 'Filter posts by language (GIN index supports array overlap queries)';
+COMMENT ON INDEX idx_posts_tags IS 'Find posts with specific tags/hashtags (GIN index)';
+COMMENT ON INDEX idx_posts_status IS 'Find non-complete posts (stubs, deleted, etc.)';
+COMMENT ON INDEX idx_posts_rkey_prunable IS 'Find posts eligible for pruning (excludes forbidden/pruned)';
+COMMENT ON INDEX idx_posts_tokens_gin IS 'Full-text search on post tokens (GIN index, partial: only indexed posts)';
+COMMENT ON INDEX idx_posts_embedded_post IS 'Find quote posts referencing a specific post';
+COMMENT ON INDEX idx_posts_mentions IS 'Find posts mentioning a specific actor (GIN index on denormalized actor_id array)';
 
 -- =============================================================================
--- POST EMBEDS
+-- Convert posts to TimescaleDB Hypertable
+-- =============================================================================
+--
+-- Configuration:
+-- - All posts (complete + stubs)
+-- - NO retention policy: Full retention for network scale (2B posts)
+-- - Chunk interval: 1 day
+-- - Compression: 12 hours (aggressive)
+-- - Expected compression: ~100:1 combined
+--
+-- Storage projections at 2B posts with 10% complete, 90% stubs:
+-- - Total: ~90 GB (vs ~540 GB traditional, 83% savings)
+--
+-- NOTE: TimescaleDB does not support foreign keys TO hypertables, so we do NOT
+-- create FK constraints pointing to posts table. Referential integrity is
+-- enforced at the application level.
 -- =============================================================================
 
--- Post Embeds: Images
-CREATE TABLE post_embed_images (
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    seq SMALLINT NOT NULL,
-    mime_type image_mime_type NOT NULL,
-    cid BYTEA NOT NULL,
-    alt TEXT,
-    width INTEGER,
-    height INTEGER,
-    PRIMARY KEY (post_id, seq)
-);
-
--- Post Embeds: Video
-CREATE TABLE post_embed_video (
-    post_id BIGINT PRIMARY KEY REFERENCES posts(id) ON DELETE CASCADE,
-    mime_type video_mime_type NOT NULL,
-    cid BYTEA NOT NULL,
-    alt TEXT,
-    width INTEGER,
-    height INTEGER
-);
-
--- Post Embeds: Video Captions
-CREATE TABLE post_embed_video_captions (
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    language language_code NOT NULL,
-    mime_type caption_mime_type NOT NULL,
-    cid BYTEA NOT NULL,
-    PRIMARY KEY (post_id, language)
-);
-
--- Post Embeds: External Links
-CREATE TABLE post_embed_ext (
-    post_id BIGINT PRIMARY KEY REFERENCES posts(id) ON DELETE CASCADE,
-    uri TEXT NOT NULL,
-    title TEXT NOT NULL,
-    description TEXT NOT NULL,
-    thumb_mime_type image_mime_type,
-    thumb_cid BYTEA
-);
-
--- Post Embeds: Record Embeds (quote posts)
-CREATE TABLE post_embed_record (
-    post_id BIGINT PRIMARY KEY REFERENCES posts(id) ON DELETE CASCADE,
-    embedded_post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    detached BOOLEAN NOT NULL DEFAULT FALSE
+-- Convert to hypertable using rkey column (time-based partitioning via tid_timestamp)
+SELECT create_hypertable(
+    'posts',
+    'rkey',
+    chunk_time_interval => (86400000000::bigint * 1000),  -- 1 day in microseconds
+    time_partitioning_func => 'tid_timestamp',
+    migrate_data => true,
+    if_not_exists => true
 );
 
-CREATE INDEX idx_post_embed_record_embedded ON post_embed_record(embedded_post_id);
-
--- =============================================================================
--- POST FACETS & MENTIONS
--- =============================================================================
+COMMENT ON TABLE posts IS 'TimescaleDB hypertable for posts (complete + stubs). Partitioned by rkey with 1-day chunks. Auto-compressed after 12 hours. No retention policy (full network scale).';
 
--- Post Facets (mentions, links, tags)
-CREATE TABLE post_facets (
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    facet_index SMALLINT NOT NULL,
-    byte_start INTEGER NOT NULL,
-    byte_end INTEGER NOT NULL,
-    facet_type facet_type NOT NULL,
-    mention_actor_id INTEGER REFERENCES actors(id) ON DELETE CASCADE,
-    link_uri_id INTEGER,
-    tag TEXT,
-    PRIMARY KEY (post_id, facet_index)
+-- Enable compression
+ALTER TABLE posts SET (
+    timescaledb.compress,
+    timescaledb.compress_segmentby = 'actor_id',
+    timescaledb.compress_orderby = 'rkey DESC'
 );
 
-CREATE INDEX idx_post_facets_mention ON post_facets(mention_actor_id) WHERE mention_actor_id IS NOT NULL;
-CREATE INDEX idx_post_facets_link_uri ON post_facets(link_uri_id);
-
--- Post Mentions (denormalized for fast lookups)
-CREATE TABLE post_mentions (
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    mentioned_actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    PRIMARY KEY (post_id, mentioned_actor_id)
+-- Add aggressive compression policy (12 hours)
+SELECT add_compression_policy(
+    'posts',
+    compress_after => INTERVAL '12 hours',
+    if_not_exists => true
 );
 
-CREATE INDEX idx_post_mentions_actor ON post_mentions(mentioned_actor_id);
+-- Enable chunk skipping for parent/root post queries
+SELECT enable_chunk_skipping('posts', 'parent_post_actor_id');
+SELECT enable_chunk_skipping('posts', 'root_post_actor_id');
+SELECT enable_chunk_skipping('posts', 'actor_id');
 
 -- =============================================================================
 -- URI DEDUPLICATION
 -- =============================================================================
 
--- URI deduplication table
+-- URI deduplication table (for facet link URIs)
 CREATE TABLE uris (
     id SERIAL PRIMARY KEY,
     uri TEXT NOT NULL UNIQUE,
     created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
 );
 
--- Add FK from post_facets to uris
-ALTER TABLE post_facets ADD CONSTRAINT fk_post_facets_uri
-    FOREIGN KEY (link_uri_id) REFERENCES uris(id) ON DELETE CASCADE;
+COMMENT ON TABLE uris IS 'URI deduplication: stores unique URIs once, referenced by ID. Currently unused (facets denormalized).';
+COMMENT ON COLUMN uris.id IS 'Synthetic URI ID for deduplication';
+COMMENT ON COLUMN uris.uri IS 'Unique URI string';
+COMMENT ON COLUMN uris.created_at IS 'First seen timestamp';
diff --git a/migrations/2025-11-01-114840_social_graph/up.sql b/migrations/2025-11-01-114840_social_graph/up.sql
index 0b8f7adc..e0ca8cc3 100644
--- a/migrations/2025-11-01-114840_social_graph/up.sql
+++ b/migrations/2025-11-01-114840_social_graph/up.sql
@@ -8,12 +8,10 @@
 
 -- Follows
 CREATE TABLE follows (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
-    cid BYTEA NOT NULL,
     subject_actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    UNIQUE (actor_id, rkey),
+    PRIMARY KEY (actor_id, rkey),
     UNIQUE (actor_id, subject_actor_id)
 );
 
@@ -22,12 +20,10 @@ CREATE INDEX idx_follows_subject ON follows(subject_actor_id);
 
 -- Blocks
 CREATE TABLE blocks (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
-    cid BYTEA NOT NULL,
     subject_actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    UNIQUE (actor_id, rkey),
+    PRIMARY KEY (actor_id, rkey),
     UNIQUE (actor_id, subject_actor_id)
 );
 
diff --git a/migrations/2025-11-01-114841_engagement/up.sql b/migrations/2025-11-01-114841_engagement/up.sql
index 53d3e77f..1cc26464 100644
--- a/migrations/2025-11-01-114841_engagement/up.sql
+++ b/migrations/2025-11-01-114841_engagement/up.sql
@@ -10,60 +10,186 @@
 -- ENGAGEMENT TYPES
 -- =============================================================================
 
-CREATE TYPE like_subject_type AS ENUM ('post', 'feedgen', 'labeler');
 CREATE TYPE bookmark_subject_type AS ENUM ('post', 'feedgen', 'list', 'labeler', 'starterpack');
+CREATE TYPE repost_status AS ENUM ('complete', 'stub');
+
+COMMENT ON TYPE bookmark_subject_type IS 'Bookmark target types: post (most common), feedgen, list, labeler, starterpack';
+COMMENT ON TYPE repost_status IS 'Repost lifecycle state: complete (normal repost), stub (placeholder awaiting fetch)';
 
 -- =============================================================================
--- LIKES
+-- POST LIKES (99.98% of likes) - TimescaleDB Hypertable
 -- =============================================================================
 
--- Likes (99.98% target posts, 0.02% target feedgens/labelers)
-CREATE TABLE likes (
-    id BIGSERIAL PRIMARY KEY,
+CREATE TABLE post_likes (
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
-    cid BYTEA NOT NULL,
-    subject_type like_subject_type NOT NULL DEFAULT 'post',
-    subject_id BIGINT,
-    via_post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE,
-    UNIQUE (actor_id, rkey)
+    post_actor_id INTEGER NOT NULL,
+    post_rkey BIGINT NOT NULL,
+    via_repost_actor_id INTEGER,
+    via_repost_rkey BIGINT,
+    PRIMARY KEY (actor_id, rkey)
+);
+
+CREATE INDEX idx_post_likes_post_actor_id_post_rkey
+    ON post_likes(post_actor_id, post_rkey);
+CREATE INDEX idx_post_likes_rkey ON post_likes(rkey);
+CREATE INDEX idx_post_likes_via_repost
+    ON post_likes(via_repost_actor_id, via_repost_rkey)
+    WHERE via_repost_actor_id IS NOT NULL;
+
+COMMENT ON INDEX idx_post_likes_post_actor_id_post_rkey IS 'Find all likes on a specific post';
+COMMENT ON INDEX idx_post_likes_rkey IS 'Find likes by timestamp (TID-based)';
+COMMENT ON INDEX idx_post_likes_via_repost IS 'Find likes that came via a specific repost (partial: only WHERE via_repost exists)';
+
+COMMENT ON TABLE post_likes IS 'Likes on posts (99.98% of all likes). Uses natural keys for post references.';
+COMMENT ON COLUMN post_likes.actor_id IS 'Who liked the post (references actors)';
+COMMENT ON COLUMN post_likes.rkey IS 'Like record TID converted to INT8 (natural key part 2, timestamp-based)';
+COMMENT ON COLUMN post_likes.post_actor_id IS 'Liked post author (natural key part 1)';
+COMMENT ON COLUMN post_likes.post_rkey IS 'Liked post TID (natural key part 2)';
+COMMENT ON COLUMN post_likes.via_repost_actor_id IS 'If liked via repost, repost author (for notification context)';
+COMMENT ON COLUMN post_likes.via_repost_rkey IS 'If liked via repost, repost TID (for notification context)';
+
+-- Convert post_likes to TimescaleDB hypertable partitioned by rkey (timestamp-based TID)
+-- Use tid_timestamp function to extract timestamp from TID for partitioning
+-- Chunk interval: 4 hours (optimal for network scale: 20M likes/day)
+SELECT create_hypertable(
+    'post_likes',
+    'rkey',
+    chunk_time_interval => (14400000000::bigint * 1000),  -- 4 hours in microseconds
+    time_partitioning_func => 'tid_timestamp',
+    migrate_data => true,
+    if_not_exists => true
+);
+
+-- Enable compression on old chunks
+-- compress_segmentby: Group by actor_id for better compression (likes from same actor)
+-- compress_orderby: Order by rkey DESC for time-series queries
+ALTER TABLE post_likes SET (
+    timescaledb.compress,
+    timescaledb.compress_segmentby = 'actor_id',
+    timescaledb.compress_orderby = 'rkey DESC'
 );
 
-CREATE INDEX idx_likes_actor ON likes(actor_id);
-CREATE INDEX idx_likes_subject ON likes(subject_type, subject_id);
-CREATE INDEX idx_likes_rkey ON likes(rkey);
-CREATE INDEX idx_likes_via_post ON likes(via_post_id);
+-- Add aggressive compression policy: compress chunks older than 12 hours
+-- This is safe because 99.9% of writes are <5 seconds old
+SELECT add_compression_policy(
+    'post_likes',
+    compress_after => INTERVAL '12 hours',
+    if_not_exists => true
+);
+
+-- Enable chunk skipping on post_id for efficient query exclusion
+-- Chunk skipping allows queries on specific posts to skip irrelevant chunks
+SELECT enable_chunk_skipping('post_likes', 'post_actor_id');
+
+-- Enable chunk skipping on actor_id for timeline queries
+-- Allows viewer state queries to skip chunks without the viewer's likes
+SELECT enable_chunk_skipping('post_likes', 'actor_id');
+
+COMMENT ON TABLE post_likes IS 'TimescaleDB hypertable for post likes. Partitioned by rkey (timestamp) with 4-hour chunks. Auto-compressed after 12 hours. NO retention policy (full network scale). Chunk skipping enabled on post_actor_id and actor_id.';
+
+-- =============================================================================
+-- FEEDGEN LIKES (rare, ~0.01% of likes)
+-- =============================================================================
+
+CREATE TABLE feedgen_likes (
+    actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
+    rkey INT8 NOT NULL,
+    feedgen_id BIGINT NOT NULL,  -- FK to feedgens added in feeds migration
+    PRIMARY KEY (actor_id, rkey)
+);
+
+CREATE INDEX idx_feedgen_likes_feedgen ON feedgen_likes(feedgen_id);
+CREATE INDEX idx_feedgen_likes_rkey ON feedgen_likes(rkey);
+
+COMMENT ON TABLE feedgen_likes IS 'Likes on feed generators (rare, ~0.01% of likes).';
+
+-- =============================================================================
+-- LABELER LIKES (rare, ~0.01% of likes)
+-- =============================================================================
+
+CREATE TABLE labeler_likes (
+    actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
+    rkey INT8 NOT NULL,
+    labeler_actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
+    PRIMARY KEY (actor_id, rkey)
+);
+
+CREATE INDEX idx_labeler_likes_labeler ON labeler_likes(labeler_actor_id);
+CREATE INDEX idx_labeler_likes_rkey ON labeler_likes(rkey);
+
+COMMENT ON TABLE labeler_likes IS 'Likes on labeler services (rare, ~0.01% of likes).';
 
 -- =============================================================================
 -- REPOSTS
 -- =============================================================================
 
 CREATE TABLE reposts (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
-    cid BYTEA NOT NULL,
-    post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE,
-    via_post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE,
-    UNIQUE (actor_id, rkey)
+    post_actor_id INTEGER NOT NULL,
+    post_rkey BIGINT NOT NULL,
+    via_repost_actor_id INTEGER,
+    via_repost_rkey BIGINT,
+    status repost_status NOT NULL DEFAULT 'complete',
+    PRIMARY KEY (actor_id, rkey),
+    UNIQUE (actor_id, post_actor_id, post_rkey)
 );
 
-CREATE INDEX idx_reposts_actor ON reposts(actor_id);
-CREATE INDEX idx_reposts_post ON reposts(post_id);
+CREATE INDEX idx_reposts_post_actor_id_post_rkey
+    ON reposts(post_actor_id, post_rkey);
 CREATE INDEX idx_reposts_rkey ON reposts(rkey);
-CREATE INDEX idx_reposts_via_post ON reposts(via_post_id);
+CREATE INDEX idx_reposts_via_repost
+    ON reposts(via_repost_actor_id, via_repost_rkey)
+    WHERE via_repost_actor_id IS NOT NULL;
+CREATE INDEX idx_reposts_status ON reposts(status) WHERE status != 'complete';
+
+COMMENT ON TABLE reposts IS 'Reposts (quote-less shares): uses natural keys (actor_id, rkey), tracks via_repost for chain context';
+COMMENT ON COLUMN reposts.status IS 'Repost lifecycle state: complete (normal), stub (placeholder, needs fetch)';
+COMMENT ON COLUMN reposts.via_repost_actor_id IS 'If reposted via another repost, source repost author';
+COMMENT ON COLUMN reposts.via_repost_rkey IS 'If reposted via another repost, source repost TID';
+
+COMMENT ON INDEX idx_reposts_post_actor_id_post_rkey IS 'Find all reposts of a specific post';
+COMMENT ON INDEX idx_reposts_rkey IS 'Find reposts by timestamp (TID-based)';
+COMMENT ON INDEX idx_reposts_via_repost IS 'Find reposts that came via another repost (partial: only WHERE via_repost exists)';
+COMMENT ON INDEX idx_reposts_status IS 'Find stub reposts needing fetch (partial: only non-complete)';
 
 -- =============================================================================
--- BOOKMARKS
+-- BOOKMARKS (off-protocol, posts only)
 -- =============================================================================
 
--- Bookmarks (off-protocol, supports multiple entity types)
 CREATE TABLE bookmarks (
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
-    subject_type bookmark_subject_type NOT NULL,
-    subject_id BIGINT,
+    post_actor_id INTEGER NOT NULL,
+    post_rkey BIGINT NOT NULL,
     PRIMARY KEY (actor_id, rkey)
 );
 
-CREATE INDEX idx_bookmarks_subject ON bookmarks(subject_type, subject_id) WHERE subject_id IS NOT NULL;
+CREATE INDEX idx_bookmarks_post ON bookmarks(post_actor_id, post_rkey);
+
+-- Viewer-first composite index for efficient viewer state lookups
+CREATE INDEX idx_bookmarks_viewer_post
+    ON bookmarks (actor_id, post_actor_id, post_rkey);
+
+COMMENT ON TABLE bookmarks IS 'Bookmarks (off-protocol, posts only): saved posts per user. Uses natural keys for post references.';
+COMMENT ON INDEX idx_bookmarks_post IS 'Find all bookmarks of a specific post';
+COMMENT ON INDEX idx_bookmarks_viewer_post IS 'Viewer state: check if viewer bookmarked a post (viewer-first composite index)';
+
+-- =============================================================================
+-- VIEWER STATE OPTIMIZATIONS
+-- =============================================================================
+
+-- Viewer-first composite indexes for efficient viewer state lookups
+-- Pattern: (viewer_actor_id, post_actor_id, post_rkey)
+-- This allows PostgreSQL to seek directly to the viewer's records, then filter by post
+
+CREATE INDEX idx_post_likes_viewer_post
+    ON post_likes (actor_id, post_actor_id, post_rkey);
+
+COMMENT ON INDEX idx_post_likes_viewer_post IS 'Viewer state: check if viewer liked a post (viewer-first composite index)';
+
+CREATE INDEX idx_reposts_viewer_post
+    ON reposts (actor_id, post_actor_id, post_rkey);
+
+COMMENT ON INDEX idx_reposts_viewer_post IS 'Viewer state: check if viewer reposted a post (viewer-first composite index)';
diff --git a/migrations/2025-11-01-114843_thread_moderation/up.sql b/migrations/2025-11-01-114843_thread_moderation/up.sql
index e038714d..dcf50ebe 100644
--- a/migrations/2025-11-01-114843_thread_moderation/up.sql
+++ b/migrations/2025-11-01-114843_thread_moderation/up.sql
@@ -28,29 +28,30 @@ CREATE TYPE postgate_rule AS ENUM (
 -- =============================================================================
 
 CREATE TABLE threadgates (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
     cid BYTEA NOT NULL,
-    post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE,
+    post_actor_id INTEGER,
+    post_rkey BIGINT,
     allow threadgate_rule[],
-    UNIQUE (actor_id, rkey)
+    PRIMARY KEY (actor_id, rkey)
 );
 
-CREATE INDEX idx_threadgates_post ON threadgates(post_id);
-
 -- Threadgate Hidden Replies
 CREATE TABLE threadgate_hidden_replies (
-    threadgate_id BIGINT NOT NULL REFERENCES threadgates(id) ON DELETE CASCADE,
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    PRIMARY KEY (threadgate_id, post_id)
+    post_actor_id INTEGER NOT NULL,
+    post_rkey BIGINT NOT NULL,
+    hidden_post_actor_id INTEGER NOT NULL,
+    hidden_post_rkey BIGINT NOT NULL,
+    PRIMARY KEY (post_actor_id, post_rkey, hidden_post_actor_id, hidden_post_rkey)
 );
 
 -- Threadgate Allowed Lists
 CREATE TABLE threadgate_allowed_lists (
-    threadgate_id BIGINT NOT NULL REFERENCES threadgates(id) ON DELETE CASCADE,
+    post_actor_id INTEGER NOT NULL,
+    post_rkey BIGINT NOT NULL,
     list_id BIGINT NOT NULL, -- FK added in lists migration
-    PRIMARY KEY (threadgate_id, list_id)
+    PRIMARY KEY (post_actor_id, post_rkey, list_id)
 );
 
 -- =============================================================================
@@ -58,51 +59,56 @@ CREATE TABLE threadgate_allowed_lists (
 -- =============================================================================
 
 CREATE TABLE postgates (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
     cid BYTEA NOT NULL,
-    post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE,
+    post_actor_id INTEGER,
+    post_rkey BIGINT,
     rules postgate_rule[] NOT NULL DEFAULT '{}',
-    UNIQUE (actor_id, rkey)
+    PRIMARY KEY (actor_id, rkey)
 );
 
-CREATE INDEX idx_postgates_post ON postgates(post_id);
-
 -- Postgate Detached Embeds
 CREATE TABLE postgate_detached (
-    postgate_id BIGINT NOT NULL REFERENCES postgates(id) ON DELETE CASCADE,
-    detached_post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    PRIMARY KEY (postgate_id, detached_post_id)
+    post_actor_id INTEGER NOT NULL,
+    post_rkey BIGINT NOT NULL,
+    detached_post_actor_id INTEGER NOT NULL,
+    detached_post_rkey BIGINT NOT NULL,
+    PRIMARY KEY (post_actor_id, post_rkey, detached_post_actor_id, detached_post_rkey)
 );
 
 -- =============================================================================
 -- POSTGATE MAINTENANCE FUNCTION
 -- =============================================================================
 
--- Updates the 'detached' flag on existing post_embed_record entries when a postgate changes
+-- Updates the 'detached' flag on posts.record_detached when a postgate changes
 CREATE OR REPLACE FUNCTION maintain_postgates(
     post_uri TEXT,
     detached_uris TEXT[],
     disable_effective TIMESTAMP
 ) RETURNS BIGINT AS $$
 DECLARE
-    target_post_id BIGINT;
+    target_actor_id INTEGER;
+    target_rkey BIGINT;
     rows_updated BIGINT;
 BEGIN
-    -- Look up the post_id from the post URI
-    SELECT p.id INTO target_post_id
-    FROM posts p
-    INNER JOIN actors a ON p.actor_id = a.id
-    WHERE 'at://' || a.did || '/app.bsky.feed.post/' || i64_to_tid(p.rkey) = post_uri;
-
-    IF target_post_id IS NULL THEN
+    -- Look up the post from the post URI
+    SELECT a.id, tid_to_i64(parts.rkey)
+    INTO target_actor_id, target_rkey
+    FROM (
+        SELECT
+            SPLIT_PART(SUBSTRING(post_uri FROM 6), '/', 1) as did,
+            SPLIT_PART(SUBSTRING(post_uri FROM 6), '/', 3) as rkey
+    ) parts
+    INNER JOIN actors a ON a.did = parts.did;
+
+    IF target_actor_id IS NULL THEN
         RETURN 0;
     END IF;
 
-    -- Update post_embed_record entries to mark quotes as detached
-    WITH detached_post_ids AS (
-        SELECT p.id
+    -- Update posts.record_detached for quotes of the target post
+    WITH detached_posts AS (
+        SELECT p.actor_id, p.rkey
         FROM UNNEST(detached_uris) AS uri
         CROSS JOIN LATERAL (
             SELECT
@@ -111,30 +117,22 @@ BEGIN
         ) parts
         INNER JOIN actors a ON a.did = parts.did
         INNER JOIN posts p ON p.actor_id = a.id AND p.rkey = tid_to_i64(parts.rkey)
-    ),
-    posts_to_update AS (
-        SELECT per.post_id
-        FROM post_embed_record per
-        WHERE per.embedded_post_id = target_post_id
     )
-    UPDATE post_embed_record per
-    SET detached = (
-        per.post_id IN (SELECT id FROM detached_post_ids)
+    UPDATE posts
+    SET record_detached = (
+        (actor_id, rkey) IN (SELECT actor_id, rkey FROM detached_posts)
         OR
-        (disable_effective IS NOT NULL AND (
-            SELECT tid_timestamp(p.rkey)
-            FROM posts p
-            WHERE p.id = per.post_id
-        ) > disable_effective)
+        (disable_effective IS NOT NULL AND tid_timestamp(rkey) > disable_effective)
     )
-    WHERE per.embedded_post_id = target_post_id;
+    WHERE embedded_post_actor_id = target_actor_id
+      AND embedded_post_rkey = target_rkey;
 
     GET DIAGNOSTICS rows_updated = ROW_COUNT;
     RETURN rows_updated;
 END;
 $$ LANGUAGE plpgsql;
 
-COMMENT ON FUNCTION maintain_postgates IS 'Updates detached flag on post_embed_record when postgates change';
+COMMENT ON FUNCTION maintain_postgates IS 'Updates record_detached flag on posts table when postgates change';
 
 -- =============================================================================
 -- THREAD MUTES
@@ -143,7 +141,8 @@ COMMENT ON FUNCTION maintain_postgates IS 'Updates detached flag on post_embed_r
 -- Thread Mutes (off-protocol)
 CREATE TABLE thread_mutes (
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    root_post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
+    root_post_actor_id INTEGER NOT NULL,
+    root_post_rkey BIGINT NOT NULL,
     created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    PRIMARY KEY (actor_id, root_post_id)
+    PRIMARY KEY (actor_id, root_post_actor_id, root_post_rkey)
 );
diff --git a/migrations/2025-11-01-114844_lists/up.sql b/migrations/2025-11-01-114844_lists/up.sql
index 5265d5ae..b81728ec 100644
--- a/migrations/2025-11-01-114844_lists/up.sql
+++ b/migrations/2025-11-01-114844_lists/up.sql
@@ -49,13 +49,12 @@ CREATE INDEX idx_lists_type ON lists(list_type) WHERE list_type IS NOT NULL;
 -- =============================================================================
 
 CREATE TABLE list_items (
-    id BIGSERIAL PRIMARY KEY,
     actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
     rkey INT8 NOT NULL,
     cid BYTEA NOT NULL,
     list_id BIGINT NOT NULL REFERENCES lists(id) ON DELETE CASCADE,
     subject_actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    UNIQUE (actor_id, rkey)
+    PRIMARY KEY (actor_id, rkey)
 );
 
 CREATE INDEX idx_list_items_list ON list_items(list_id);
diff --git a/migrations/2025-11-01-114847_queues_and_extensions/up.sql b/migrations/2025-11-01-114847_queues_and_extensions/up.sql
index d243cfa8..5725484c 100644
--- a/migrations/2025-11-01-114847_queues_and_extensions/up.sql
+++ b/migrations/2025-11-01-114847_queues_and_extensions/up.sql
@@ -35,11 +35,13 @@ CREATE EXTENSION IF NOT EXISTS postgis;
 -- - Multi-hop neighborhood queries (N-degree connections)
 CREATE EXTENSION IF NOT EXISTS pgrouting;
 
--- Note: timescaledb and pgvector are available via flake.nix but not enabled yet
---
--- timescaledb: Time-series optimization for notifications and stats tables
--- Deferred until we design schema migration strategy (hypertables, partitioning)
--- Future use: trending posts, notification retention policies, time-bucketing
+-- TimescaleDB: Time-series optimization for hypertables
+-- Provides efficient compression and querying for time-series data (posts, post_likes)
+-- Used for: compressed storage, chunk-based partitioning, chunk skipping
+-- NOTE: TimescaleDB extension is installed in the posts migration (runs earlier)
+-- CREATE EXTENSION IF NOT EXISTS timescaledb CASCADE;
+
+-- Note: pgvector is available via flake.nix but not enabled yet
 --
 -- pgvector: Vector similarity search for semantic/ML-based features
 -- Deferred until concrete use case emerges (embeddings, semantic search)
@@ -147,3 +149,24 @@ CREATE INDEX idx_constellation_status_created
     WHERE status IN ('pending', 'processing');
 
 COMMENT ON TABLE constellation_enrichment_queue IS 'Queue for asynchronous Constellation API enrichment';
+
+-- =============================================================================
+-- DIESEL SCHEMA INFERENCE TABLE
+-- =============================================================================
+
+-- Dummy table to force Diesel to generate SQL type definitions
+--
+-- This table is never used by the application, but exists solely to ensure
+-- that Diesel generates type definitions for enums and composite types that are
+-- only used inside other composite types.
+--
+-- Without this table, Diesel wouldn't generate these types in schema.rs since
+-- they're not directly used as column types in any real tables.
+
+CREATE TABLE _diesel_schema_inference (
+    id INTEGER PRIMARY KEY,
+    video_mime_type video_mime_type NOT NULL,
+    caption_mime_type caption_mime_type NOT NULL,
+    facet_type facet_type NOT NULL,
+    video_caption post_video_caption NOT NULL
+);
diff --git a/migrations/2025-11-01-114848_aggregate_stats/up.sql b/migrations/2025-11-01-114848_aggregate_stats/up.sql
index 34a11ea6..2230a828 100644
--- a/migrations/2025-11-01-114848_aggregate_stats/up.sql
+++ b/migrations/2025-11-01-114848_aggregate_stats/up.sql
@@ -35,14 +35,22 @@ CREATE TYPE actor_stat_type AS ENUM (
 -- Post aggregate statistics
 -- Stores engagement metrics for posts with automatic cleanup
 CREATE TABLE post_aggregate_stats (
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
+    actor_id INTEGER NOT NULL,
+    rkey BIGINT NOT NULL,
     stat_type post_stat_type NOT NULL,
     value INTEGER NOT NULL,
     updated_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT now(),
 
-    PRIMARY KEY (post_id, stat_type)
+    PRIMARY KEY (actor_id, rkey, stat_type)
 );
 
+COMMENT ON TABLE post_aggregate_stats IS 'Post aggregate statistics using natural keys (actor_id, rkey)';
+COMMENT ON COLUMN post_aggregate_stats.actor_id IS 'Actor who created the post (natural key part 1)';
+COMMENT ON COLUMN post_aggregate_stats.rkey IS 'Record key as i64 (TID converted) - natural key part 2';
+COMMENT ON COLUMN post_aggregate_stats.stat_type IS 'Type of statistic (like, reply, repost, quote)';
+COMMENT ON COLUMN post_aggregate_stats.value IS 'Current count value';
+COMMENT ON COLUMN post_aggregate_stats.updated_at IS 'Last time value changed (NOT last write time)';
+
 -- Actor aggregate statistics
 -- Stores profile metrics for actors with automatic cleanup
 CREATE TABLE actor_aggregate_stats (
@@ -60,7 +68,13 @@ CREATE TABLE actor_aggregate_stats (
 
 -- Index for finding hot posts (recently updated engagement)
 CREATE INDEX idx_post_aggregate_stats_updated_at
-    ON post_aggregate_stats (updated_at DESC);
+    ON post_aggregate_stats(updated_at DESC)
+    WHERE value > 0;
+
+-- Index for hot content queries (recently updated high-value stats)
+CREATE INDEX idx_post_aggregate_stats_hot_content
+    ON post_aggregate_stats(stat_type, updated_at DESC, value DESC)
+    WHERE value > 10;
 
 -- Index for finding hot actors (recently updated profiles)
 CREATE INDEX idx_actor_aggregate_stats_updated_at
diff --git a/migrations/2025-11-15-171545_add_gif_to_video_mime_type/down.sql b/migrations/2025-11-15-171545_add_gif_to_video_mime_type/down.sql
deleted file mode 100644
index 8fad3299..00000000
--- a/migrations/2025-11-15-171545_add_gif_to_video_mime_type/down.sql
+++ /dev/null
@@ -1,11 +0,0 @@
--- Note: PostgreSQL does not support removing enum values without dropping and recreating the enum.
--- Since these mime types are backward compatible (the old values still work), and removing them
--- would require dropping columns/constraints/etc, we leave the enum values in place.
---
--- If you absolutely need to revert, you would need to:
--- 1. Drop all columns using these enum types
--- 2. Drop the enum types
--- 3. Recreate the enum types with original values
--- 4. Recreate all columns
---
--- This is not implemented as it would be destructive and the added values don't break anything.
diff --git a/migrations/2025-11-15-171545_add_gif_to_video_mime_type/up.sql b/migrations/2025-11-15-171545_add_gif_to_video_mime_type/up.sql
deleted file mode 100644
index 8f039833..00000000
--- a/migrations/2025-11-15-171545_add_gif_to_video_mime_type/up.sql
+++ /dev/null
@@ -1,20 +0,0 @@
--- Expand mime type enums to be more permissive and prevent future errors
--- This adds commonly used formats that might appear in the wild
-
--- Video mime types (GIFs are treated as video embeds in Bluesky)
-ALTER TYPE video_mime_type ADD VALUE IF NOT EXISTS 'image/gif';
-ALTER TYPE video_mime_type ADD VALUE IF NOT EXISTS 'video/x-m4v';
-ALTER TYPE video_mime_type ADD VALUE IF NOT EXISTS 'video/3gpp';
-ALTER TYPE video_mime_type ADD VALUE IF NOT EXISTS 'video/x-msvideo';
-ALTER TYPE video_mime_type ADD VALUE IF NOT EXISTS 'video/x-matroska';
-ALTER TYPE video_mime_type ADD VALUE IF NOT EXISTS 'video/ogg';
-
--- Image mime types (add less common but valid formats)
-ALTER TYPE image_mime_type ADD VALUE IF NOT EXISTS 'image/x-icon';
-ALTER TYPE image_mime_type ADD VALUE IF NOT EXISTS 'image/vnd.microsoft.icon';
-
--- Caption mime types (add common subtitle formats)
-ALTER TYPE caption_mime_type ADD VALUE IF NOT EXISTS 'text/srt';
-ALTER TYPE caption_mime_type ADD VALUE IF NOT EXISTS 'application/x-subrip';
-ALTER TYPE caption_mime_type ADD VALUE IF NOT EXISTS 'text/x-ssa';
-ALTER TYPE caption_mime_type ADD VALUE IF NOT EXISTS 'text/x-ass';
diff --git a/migrations/2025-11-15-203456_drop_likes_id_column/down.sql b/migrations/2025-11-15-203456_drop_likes_id_column/down.sql
deleted file mode 100644
index cb68d77c..00000000
--- a/migrations/2025-11-15-203456_drop_likes_id_column/down.sql
+++ /dev/null
@@ -1,6 +0,0 @@
--- Rollback: restore BIGSERIAL id column as primary key
--- Note: id values will not be preserved on rollback (new sequence starts)
-
-ALTER TABLE likes DROP CONSTRAINT likes_pkey;
-ALTER TABLE likes ADD COLUMN id BIGSERIAL;
-ALTER TABLE likes ADD PRIMARY KEY (id);
diff --git a/migrations/2025-11-15-203456_drop_likes_id_column/up.sql b/migrations/2025-11-15-203456_drop_likes_id_column/up.sql
deleted file mode 100644
index 934a42c4..00000000
--- a/migrations/2025-11-15-203456_drop_likes_id_column/up.sql
+++ /dev/null
@@ -1,7 +0,0 @@
--- Drop the BIGSERIAL id column and use composite primary key (actor_id, rkey)
--- This saves 8 bytes per row + primary key index overhead
--- Safe because: no FK references to likes(id) exist in the codebase
-
-ALTER TABLE likes DROP CONSTRAINT likes_pkey;
-ALTER TABLE likes ADD PRIMARY KEY (actor_id, rkey);
-ALTER TABLE likes DROP COLUMN id;
diff --git a/migrations/2025-11-15-203642_drop_list_items_id_column/down.sql b/migrations/2025-11-15-203642_drop_list_items_id_column/down.sql
deleted file mode 100644
index a85da302..00000000
--- a/migrations/2025-11-15-203642_drop_list_items_id_column/down.sql
+++ /dev/null
@@ -1,6 +0,0 @@
--- Rollback: restore BIGSERIAL id column as primary key
--- Note: id values will not be preserved on rollback (new sequence starts)
-
-ALTER TABLE list_items DROP CONSTRAINT list_items_pkey;
-ALTER TABLE list_items ADD COLUMN id BIGSERIAL;
-ALTER TABLE list_items ADD PRIMARY KEY (id);
diff --git a/migrations/2025-11-15-203642_drop_list_items_id_column/up.sql b/migrations/2025-11-15-203642_drop_list_items_id_column/up.sql
deleted file mode 100644
index 40000446..00000000
--- a/migrations/2025-11-15-203642_drop_list_items_id_column/up.sql
+++ /dev/null
@@ -1,7 +0,0 @@
--- Drop the BIGSERIAL id column and use composite primary key (actor_id, rkey)
--- This saves 8 bytes per row + primary key index overhead
--- Safe because: no FK references to list_items(id) exist in the codebase
-
-ALTER TABLE list_items DROP CONSTRAINT list_items_pkey;
-ALTER TABLE list_items ADD PRIMARY KEY (actor_id, rkey);
-ALTER TABLE list_items DROP COLUMN id;
diff --git a/migrations/2025-11-15-203734_drop_follows_id_column/down.sql b/migrations/2025-11-15-203734_drop_follows_id_column/down.sql
deleted file mode 100644
index 94bbbe3e..00000000
--- a/migrations/2025-11-15-203734_drop_follows_id_column/down.sql
+++ /dev/null
@@ -1,6 +0,0 @@
--- Rollback: restore BIGSERIAL id column as primary key
--- Note: id values will not be preserved on rollback (new sequence starts)
-
-ALTER TABLE follows DROP CONSTRAINT follows_pkey;
-ALTER TABLE follows ADD COLUMN id BIGSERIAL;
-ALTER TABLE follows ADD PRIMARY KEY (id);
diff --git a/migrations/2025-11-15-203734_drop_follows_id_column/up.sql b/migrations/2025-11-15-203734_drop_follows_id_column/up.sql
deleted file mode 100644
index 8d3770dd..00000000
--- a/migrations/2025-11-15-203734_drop_follows_id_column/up.sql
+++ /dev/null
@@ -1,7 +0,0 @@
--- Drop the BIGSERIAL id column and use composite primary key (actor_id, rkey)
--- This saves 8 bytes per row + primary key index overhead
--- Safe because: no FK references to follows(id) exist in the codebase
-
-ALTER TABLE follows DROP CONSTRAINT follows_pkey;
-ALTER TABLE follows ADD PRIMARY KEY (actor_id, rkey);
-ALTER TABLE follows DROP COLUMN id;
diff --git a/migrations/2025-11-15-203834_drop_blocks_id_column/down.sql b/migrations/2025-11-15-203834_drop_blocks_id_column/down.sql
deleted file mode 100644
index 49a0ad97..00000000
--- a/migrations/2025-11-15-203834_drop_blocks_id_column/down.sql
+++ /dev/null
@@ -1,6 +0,0 @@
--- Rollback: restore BIGSERIAL id column as primary key
--- Note: id values will not be preserved on rollback (new sequence starts)
-
-ALTER TABLE blocks DROP CONSTRAINT blocks_pkey;
-ALTER TABLE blocks ADD COLUMN id BIGSERIAL;
-ALTER TABLE blocks ADD PRIMARY KEY (id);
diff --git a/migrations/2025-11-15-203834_drop_blocks_id_column/up.sql b/migrations/2025-11-15-203834_drop_blocks_id_column/up.sql
deleted file mode 100644
index 135c8e79..00000000
--- a/migrations/2025-11-15-203834_drop_blocks_id_column/up.sql
+++ /dev/null
@@ -1,7 +0,0 @@
--- Drop the BIGSERIAL id column and use composite primary key (actor_id, rkey)
--- This saves 8 bytes per row + primary key index overhead
--- Safe because: no FK references to blocks(id) exist in the codebase
-
-ALTER TABLE blocks DROP CONSTRAINT blocks_pkey;
-ALTER TABLE blocks ADD PRIMARY KEY (actor_id, rkey);
-ALTER TABLE blocks DROP COLUMN id;
diff --git a/migrations/2025-11-15-205803_add_via_repost_id_to_likes/down.sql b/migrations/2025-11-15-205803_add_via_repost_id_to_likes/down.sql
deleted file mode 100644
index a2cb7b65..00000000
--- a/migrations/2025-11-15-205803_add_via_repost_id_to_likes/down.sql
+++ /dev/null
@@ -1,4 +0,0 @@
--- Rollback: remove via_repost_id column from likes
-
-DROP INDEX IF EXISTS idx_likes_via_repost;
-ALTER TABLE likes DROP COLUMN via_repost_id;
diff --git a/migrations/2025-11-15-205803_add_via_repost_id_to_likes/up.sql b/migrations/2025-11-15-205803_add_via_repost_id_to_likes/up.sql
deleted file mode 100644
index 366749e2..00000000
--- a/migrations/2025-11-15-205803_add_via_repost_id_to_likes/up.sql
+++ /dev/null
@@ -1,8 +0,0 @@
--- Add via_repost_id to likes table to properly track "liked via repost" context
--- When a user likes a post through viewing a repost, this stores the repost record
--- This enables notifications like "Alice liked your repost of Bob's post"
-
-ALTER TABLE likes ADD COLUMN via_repost_id BIGINT REFERENCES reposts(id) ON DELETE CASCADE;
-
--- Partial index only for rows that have via_repost_id set (sparse index)
-CREATE INDEX idx_likes_via_repost ON likes(via_repost_id) WHERE via_repost_id IS NOT NULL;
diff --git a/migrations/2025-11-15-205826_add_via_repost_id_to_reposts/down.sql b/migrations/2025-11-15-205826_add_via_repost_id_to_reposts/down.sql
deleted file mode 100644
index 74dc89ba..00000000
--- a/migrations/2025-11-15-205826_add_via_repost_id_to_reposts/down.sql
+++ /dev/null
@@ -1,4 +0,0 @@
--- Rollback: remove via_repost_id column from reposts
-
-DROP INDEX IF EXISTS idx_reposts_via_repost;
-ALTER TABLE reposts DROP COLUMN via_repost_id;
diff --git a/migrations/2025-11-15-205826_add_via_repost_id_to_reposts/up.sql b/migrations/2025-11-15-205826_add_via_repost_id_to_reposts/up.sql
deleted file mode 100644
index 6e1c11e9..00000000
--- a/migrations/2025-11-15-205826_add_via_repost_id_to_reposts/up.sql
+++ /dev/null
@@ -1,8 +0,0 @@
--- Add via_repost_id to reposts table to track "repost via repost" context
--- When a user reposts a post through viewing another repost, this stores that context
--- This enables tracking repost chains and proper notification attribution
-
-ALTER TABLE reposts ADD COLUMN via_repost_id BIGINT REFERENCES reposts(id) ON DELETE CASCADE;
-
--- Partial index only for rows that have via_repost_id set (sparse index)
-CREATE INDEX idx_reposts_via_repost ON reposts(via_repost_id) WHERE via_repost_id IS NOT NULL;
diff --git a/migrations/2025-11-15-210615_drop_via_post_id_from_likes/down.sql b/migrations/2025-11-15-210615_drop_via_post_id_from_likes/down.sql
deleted file mode 100644
index 13cdc6a0..00000000
--- a/migrations/2025-11-15-210615_drop_via_post_id_from_likes/down.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Rollback: restore via_post_id column (though it was never used)
-
-ALTER TABLE likes ADD COLUMN via_post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE;
diff --git a/migrations/2025-11-15-210615_drop_via_post_id_from_likes/up.sql b/migrations/2025-11-15-210615_drop_via_post_id_from_likes/up.sql
deleted file mode 100644
index 6c7849ce..00000000
--- a/migrations/2025-11-15-210615_drop_via_post_id_from_likes/up.sql
+++ /dev/null
@@ -1,5 +0,0 @@
--- Drop legacy via_post_id column from likes table
--- This column was never properly used - via_repost_id is the correct column
--- for tracking "liked via repost" context
-
-ALTER TABLE likes DROP COLUMN via_post_id;
diff --git a/migrations/2025-11-15-210633_drop_via_post_id_from_reposts/down.sql b/migrations/2025-11-15-210633_drop_via_post_id_from_reposts/down.sql
deleted file mode 100644
index e53eef5f..00000000
--- a/migrations/2025-11-15-210633_drop_via_post_id_from_reposts/down.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Rollback: restore via_post_id column (though it was never used)
-
-ALTER TABLE reposts ADD COLUMN via_post_id BIGINT REFERENCES posts(id) ON DELETE CASCADE;
diff --git a/migrations/2025-11-15-210633_drop_via_post_id_from_reposts/up.sql b/migrations/2025-11-15-210633_drop_via_post_id_from_reposts/up.sql
deleted file mode 100644
index 33e34105..00000000
--- a/migrations/2025-11-15-210633_drop_via_post_id_from_reposts/up.sql
+++ /dev/null
@@ -1,5 +0,0 @@
--- Drop legacy via_post_id column from reposts table
--- This column was never properly used - via_repost_id is the correct column
--- for tracking "repost via repost" context
-
-ALTER TABLE reposts DROP COLUMN via_post_id;
diff --git a/migrations/2025-11-15-213700_add_repost_status/down.sql b/migrations/2025-11-15-213700_add_repost_status/down.sql
deleted file mode 100644
index 58533084..00000000
--- a/migrations/2025-11-15-213700_add_repost_status/down.sql
+++ /dev/null
@@ -1,10 +0,0 @@
--- Rollback: Remove status column and repost_status enum from reposts table
-
--- Drop index
-DROP INDEX IF EXISTS idx_reposts_status;
-
--- Drop status column
-ALTER TABLE reposts DROP COLUMN status;
-
--- Drop enum type
-DROP TYPE repost_status;
diff --git a/migrations/2025-11-15-213700_add_repost_status/up.sql b/migrations/2025-11-15-213700_add_repost_status/up.sql
deleted file mode 100644
index 04e4a81e..00000000
--- a/migrations/2025-11-15-213700_add_repost_status/up.sql
+++ /dev/null
@@ -1,19 +0,0 @@
--- Add status column to reposts table for stub tracking
---
--- This enables repost stub creation for eventual consistency handling.
--- When likes/reposts reference reposts that haven't arrived yet, we create
--- stub reposts that can be upgraded later when the actual repost arrives.
-
--- Create repost_status enum type
-CREATE TYPE repost_status AS ENUM ('complete', 'stub');
-
-COMMENT ON TYPE repost_status IS 'Repost lifecycle state: complete (normal repost), stub (placeholder awaiting fetch)';
-
--- Add status column to reposts table
-ALTER TABLE reposts ADD COLUMN status repost_status NOT NULL DEFAULT 'complete';
-
-COMMENT ON COLUMN reposts.status IS 'Repost lifecycle state: complete (normal), stub (placeholder, needs fetch)';
-
--- Index on stub reposts for fetch queue queries
--- Partial index only indexes non-complete reposts (very small set)
-CREATE INDEX idx_reposts_status ON reposts(status) WHERE status != 'complete';
diff --git a/migrations/2025-11-16-005212_drop_likes_cid/down.sql b/migrations/2025-11-16-005212_drop_likes_cid/down.sql
deleted file mode 100644
index 619a53bd..00000000
--- a/migrations/2025-11-16-005212_drop_likes_cid/down.sql
+++ /dev/null
@@ -1,4 +0,0 @@
--- Restore the cid column to likes table
--- Note: This cannot restore the original CID data - column will be populated with empty values
--- Synthetic CIDs will continue to be generated for notifications
-ALTER TABLE likes ADD COLUMN cid BYTEA NOT NULL DEFAULT '';
diff --git a/migrations/2025-11-16-005212_drop_likes_cid/up.sql b/migrations/2025-11-16-005212_drop_likes_cid/up.sql
deleted file mode 100644
index 452eac20..00000000
--- a/migrations/2025-11-16-005212_drop_likes_cid/up.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Drop the cid column from likes table
--- Like CIDs are now generated synthetically from (actor_id, rkey) instead of being stored
-ALTER TABLE likes DROP COLUMN cid;
diff --git a/migrations/2025-11-16-014100_drop_blocks_cid/down.sql b/migrations/2025-11-16-014100_drop_blocks_cid/down.sql
deleted file mode 100644
index d5770626..00000000
--- a/migrations/2025-11-16-014100_drop_blocks_cid/down.sql
+++ /dev/null
@@ -1,2 +0,0 @@
--- Rollback: Add cid column back
-ALTER TABLE blocks ADD COLUMN cid BYTEA NOT NULL DEFAULT E'\\x0000000000000000000000000000000000000000000000000000000000000000'::bytea;
diff --git a/migrations/2025-11-16-014100_drop_blocks_cid/up.sql b/migrations/2025-11-16-014100_drop_blocks_cid/up.sql
deleted file mode 100644
index 88428d2a..00000000
--- a/migrations/2025-11-16-014100_drop_blocks_cid/up.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Drop the cid column from blocks table
--- Block CIDs are now generated synthetically from (actor_id, rkey)
-ALTER TABLE blocks DROP COLUMN cid;
diff --git a/migrations/2025-11-16-014101_drop_follows_cid/down.sql b/migrations/2025-11-16-014101_drop_follows_cid/down.sql
deleted file mode 100644
index 77e0c1a0..00000000
--- a/migrations/2025-11-16-014101_drop_follows_cid/down.sql
+++ /dev/null
@@ -1,2 +0,0 @@
--- Rollback: Add cid column back
-ALTER TABLE follows ADD COLUMN cid BYTEA NOT NULL DEFAULT E'\\x0000000000000000000000000000000000000000000000000000000000000000'::bytea;
diff --git a/migrations/2025-11-16-014101_drop_follows_cid/up.sql b/migrations/2025-11-16-014101_drop_follows_cid/up.sql
deleted file mode 100644
index 07ea8bf4..00000000
--- a/migrations/2025-11-16-014101_drop_follows_cid/up.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Drop the cid column from follows table
--- Follow CIDs are now generated synthetically from (actor_id, rkey)
-ALTER TABLE follows DROP COLUMN cid;
diff --git a/migrations/2025-11-16-054411_add_tokens_to_posts/down.sql b/migrations/2025-11-16-054411_add_tokens_to_posts/down.sql
deleted file mode 100644
index 16cf07a9..00000000
--- a/migrations/2025-11-16-054411_add_tokens_to_posts/down.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Remove search tokens from posts table
-DROP INDEX IF EXISTS idx_posts_tokens_gin;
-ALTER TABLE posts DROP COLUMN IF EXISTS tokens;
diff --git a/migrations/2025-11-16-054411_add_tokens_to_posts/up.sql b/migrations/2025-11-16-054411_add_tokens_to_posts/up.sql
deleted file mode 100644
index c425bfee..00000000
--- a/migrations/2025-11-16-054411_add_tokens_to_posts/up.sql
+++ /dev/null
@@ -1,7 +0,0 @@
--- Add tokens column to posts table for search
--- Nullable to avoid storage cost for posts not yet indexed
--- GIN index for fast token containment queries
-ALTER TABLE posts ADD COLUMN tokens TEXT[];
-
--- GIN index for fast token array queries (tokens && query_tokens)
-CREATE INDEX idx_posts_tokens_gin ON posts USING GIN (tokens);
diff --git a/migrations/2025-11-16-070138_drop_search_vector/down.sql b/migrations/2025-11-16-070138_drop_search_vector/down.sql
deleted file mode 100644
index eabca2e4..00000000
--- a/migrations/2025-11-16-070138_drop_search_vector/down.sql
+++ /dev/null
@@ -1,10 +0,0 @@
--- Restore search_vector column and index
--- This rollback adds the column back but does NOT populate it
--- You would need to run a backfill to regenerate search vectors
-
--- Add the search_vector column back
-ALTER TABLE posts ADD COLUMN search_vector tsvector;
-
--- Recreate the GIN index
-CREATE INDEX idx_posts_search_vector
-    ON posts USING GIN (search_vector);
diff --git a/migrations/2025-11-16-070138_drop_search_vector/up.sql b/migrations/2025-11-16-070138_drop_search_vector/up.sql
deleted file mode 100644
index d5120648..00000000
--- a/migrations/2025-11-16-070138_drop_search_vector/up.sql
+++ /dev/null
@@ -1,9 +0,0 @@
--- Drop search_vector column and index
--- Token-based search replaces tsvector-based search
--- Storage savings: ~220 GB (198 GB column + 20-30 GB index) for 30M posts
-
--- Drop the GIN index first
-DROP INDEX IF EXISTS idx_posts_search_vector;
-
--- Drop the search_vector column
-ALTER TABLE posts DROP COLUMN IF EXISTS search_vector;
diff --git a/migrations/2025-11-16-203204_denormalize_post_embeds_and_facets/down.sql b/migrations/2025-11-16-203204_denormalize_post_embeds_and_facets/down.sql
deleted file mode 100644
index 7b63d5b0..00000000
--- a/migrations/2025-11-16-203204_denormalize_post_embeds_and_facets/down.sql
+++ /dev/null
@@ -1,30 +0,0 @@
--- Rollback: Remove denormalized columns and composite types from posts table
-
--- Drop indexes
-DROP INDEX IF EXISTS idx_posts_mentions;
-DROP INDEX IF EXISTS idx_posts_embedded_post_id;
-
--- Drop columns
-ALTER TABLE posts DROP COLUMN IF EXISTS mentions;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_8;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_7;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_6;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_5;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_4;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_3;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_2;
-ALTER TABLE posts DROP COLUMN IF EXISTS facet_1;
-ALTER TABLE posts DROP COLUMN IF EXISTS image_4;
-ALTER TABLE posts DROP COLUMN IF EXISTS image_3;
-ALTER TABLE posts DROP COLUMN IF EXISTS image_2;
-ALTER TABLE posts DROP COLUMN IF EXISTS image_1;
-ALTER TABLE posts DROP COLUMN IF EXISTS record_detached;
-ALTER TABLE posts DROP COLUMN IF EXISTS embedded_post_id;
-ALTER TABLE posts DROP COLUMN IF EXISTS video_embed;
-ALTER TABLE posts DROP COLUMN IF EXISTS ext_embed;
-
--- Drop composite types
-DROP TYPE IF EXISTS post_facet_embed;
-DROP TYPE IF EXISTS post_image_embed;
-DROP TYPE IF EXISTS post_video_embed;
-DROP TYPE IF EXISTS post_ext_embed;
diff --git a/migrations/2025-11-16-203204_denormalize_post_embeds_and_facets/up.sql b/migrations/2025-11-16-203204_denormalize_post_embeds_and_facets/up.sql
deleted file mode 100644
index 8571aac7..00000000
--- a/migrations/2025-11-16-203204_denormalize_post_embeds_and_facets/up.sql
+++ /dev/null
@@ -1,102 +0,0 @@
--- Denormalize post embeds and facets into posts table for query performance
---
--- Strategy: Use composite types with NULLable columns for sparse data
--- - External embed: 1 column (post_ext_embed)
--- - Video embed: 1 column (post_video_embed)
--- - Record embed: 2 columns (embedded_post_id, record_detached)
--- - Images: 4 NULLable columns of post_image_embed (Bluesky protocol limit)
--- - Facets: 8 NULLable columns of post_facet_embed (covers 99.3% of posts with facets)
--- - Mentions: integer[] array (Diesel-friendly, truncate to max 8)
---
--- Total columns: 17 (down from 89 separate columns!)
--- Storage impact: NULL columns compress to ~0 bytes in TimescaleDB columnstore
--- Query impact: Eliminates 5 JOINs + 1 extra query in embed loader
-
--- ============================================================================
--- Step 1: Create composite types for embeds
--- ============================================================================
-
--- External link embed (used by 0.8% of posts)
-CREATE TYPE post_ext_embed AS (
-    uri text,
-    title text,
-    description text,
-    thumb_mime_type image_mime_type,
-    thumb_cid bytea
-);
-
--- Video embed (used by 0.16% of posts)
-CREATE TYPE post_video_embed AS (
-    mime_type video_mime_type,
-    cid bytea,
-    alt text,
-    width integer,
-    height integer
-);
-
--- Image embed (used by 1.8% of posts, max 4 per Bluesky protocol)
-CREATE TYPE post_image_embed AS (
-    mime_type image_mime_type,
-    cid bytea,
-    alt text,
-    width integer,
-    height integer
-);
-
--- Facet embed (facet_type, index_start, index_end, link_uri, mention_actor_id, tag)
--- Used by 12.5% of posts, 99.3% have ≤8 facets
-CREATE TYPE post_facet_embed AS (
-    facet_type facet_type,
-    index_start integer,
-    index_end integer,
-    link_uri text,
-    mention_actor_id integer,
-    tag text
-);
-
--- ============================================================================
--- Step 2: Add composite columns to posts table
--- ============================================================================
-
--- External embed (1:0..1 relationship)
-ALTER TABLE posts ADD COLUMN ext_embed post_ext_embed;
-
--- Video embed (1:0..1 relationship)
-ALTER TABLE posts ADD COLUMN video_embed post_video_embed;
-
--- Record embed / quote post (1:0..1 relationship)
-ALTER TABLE posts ADD COLUMN embedded_post_id bigint REFERENCES posts(id) ON DELETE SET NULL;
-ALTER TABLE posts ADD COLUMN record_detached boolean;
-
--- Image embeds (1:0..4 relationship, Bluesky protocol limit)
--- NULL columns compress to ~0 bytes in TimescaleDB
-ALTER TABLE posts ADD COLUMN image_1 post_image_embed;
-ALTER TABLE posts ADD COLUMN image_2 post_image_embed;
-ALTER TABLE posts ADD COLUMN image_3 post_image_embed;
-ALTER TABLE posts ADD COLUMN image_4 post_image_embed;
-
--- Facet embeds (1:0..8 relationship, covers 99.3% of posts with facets)
--- Spam posts with 100+ facets will be truncated to 8
-ALTER TABLE posts ADD COLUMN facet_1 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_2 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_3 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_4 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_5 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_6 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_7 post_facet_embed;
-ALTER TABLE posts ADD COLUMN facet_8 post_facet_embed;
-
--- Mentions (1:many relationship, use array)
--- Array of actor_ids, truncate to max 8 (covers 99.5% of posts with mentions)
--- Arrays are Diesel-friendly (no custom FromSql/ToSql needed)
-ALTER TABLE posts ADD COLUMN mentions integer[];
-
--- ============================================================================
--- Step 3: Create indexes for common query patterns
--- ============================================================================
-
--- Index for quote posts (embedded_post_id lookups)
-CREATE INDEX idx_posts_embedded_post_id ON posts(embedded_post_id) WHERE embedded_post_id IS NOT NULL;
-
--- GIN index for mention searches (find all posts mentioning an actor)
-CREATE INDEX idx_posts_mentions ON posts USING GIN(mentions) WHERE mentions IS NOT NULL;
diff --git a/migrations/2025-11-16-232532_drop_old_normalized_tables/down.sql b/migrations/2025-11-16-232532_drop_old_normalized_tables/down.sql
deleted file mode 100644
index 8741cf85..00000000
--- a/migrations/2025-11-16-232532_drop_old_normalized_tables/down.sql
+++ /dev/null
@@ -1,9 +0,0 @@
--- This migration cannot be reverted because it drops tables that are no longer used
--- The data has been migrated to denormalized composite type columns in the posts table
--- If you need to revert, you would need to:
--- 1. Recreate the old table schemas (see previous migrations)
--- 2. Migrate data back from composite columns to normalized tables
--- 3. Update application code to use old normalized tables
-
--- Placeholder to satisfy diesel migration requirements
-SELECT 1;
diff --git a/migrations/2025-11-16-232532_drop_old_normalized_tables/up.sql b/migrations/2025-11-16-232532_drop_old_normalized_tables/up.sql
deleted file mode 100644
index 57c3aa3a..00000000
--- a/migrations/2025-11-16-232532_drop_old_normalized_tables/up.sql
+++ /dev/null
@@ -1,15 +0,0 @@
--- Drop old normalized tables that have been replaced by denormalized composite type columns
--- See .claude/plans/1114-*.md for context on composite type denormalization
-
--- Drop old post embed tables (replaced by composite columns in posts table)
-DROP TABLE IF EXISTS post_embed_video_captions CASCADE;
-DROP TABLE IF EXISTS post_embed_video CASCADE;
-DROP TABLE IF EXISTS post_embed_record CASCADE;  -- Quote posts use embedded_post_id + record_detached columns
-DROP TABLE IF EXISTS post_embed_images CASCADE;
-DROP TABLE IF EXISTS post_embed_ext CASCADE;
-
--- Drop old post facets table (replaced by facet_1..facet_8 composite columns)
-DROP TABLE IF EXISTS post_facets CASCADE;
-
--- Drop old post mentions table (replaced by mention_actor_ids array column)
-DROP TABLE IF EXISTS post_mentions CASCADE;
diff --git a/migrations/2025-11-16-233305_add_video_captions/down.sql b/migrations/2025-11-16-233305_add_video_captions/down.sql
deleted file mode 100644
index 5e4e78e2..00000000
--- a/migrations/2025-11-16-233305_add_video_captions/down.sql
+++ /dev/null
@@ -1,21 +0,0 @@
--- Revert video captions support
--- Restore original post_video_embed without captions field
-
--- Step 1: Drop video_embed column
-ALTER TABLE posts DROP COLUMN IF EXISTS video_embed;
-
--- Step 2: Drop composite types
-DROP TYPE IF EXISTS post_video_embed;
-DROP TYPE IF EXISTS post_video_caption;
-
--- Step 3: Recreate original post_video_embed without captions
-CREATE TYPE post_video_embed AS (
-    mime_type video_mime_type,
-    cid bytea,
-    alt text,
-    width integer,
-    height integer
-);
-
--- Step 4: Re-add video_embed column
-ALTER TABLE posts ADD COLUMN video_embed post_video_embed;
diff --git a/migrations/2025-11-16-233305_add_video_captions/up.sql b/migrations/2025-11-16-233305_add_video_captions/up.sql
deleted file mode 100644
index e1647260..00000000
--- a/migrations/2025-11-16-233305_add_video_captions/up.sql
+++ /dev/null
@@ -1,36 +0,0 @@
--- Add video captions support to post_video_embed composite type
---
--- Video captions (subtitles) are specified in AT Protocol via app.bsky.embed.video
--- Each caption has: lang (ISO 639-1 code), file (Blob with mime_type, cid, size)
--- Most videos have 0-3 caption tracks (e.g., English, Spanish, French)
---
--- Captions are embedded in the video composite since they only exist with videos
--- This avoids adding extra columns to posts table (already at 31 columns)
-
--- Step 1: Create composite type for individual video caption
-CREATE TYPE post_video_caption AS (
-    lang language_code,
-    mime_type caption_mime_type,
-    cid bytea
-);
-
--- Step 2: Drop existing video_embed column (must drop to alter composite type)
-ALTER TABLE posts DROP COLUMN video_embed;
-
--- Step 3: Drop old composite type
-DROP TYPE post_video_embed;
-
--- Step 4: Recreate post_video_embed with 3 optional caption fields
-CREATE TYPE post_video_embed AS (
-    mime_type video_mime_type,
-    cid bytea,
-    alt text,
-    width integer,
-    height integer,
-    caption_1 post_video_caption,  -- First caption track (e.g., English)
-    caption_2 post_video_caption,  -- Second caption track (e.g., Spanish)
-    caption_3 post_video_caption   -- Third caption track (e.g., French)
-);
-
--- Step 5: Re-add video_embed column with new type
-ALTER TABLE posts ADD COLUMN video_embed post_video_embed;
diff --git a/migrations/2025-11-17-001423_split_likes_table/down.sql b/migrations/2025-11-17-001423_split_likes_table/down.sql
deleted file mode 100644
index 85d5005c..00000000
--- a/migrations/2025-11-17-001423_split_likes_table/down.sql
+++ /dev/null
@@ -1,41 +0,0 @@
--- =============================================================================
--- ROLLBACK: Recreate unified likes table
--- =============================================================================
-
--- Recreate enum type
-CREATE TYPE like_subject_type AS ENUM ('post', 'feedgen', 'labeler');
-
--- Recreate unified table
-CREATE TABLE likes (
-    id BIGSERIAL PRIMARY KEY,
-    actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    rkey INT8 NOT NULL,
-    subject_type like_subject_type NOT NULL DEFAULT 'post',
-    subject_id BIGINT,
-    via_repost_id BIGINT REFERENCES reposts(id) ON DELETE CASCADE,
-    UNIQUE (actor_id, rkey)
-);
-
--- Migrate data back from split tables
-INSERT INTO likes (actor_id, rkey, subject_type, subject_id, via_repost_id)
-SELECT actor_id, rkey, 'post'::like_subject_type, post_id, via_repost_id
-FROM post_likes;
-
-INSERT INTO likes (actor_id, rkey, subject_type, subject_id, via_repost_id)
-SELECT actor_id, rkey, 'feedgen'::like_subject_type, feedgen_id, NULL
-FROM feedgen_likes;
-
-INSERT INTO likes (actor_id, rkey, subject_type, subject_id, via_repost_id)
-SELECT actor_id, rkey, 'labeler'::like_subject_type, labeler_actor_id::bigint, NULL
-FROM labeler_likes;
-
--- Recreate indexes
-CREATE INDEX idx_likes_actor ON likes(actor_id);
-CREATE INDEX idx_likes_subject ON likes(subject_type, subject_id);
-CREATE INDEX idx_likes_rkey ON likes(rkey);
-CREATE INDEX idx_likes_via_repost ON likes(via_repost_id) WHERE via_repost_id IS NOT NULL;
-
--- Drop new tables
-DROP TABLE IF EXISTS labeler_likes CASCADE;
-DROP TABLE IF EXISTS feedgen_likes CASCADE;
-DROP TABLE IF EXISTS post_likes CASCADE;
diff --git a/migrations/2025-11-17-001423_split_likes_table/up.sql b/migrations/2025-11-17-001423_split_likes_table/up.sql
deleted file mode 100644
index b7c16d7d..00000000
--- a/migrations/2025-11-17-001423_split_likes_table/up.sql
+++ /dev/null
@@ -1,114 +0,0 @@
--- =============================================================================
--- SPLIT LIKES TABLE - Create new tables with proper foreign keys
--- =============================================================================
---
--- Background: The polymorphic likes table uses subject_type enum to determine
--- what subject_id references. This prevents foreign key constraints and adds
--- query overhead. Splitting into three tables enables:
--- - Proper referential integrity with FK constraints
--- - Better query performance (no discriminator filtering)
--- - Smaller indexes (no subject_type in every index)
--- - TimescaleDB hypertable conversion for post_likes
---
--- Distribution: 99.98% post likes, 0.02% feedgen/labeler likes
---
--- =============================================================================
-
--- 1. POST LIKES (99.98% of likes)
--- Will become a TimescaleDB hypertable later
-CREATE TABLE post_likes (
-    actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    rkey INT8 NOT NULL,
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    via_repost_id BIGINT REFERENCES reposts(id) ON DELETE CASCADE,
-    PRIMARY KEY (actor_id, rkey)
-);
-
-CREATE INDEX idx_post_likes_post ON post_likes(post_id);
-CREATE INDEX idx_post_likes_rkey ON post_likes(rkey);
-CREATE INDEX idx_post_likes_via_repost ON post_likes(via_repost_id)
-    WHERE via_repost_id IS NOT NULL;
-
-COMMENT ON TABLE post_likes IS 'Likes on posts (99.98% of all likes). FK to posts ensures referential integrity.';
-COMMENT ON COLUMN post_likes.via_repost_id IS 'If user liked via a repost, reference to the repost record (for notification context).';
-
--- 2. FEEDGEN LIKES (rare)
-CREATE TABLE feedgen_likes (
-    actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    rkey INT8 NOT NULL,
-    feedgen_id BIGINT NOT NULL REFERENCES feedgens(id) ON DELETE CASCADE,
-    PRIMARY KEY (actor_id, rkey)
-);
-
-CREATE INDEX idx_feedgen_likes_feedgen ON feedgen_likes(feedgen_id);
-CREATE INDEX idx_feedgen_likes_rkey ON feedgen_likes(rkey);
-
-COMMENT ON TABLE feedgen_likes IS 'Likes on feed generators (rare, ~0.01% of likes).';
-
--- 3. LABELER LIKES (rare)
-CREATE TABLE labeler_likes (
-    actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    rkey INT8 NOT NULL,
-    labeler_actor_id INTEGER NOT NULL REFERENCES actors(id) ON DELETE CASCADE,
-    PRIMARY KEY (actor_id, rkey)
-);
-
-CREATE INDEX idx_labeler_likes_labeler ON labeler_likes(labeler_actor_id);
-CREATE INDEX idx_labeler_likes_rkey ON labeler_likes(rkey);
-
-COMMENT ON TABLE labeler_likes IS 'Likes on labeler services (rare, ~0.01% of likes).';
-
--- =============================================================================
--- MIGRATE DATA from old likes table
--- =============================================================================
-
--- Migrate post likes (should be 99.98% of data)
-INSERT INTO post_likes (actor_id, rkey, post_id, via_repost_id)
-SELECT actor_id, rkey, subject_id, via_repost_id
-FROM likes
-WHERE subject_type = 'post'
-  AND subject_id IS NOT NULL;  -- Only migrate if we have a valid subject_id
-
--- Migrate feedgen likes
-INSERT INTO feedgen_likes (actor_id, rkey, feedgen_id)
-SELECT actor_id, rkey, subject_id
-FROM likes
-WHERE subject_type = 'feedgen'
-  AND subject_id IS NOT NULL;
-
--- Migrate labeler likes (note: labelers use actor_id, not a separate labeler table ID)
-INSERT INTO labeler_likes (actor_id, rkey, labeler_actor_id)
-SELECT actor_id, rkey, subject_id::integer  -- Cast to integer since labelers reference actors(id)
-FROM likes
-WHERE subject_type = 'labeler'
-  AND subject_id IS NOT NULL;
-
--- Log migration results
-DO $$
-DECLARE
-    post_count INTEGER;
-    feedgen_count INTEGER;
-    labeler_count INTEGER;
-    orphaned_count INTEGER;
-BEGIN
-    SELECT COUNT(*) INTO post_count FROM post_likes;
-    SELECT COUNT(*) INTO feedgen_count FROM feedgen_likes;
-    SELECT COUNT(*) INTO labeler_count FROM labeler_likes;
-    SELECT COUNT(*) INTO orphaned_count FROM likes WHERE subject_id IS NULL;
-
-    RAISE NOTICE 'Migration complete:';
-    RAISE NOTICE '  post_likes: % rows', post_count;
-    RAISE NOTICE '  feedgen_likes: % rows', feedgen_count;
-    RAISE NOTICE '  labeler_likes: % rows', labeler_count;
-    RAISE NOTICE '  orphaned (skipped): % rows', orphaned_count;
-END $$;
-
--- =============================================================================
--- DROP OLD TABLE
--- =============================================================================
-
--- Drop the old polymorphic table immediately
-DROP TABLE likes CASCADE;
-
--- Drop the enum type (no longer needed)
-DROP TYPE like_subject_type;
diff --git a/migrations/2025-11-17-005012_add_diesel_type_generator_table/down.sql b/migrations/2025-11-17-005012_add_diesel_type_generator_table/down.sql
deleted file mode 100644
index 75c196e5..00000000
--- a/migrations/2025-11-17-005012_add_diesel_type_generator_table/down.sql
+++ /dev/null
@@ -1,2 +0,0 @@
--- Drop the dummy schema inference table
-DROP TABLE IF EXISTS _diesel_schema_inference;
diff --git a/migrations/2025-11-17-005012_add_diesel_type_generator_table/up.sql b/migrations/2025-11-17-005012_add_diesel_type_generator_table/up.sql
deleted file mode 100644
index 47f9d7ce..00000000
--- a/migrations/2025-11-17-005012_add_diesel_type_generator_table/up.sql
+++ /dev/null
@@ -1,16 +0,0 @@
--- Dummy table to force Diesel to generate SQL type definitions
---
--- This table is never used by the application, but exists solely to ensure
--- that Diesel generates type definitions for enums and composite types that are
--- only used inside other composite types.
---
--- Without this table, Diesel wouldn't generate these types in schema.rs since
--- they're not directly used as column types in any real tables.
-
-CREATE TABLE _diesel_schema_inference (
-    id INTEGER PRIMARY KEY,
-    video_mime_type video_mime_type NOT NULL,
-    caption_mime_type caption_mime_type NOT NULL,
-    facet_type facet_type NOT NULL,
-    video_caption post_video_caption NOT NULL
-);
diff --git a/migrations/2025-11-17-054507_enable_timescaledb/down.sql b/migrations/2025-11-17-054507_enable_timescaledb/down.sql
deleted file mode 100644
index 3e827c14..00000000
--- a/migrations/2025-11-17-054507_enable_timescaledb/down.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Remove TimescaleDB extension
--- Warning: This will drop all hypertables and their data
-DROP EXTENSION IF EXISTS timescaledb CASCADE;
diff --git a/migrations/2025-11-17-054507_enable_timescaledb/up.sql b/migrations/2025-11-17-054507_enable_timescaledb/up.sql
deleted file mode 100644
index d71344cd..00000000
--- a/migrations/2025-11-17-054507_enable_timescaledb/up.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Enable TimescaleDB extension
--- This adds TimescaleDB functions and capabilities to the database
-CREATE EXTENSION IF NOT EXISTS timescaledb CASCADE;
diff --git a/migrations/2025-11-17-054558_post_likes_hypertable/down.sql b/migrations/2025-11-17-054558_post_likes_hypertable/down.sql
deleted file mode 100644
index 849852f8..00000000
--- a/migrations/2025-11-17-054558_post_likes_hypertable/down.sql
+++ /dev/null
@@ -1,35 +0,0 @@
--- =============================================================================
--- Revert post_likes hypertable to regular table
--- =============================================================================
-
--- Remove retention policy first
-SELECT remove_retention_policy('post_likes', if_exists => true);
-
--- Remove compression policy
-SELECT remove_compression_policy('post_likes', if_exists => true);
-
--- Decompress all compressed chunks before converting back
-DO $$
-DECLARE
-    chunk_record RECORD;
-BEGIN
-    FOR chunk_record IN
-        SELECT chunk_schema, chunk_name
-        FROM timescaledb_information.chunks
-        WHERE hypertable_name = 'post_likes' AND is_compressed
-    LOOP
-        EXECUTE format('SELECT decompress_chunk(%L)',
-                      chunk_record.chunk_schema || '.' || chunk_record.chunk_name);
-    END LOOP;
-END $$;
-
--- Note: There's no direct way to convert a hypertable back to a regular table
--- The data remains accessible, but the hypertable structure stays
--- To fully revert, you would need to:
--- 1. Create a new regular table
--- 2. Copy all data from the hypertable
--- 3. Drop the hypertable
--- 4. Rename the new table
--- This is intentionally not automated to prevent accidental data loss
-RAISE NOTICE 'Retention and compression policies removed. Hypertable structure remains.';
-RAISE NOTICE 'To fully revert to regular table, manual intervention required.';
diff --git a/migrations/2025-11-17-054558_post_likes_hypertable/up.sql b/migrations/2025-11-17-054558_post_likes_hypertable/up.sql
deleted file mode 100644
index 9e0434d7..00000000
--- a/migrations/2025-11-17-054558_post_likes_hypertable/up.sql
+++ /dev/null
@@ -1,41 +0,0 @@
--- =============================================================================
--- Convert post_likes to TimescaleDB hypertable
--- =============================================================================
-
--- Convert post_likes to hypertable partitioned by rkey (timestamp-based TID)
--- Use tid_timestamp function to extract timestamp from TID for partitioning
--- Chunk interval: 1 day (86400000000 microseconds in TID format)
-SELECT create_hypertable(
-    'post_likes',
-    by_range('rkey', partition_func => 'tid_timestamp', partition_interval => INTERVAL '1 day'),
-    migrate_data => true,
-    if_not_exists => true
-);
-
--- Enable compression on old chunks
--- compress_segmentby: Group by actor_id for better compression (likes from same actor)
--- compress_orderby: Order by rkey DESC for time-series queries
-ALTER TABLE post_likes SET (
-    timescaledb.compress,
-    timescaledb.compress_segmentby = 'actor_id',
-    timescaledb.compress_orderby = 'rkey DESC'
-);
-
--- Add compression policy: compress chunks older than 7 days
--- This happens automatically in the background
-SELECT add_compression_policy(
-    'post_likes',
-    compress_after => INTERVAL '7 days',
-    if_not_exists => true
-);
-
--- Add retention policy: drop chunks older than 31 days
--- This replaces the manual cleanup in CleanupWorker
-SELECT add_retention_policy(
-    'post_likes',
-    drop_after => INTERVAL '31 days',
-    if_not_exists => true
-);
-
--- Create helpful comment
-COMMENT ON TABLE post_likes IS 'TimescaleDB hypertable for post likes. Partitioned by rkey (timestamp). Auto-compressed after 7 days, auto-dropped after 31 days.';
diff --git a/migrations/2025-11-17-062926_optimize_post_likes_hypertable/down.sql b/migrations/2025-11-17-062926_optimize_post_likes_hypertable/down.sql
deleted file mode 100644
index de967e57..00000000
--- a/migrations/2025-11-17-062926_optimize_post_likes_hypertable/down.sql
+++ /dev/null
@@ -1,62 +0,0 @@
--- =============================================================================
--- Revert post_likes Hypertable Optimizations
--- =============================================================================
-
--- Note: Reverting chunk interval only affects NEW chunks.
--- Existing 4-hour chunks will remain until they age out (31-day retention).
-
--- -----------------------------------------------------------------------------
--- 1. Revert Chunk Skipping
--- -----------------------------------------------------------------------------
-
--- TimescaleDB doesn't provide disable_chunk_skipping as of v2.17
--- Chunk skipping will remain enabled but won't cause issues
-RAISE NOTICE 'Chunk skipping cannot be disabled once enabled';
-RAISE NOTICE 'This is not harmful - it only provides query optimization';
-
--- -----------------------------------------------------------------------------
--- 2. Revert Compression Policy: 12 hours → 7 days
--- -----------------------------------------------------------------------------
-
-SELECT remove_compression_policy('post_likes', if_exists => true);
-
-SELECT add_compression_policy(
-    'post_likes',
-    compress_after => INTERVAL '7 days',
-    if_not_exists => true
-);
-
--- -----------------------------------------------------------------------------
--- 3. Revert Chunk Interval: 4 hours → 1 day
--- -----------------------------------------------------------------------------
-
-SELECT set_chunk_time_interval('post_likes', INTERVAL '1 day');
-
--- -----------------------------------------------------------------------------
--- 4. Update Table Comment
--- -----------------------------------------------------------------------------
-
-COMMENT ON TABLE post_likes IS 'TimescaleDB hypertable for post likes. Partitioned by rkey (timestamp). Auto-compressed after 7 days, auto-dropped after 31 days.';
-
--- -----------------------------------------------------------------------------
--- 5. Verification
--- -----------------------------------------------------------------------------
-
-SELECT
-    h.hypertable_name,
-    d.time_interval as chunk_interval,
-    h.compression_enabled
-FROM timescaledb_information.hypertables h
-JOIN timescaledb_information.dimensions d ON h.hypertable_name = d.hypertable_name
-WHERE h.hypertable_name = 'post_likes';
-
-SELECT
-    hypertable_name,
-    proc_name,
-    config
-FROM timescaledb_information.jobs
-WHERE hypertable_name = 'post_likes'
-ORDER BY proc_name;
-
-RAISE NOTICE 'Reverted to 1-day chunks and 7-day compression';
-RAISE NOTICE 'Existing 4-hour chunks will remain until 31-day retention drops them';
diff --git a/migrations/2025-11-17-062926_optimize_post_likes_hypertable/up.sql b/migrations/2025-11-17-062926_optimize_post_likes_hypertable/up.sql
deleted file mode 100644
index 75dc8f58..00000000
--- a/migrations/2025-11-17-062926_optimize_post_likes_hypertable/up.sql
+++ /dev/null
@@ -1,159 +0,0 @@
--- =============================================================================
--- Optimize post_likes Hypertable for Network Scale (20M likes/day)
--- =============================================================================
---
--- Based on performance testing and analysis:
--- - Current: 812K rows, 1-day chunks (4 MB each)
--- - Target: 20M rows/day, 4-hour chunks (52 MB each)
--- - Compression ratio: 120:1 measured
--- - Write performance: 172K rows/sec (COPY), only 7% slower on compressed chunks
---
--- Storage projections at 31-day retention:
--- - Before: 30 GB (2-day compression)
--- - After:  26 GB (12-hour compression)
--- - Savings: 15% overall, 65% less uncompressed data
---
--- References:
--- - .claude/reference/NETWORK_SCALE_STORAGE_ANALYSIS.md
--- - .claude/reference/COMPRESSION_PERFORMANCE_TEST_RESULTS.md
--- =============================================================================
-
--- -----------------------------------------------------------------------------
--- 1. Optimize Chunk Interval: 1 day → 4 hours
--- -----------------------------------------------------------------------------
---
--- Why 4 hours:
--- - Creates 6 chunks/day (manageable metadata overhead)
--- - ~52 MB per chunk (optimal: 10-100 MB range)
--- - ~3.3M rows per chunk at 20M/day
--- - 180 total chunks at 31-day retention
---
--- Note: Only affects NEW chunks. Existing chunks remain at 1-day interval.
--- This is expected and acceptable - they'll age out within 31 days.
---
-
-SELECT set_chunk_time_interval('post_likes', INTERVAL '4 hours');
-
--- Verify the change
-DO $$
-DECLARE
-    interval_hours NUMERIC;
-BEGIN
-    SELECT
-        EXTRACT(EPOCH FROM d.time_interval) / 3600 INTO interval_hours
-    FROM timescaledb_information.dimensions d
-    WHERE d.hypertable_name = 'post_likes';
-
-    IF interval_hours != 4 THEN
-        RAISE EXCEPTION 'Chunk interval not set correctly. Expected 4 hours, got % hours', interval_hours;
-    END IF;
-
-    RAISE NOTICE 'Chunk interval set to 4 hours successfully';
-END $$;
-
--- -----------------------------------------------------------------------------
--- 2. Update Compression Policy: 7 days → 12 hours (aggressive!)
--- -----------------------------------------------------------------------------
---
--- Why 12 hours:
--- - 99.9% of writes are <5 seconds old → hit uncompressed chunks (6.9ms)
--- - Only 7% performance penalty on compressed chunks (7.4ms vs 6.9ms)
--- - 65% reduction in uncompressed data (1.5 GB vs 6.1 GB)
--- - Safe buffer for Jetstream catch-up scenarios (<48 hours)
---
--- Performance testing showed:
--- - COPY to uncompressed: 6.9ms per 1,000 rows
--- - COPY to compressed:   7.4ms per 1,000 rows
--- - Difference: negligible for our use case!
---
-
--- Remove existing compression policy
-SELECT remove_compression_policy('post_likes', if_exists => true);
-
--- Add new aggressive compression policy
-SELECT add_compression_policy(
-    'post_likes',
-    compress_after => INTERVAL '12 hours',
-    if_not_exists => true
-);
-
-COMMENT ON TABLE post_likes IS 'TimescaleDB hypertable for post likes. Partitioned by rkey (timestamp) with 4-hour chunks. Auto-compressed after 12 hours, auto-dropped after 31 days. Chunk skipping enabled on post_id.';
-
--- -----------------------------------------------------------------------------
--- 3. Enable Chunk Skipping on post_id
--- -----------------------------------------------------------------------------
---
--- Why post_id:
--- - Highly correlated with time (likes cluster near post creation)
--- - 568K unique posts, most with 1-2 likes
--- - Top post spans only 14 days, not full 30 days
--- - Estimated 50-75% chunk exclusion for post queries
---
--- Why NOT actor_id:
--- - Not correlated with time (actors span entire 30-day window)
--- - Only 297 unique actors across all data
--- - Would provide minimal benefit
---
--- Note: Only works on chunks compressed AFTER enabling this feature.
--- Existing compressed chunks won't benefit, but they'll age out.
---
-
--- Enable chunk skipping feature (tech preview as of TimescaleDB 2.17)
-SET timescaledb.enable_chunk_skipping = on;
-
-SELECT enable_chunk_skipping('post_likes', 'post_id');
-
--- Verify chunk skipping is enabled
-DO $$
-DECLARE
-    skipping_enabled BOOLEAN;
-BEGIN
-    -- Check if chunk skipping configuration exists
-    -- Note: This is a simplified check. Actual verification would query
-    -- timescaledb internal catalogs which may vary by version.
-    RAISE NOTICE 'Chunk skipping enabled on post_id column';
-    RAISE NOTICE 'This will take effect on chunks compressed after this migration';
-END $$;
-
--- -----------------------------------------------------------------------------
--- 4. Summary of Configuration
--- -----------------------------------------------------------------------------
-
--- Display current hypertable configuration
-SELECT
-    h.hypertable_name,
-    h.num_chunks,
-    h.compression_enabled,
-    d.column_name as partition_column,
-    d.time_interval as chunk_interval
-FROM timescaledb_information.hypertables h
-JOIN timescaledb_information.dimensions d ON h.hypertable_name = d.hypertable_name
-WHERE h.hypertable_name = 'post_likes';
-
--- Display policies
-SELECT
-    hypertable_name,
-    proc_name,
-    config,
-    schedule_interval,
-    next_start
-FROM timescaledb_information.jobs
-WHERE hypertable_name = 'post_likes'
-ORDER BY proc_name;
-
--- Display chunk statistics
-SELECT
-    COUNT(*) as total_chunks,
-    COUNT(*) FILTER (WHERE is_compressed) as compressed,
-    COUNT(*) FILTER (WHERE NOT is_compressed) as uncompressed,
-    pg_size_pretty(SUM(pg_total_relation_size(chunk_schema || '.' || chunk_name))) as total_size
-FROM timescaledb_information.chunks
-WHERE hypertable_name = 'post_likes';
-
--- Expected behavior after migration:
--- ✅ New chunks created every 4 hours (6 per day)
--- ✅ Chunks compressed automatically after 12 hours
--- ✅ Chunks dropped automatically after 31 days (existing retention policy)
--- ✅ Queries on post_id will skip irrelevant chunks (50-75% reduction)
--- ✅ Write performance: 172K rows/sec, minimal impact from compression (7%)
--- ✅ Storage at 31-day retention: ~26 GB (vs 30 GB before optimization)
diff --git a/migrations/2025-11-17-072850_convert_posts_to_hypertable/down.sql b/migrations/2025-11-17-072850_convert_posts_to_hypertable/down.sql
deleted file mode 100644
index aff681b6..00000000
--- a/migrations/2025-11-17-072850_convert_posts_to_hypertable/down.sql
+++ /dev/null
@@ -1,75 +0,0 @@
--- =============================================================================
--- Rollback: Convert posts Hypertable Back to Regular Table
--- =============================================================================
---
--- WARNING: This migration will:
--- 1. Decompress all chunks (may take time for large tables)
--- 2. Convert hypertable back to regular PostgreSQL table
--- 3. Restore original PRIMARY KEY structure
---
--- Note: Chunk skipping cannot be fully rolled back, but it's harmless on regular tables
--- =============================================================================
-
--- -----------------------------------------------------------------------------
--- Step 1: Remove Compression Policy
--- -----------------------------------------------------------------------------
-
-SELECT remove_compression_policy('posts', if_exists => true);
-
--- -----------------------------------------------------------------------------
--- Step 2: Decompress All Chunks
--- -----------------------------------------------------------------------------
-
--- This may take some time for large tables
-SELECT decompress_chunk(chunk_schema || '.' || chunk_name)
-FROM timescaledb_information.chunks
-WHERE hypertable_name = 'posts' AND is_compressed;
-
--- -----------------------------------------------------------------------------
--- Step 3: Disable Compression
--- -----------------------------------------------------------------------------
-
-ALTER TABLE posts SET (
-    timescaledb.compress = false
-);
-
--- -----------------------------------------------------------------------------
--- Step 4: Convert Back to Regular Table
--- -----------------------------------------------------------------------------
-
--- Note: This does NOT migrate data back - it just removes the hypertable metadata
--- The data remains in the chunks which become regular tables
--- For a full rollback with data migration, you would need to:
--- 1. Create a new regular table
--- 2. Copy all data from the hypertable
--- 3. Drop the hypertable
--- 4. Rename the new table
-
--- For now, we'll just raise a notice
-DO $$
-BEGIN
-    RAISE NOTICE 'Rollback of hypertable conversion is complex and may not be needed.';
-    RAISE NOTICE 'The table will remain a hypertable but with compression disabled.';
-    RAISE NOTICE 'To fully convert back, manual steps are required:';
-    RAISE NOTICE '1. CREATE TABLE posts_backup AS SELECT * FROM posts';
-    RAISE NOTICE '2. DROP TABLE posts CASCADE';
-    RAISE NOTICE '3. ALTER TABLE posts_backup RENAME TO posts';
-    RAISE NOTICE '4. Recreate indexes and constraints';
-END $$;
-
--- -----------------------------------------------------------------------------
--- Step 5: Restore Original PRIMARY KEY (if needed)
--- -----------------------------------------------------------------------------
-
--- Drop composite primary key
-ALTER TABLE posts DROP CONSTRAINT posts_pkey;
-
--- Restore original single-column primary key
--- Note: This assumes the original was just (id)
-ALTER TABLE posts ADD PRIMARY KEY (id);
-
-COMMENT ON COLUMN posts.id IS 'Primary key';
-COMMENT ON COLUMN posts.rkey IS 'TID timestamp identifier';
-
--- Remove table comment
-COMMENT ON TABLE posts IS NULL;
diff --git a/migrations/2025-11-17-072850_convert_posts_to_hypertable/up.sql b/migrations/2025-11-17-072850_convert_posts_to_hypertable/up.sql
deleted file mode 100644
index f73b138a..00000000
--- a/migrations/2025-11-17-072850_convert_posts_to_hypertable/up.sql
+++ /dev/null
@@ -1,123 +0,0 @@
--- =============================================================================
--- Convert posts Table to TimescaleDB Hypertable (Full Network Scale)
--- =============================================================================
---
--- Configuration:
--- - All posts (complete + stubs): Stubs will have lower compression but simplified architecture
--- - NO retention policy: Full retention for network scale (2B posts)
--- - Chunk interval: 1 day (adjust to 4 hours at network scale via set_chunk_time_interval)
--- - Compression: 12 hours (aggressive, like post_likes)
--- - Expected compression: ~6,496:1 for complete posts, ~20:1 for stubs, combined ~100:1
---
--- Storage projections at 2B posts with 10% complete, 90% stubs:
--- - Complete: 200M × 0.43 bytes = 86 MB
--- - Stubs: 1.8B × ~50 bytes = 90 GB
--- - Total: ~90 GB (vs ~540 GB traditional, 83% savings)
---
--- Note: PRIMARY KEY must include rkey for hypertable partitioning
--- =============================================================================
-
--- -----------------------------------------------------------------------------
--- Step 1: Add rkey to PRIMARY KEY
--- -----------------------------------------------------------------------------
-
--- Drop existing primary key (CASCADE to drop dependent foreign keys)
--- TimescaleDB does not support foreign keys to hypertables, so we will NOT recreate them
--- Referential integrity will be enforced at the application level
-ALTER TABLE posts DROP CONSTRAINT posts_pkey CASCADE;
-
--- Recreate primary key with (id, rkey) for hypertable compatibility
-ALTER TABLE posts ADD PRIMARY KEY (id, rkey);
-
-COMMENT ON COLUMN posts.id IS 'Primary key (first part of composite key with rkey for hypertable)';
-COMMENT ON COLUMN posts.rkey IS 'TID timestamp identifier (primary key second part, partitioning column)';
-
--- Note: Foreign keys to posts table have been dropped and will NOT be recreated
--- TimescaleDB limitation: cannot have FK constraints to hypertables
--- Referential integrity enforced at application level
-
--- -----------------------------------------------------------------------------
--- Step 2: Convert to Hypertable
--- -----------------------------------------------------------------------------
-
-SELECT create_hypertable(
-    'posts',
-    by_range('rkey', partition_func => 'tid_timestamp', partition_interval => INTERVAL '1 day'),
-    migrate_data => true,
-    if_not_exists => true
-);
-
-COMMENT ON TABLE posts IS 'TimescaleDB hypertable for posts (complete + stubs). Partitioned by rkey with 1-day chunks. Auto-compressed after 12 hours. No retention policy (full network scale).';
-
--- -----------------------------------------------------------------------------
--- Step 3: Enable Compression
--- -----------------------------------------------------------------------------
-
-ALTER TABLE posts SET (
-    timescaledb.compress,
-    timescaledb.compress_segmentby = 'actor_id',
-    timescaledb.compress_orderby = 'rkey DESC'
-);
-
--- -----------------------------------------------------------------------------
--- Step 4: Add Aggressive Compression Policy (12 hours, like post_likes)
--- -----------------------------------------------------------------------------
-
-SELECT add_compression_policy(
-    'posts',
-    compress_after => INTERVAL '12 hours',
-    if_not_exists => true
-);
-
--- NO retention policy - we want full retention for network scale!
-
--- -----------------------------------------------------------------------------
--- Step 5: Enable Chunk Skipping (Tech Preview Feature)
--- -----------------------------------------------------------------------------
-
--- Enable chunk skipping feature for this session
-SET timescaledb.enable_chunk_skipping = on;
-
--- Enable chunk skipping on post_id for efficient parent post queries
--- This allows queries like "get replies to post X" to skip irrelevant chunks
-SELECT enable_chunk_skipping('posts', 'parent_post_id');
-
--- Enable chunk skipping on root_post_id for thread queries
-SELECT enable_chunk_skipping('posts', 'root_post_id');
-
--- -----------------------------------------------------------------------------
--- Step 6: Display Configuration Summary
--- -----------------------------------------------------------------------------
-
--- Display hypertable configuration
-DO $$
-DECLARE
-    chunk_count INTEGER;
-    compressed_count INTEGER;
-BEGIN
-    SELECT COUNT(*), COUNT(*) FILTER (WHERE is_compressed)
-    INTO chunk_count, compressed_count
-    FROM timescaledb_information.chunks
-    WHERE hypertable_name = 'posts';
-
-    RAISE NOTICE '';
-    RAISE NOTICE '✅ posts table converted to hypertable successfully';
-    RAISE NOTICE 'Configuration:';
-    RAISE NOTICE '  - Chunks: % total (% compressed)', chunk_count, compressed_count;
-    RAISE NOTICE '  - Compression: Enabled (actor_id segmentby, rkey orderby)';
-    RAISE NOTICE '  - Compression policy: 12 hours';
-    RAISE NOTICE '  - Retention policy: NONE (full network scale)';
-    RAISE NOTICE '  - Chunk skipping: Enabled on parent_post_id, root_post_id';
-    RAISE NOTICE '';
-    RAISE NOTICE 'Query the following for more details:';
-    RAISE NOTICE '  SELECT * FROM timescaledb_information.hypertables WHERE hypertable_name = ''posts'';';
-    RAISE NOTICE '  SELECT * FROM timescaledb_information.jobs WHERE hypertable_name = ''posts'';';
-END $$;
-
--- Expected behavior after migration:
--- ✅ All posts (complete + stubs) migrated to hypertable
--- ✅ Chunks created with 1-day interval (32-40 chunks for current data)
--- ✅ Chunks compressed automatically after 12 hours
--- ✅ NO retention - all data kept forever (network scale)
--- ✅ Chunk skipping enabled on parent_post_id and root_post_id
--- ✅ Storage: ~90 GB at 2B posts (vs 540 GB traditional, 83% savings)
diff --git a/migrations/2025-11-17-072937_remove_retention_policies/down.sql b/migrations/2025-11-17-072937_remove_retention_policies/down.sql
deleted file mode 100644
index 222a4eda..00000000
--- a/migrations/2025-11-17-072937_remove_retention_policies/down.sql
+++ /dev/null
@@ -1,43 +0,0 @@
--- =============================================================================
--- Restore Retention Policies (Rollback to 31-Day Retention)
--- =============================================================================
---
--- This restores the previous 31-day retention policy on post_likes
--- Note: posts table never had retention, so nothing to restore there
---
--- =============================================================================
-
--- -----------------------------------------------------------------------------
--- Restore 31-Day Retention Policy on post_likes
--- -----------------------------------------------------------------------------
-
-SELECT add_retention_policy(
-    'post_likes',
-    drop_after => INTERVAL '31 days',
-    if_not_exists => true
-);
-
-COMMENT ON TABLE post_likes IS 'TimescaleDB hypertable for post likes. Partitioned by rkey (timestamp) with 4-hour chunks. Auto-compressed after 12 hours, auto-dropped after 31 days. Chunk skipping enabled on post_id.';
-
--- -----------------------------------------------------------------------------
--- Display Restored Configuration
--- -----------------------------------------------------------------------------
-
-SELECT
-    hypertable_name,
-    proc_name,
-    config,
-    schedule_interval,
-    next_start
-FROM timescaledb_information.jobs
-WHERE hypertable_name = 'post_likes'
-ORDER BY proc_name;
-
--- Summary
-DO $$
-BEGIN
-    RAISE NOTICE '✅ Retention policy restored on post_likes';
-    RAISE NOTICE 'Configuration:';
-    RAISE NOTICE '  - post_likes: 4-hour chunks, 12-hour compression, 31-day retention';
-    RAISE NOTICE '  - posts: 1-day chunks, 12-hour compression, NO retention';
-END $$;
diff --git a/migrations/2025-11-17-072937_remove_retention_policies/up.sql b/migrations/2025-11-17-072937_remove_retention_policies/up.sql
deleted file mode 100644
index 82932e57..00000000
--- a/migrations/2025-11-17-072937_remove_retention_policies/up.sql
+++ /dev/null
@@ -1,63 +0,0 @@
--- =============================================================================
--- Remove Retention Policies for Full Network Scale Storage
--- =============================================================================
---
--- Decision: With 83-99% compression savings, we can afford full retention
--- Storage projections with compression (no retention):
--- - post_likes @ 12B records: ~15 GB (was 30 GB with 31-day retention)
--- - posts @ 2B records: ~90 GB (complete + stubs)
--- - Total: ~105 GB for full network scale (vs ~1TB+ traditional)
---
--- This allows:
--- - Historical data analysis
--- - Backfill without data loss
--- - Simpler operations (no chunk deletion)
--- - Better compression (more data to compress)
---
--- =============================================================================
-
--- -----------------------------------------------------------------------------
--- Remove Retention Policy from post_likes
--- -----------------------------------------------------------------------------
-
-SELECT remove_retention_policy('post_likes', if_exists => true);
-
-COMMENT ON TABLE post_likes IS 'TimescaleDB hypertable for post likes. Partitioned by rkey (timestamp) with 4-hour chunks. Auto-compressed after 12 hours. NO retention policy (full network scale). Chunk skipping enabled on post_id.';
-
--- -----------------------------------------------------------------------------
--- Display Updated Configuration
--- -----------------------------------------------------------------------------
-
--- Show policies for post_likes (should only show compression, no retention)
-SELECT
-    hypertable_name,
-    proc_name,
-    config,
-    schedule_interval
-FROM timescaledb_information.jobs
-WHERE hypertable_name = 'post_likes'
-ORDER BY proc_name;
-
--- Show policies for posts (should only show compression, no retention)
-SELECT
-    hypertable_name,
-    proc_name,
-    config,
-    schedule_interval
-FROM timescaledb_information.jobs
-WHERE hypertable_name = 'posts'
-ORDER BY proc_name;
-
--- Summary
-DO $$
-BEGIN
-    RAISE NOTICE '✅ Retention policies removed from all hypertables';
-    RAISE NOTICE 'Configuration:';
-    RAISE NOTICE '  - post_likes: 4-hour chunks, 12-hour compression, NO retention';
-    RAISE NOTICE '  - posts: 1-day chunks, 12-hour compression, NO retention';
-    RAISE NOTICE ' ';
-    RAISE NOTICE 'Expected storage at network scale (full retention):';
-    RAISE NOTICE '  - post_likes @ 12B records: ~15 GB';
-    RAISE NOTICE '  - posts @ 2B records: ~90 GB';
-    RAISE NOTICE '  - Total: ~105 GB (vs ~1TB+ traditional, 90%% savings)';
-END $$;
diff --git a/migrations/2025-11-18-191434_remove_post_id_natural_keys/down.sql b/migrations/2025-11-18-191434_remove_post_id_natural_keys/down.sql
deleted file mode 100644
index 7f38a88b..00000000
--- a/migrations/2025-11-18-191434_remove_post_id_natural_keys/down.sql
+++ /dev/null
@@ -1,381 +0,0 @@
--- Rollback: Restore post_id synthetic integer keys
--- This reverses the migration to natural keys
-
--- ============================================================================
--- Phase 1: Drop foreign key constraints that reference natural keys
--- ============================================================================
-
--- Child tables
-ALTER TABLE threadgate_allowed_lists DROP CONSTRAINT IF EXISTS threadgate_allowed_lists_threadgate_fkey;
-ALTER TABLE threadgate_allowed_lists DROP CONSTRAINT IF EXISTS threadgate_allowed_lists_list_id_fkey;
-ALTER TABLE threadgate_hidden_replies DROP CONSTRAINT IF EXISTS threadgate_hidden_replies_threadgate_fkey;
-ALTER TABLE postgate_detached DROP CONSTRAINT IF EXISTS postgate_detached_postgate_fkey;
-
--- Main tables
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_post_fkey;
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_via_repost_fkey;
-ALTER TABLE post_aggregate_stats DROP CONSTRAINT IF EXISTS post_aggregate_stats_post_fkey;
-ALTER TABLE threadgates DROP CONSTRAINT IF EXISTS threadgates_post_fkey;
-ALTER TABLE postgates DROP CONSTRAINT IF EXISTS postgates_post_fkey;
-ALTER TABLE profiles DROP CONSTRAINT IF EXISTS profiles_pinned_post_fkey;
-
--- ============================================================================
--- Phase 2: Drop unique constraints on natural keys
--- ============================================================================
-
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_actor_id_post_actor_id_post_rkey_key;
-
--- ============================================================================
--- Phase 3: Drop indexes on natural keys
--- ============================================================================
-
-DROP INDEX IF EXISTS idx_posts_parent;
-DROP INDEX IF EXISTS idx_posts_root;
-DROP INDEX IF EXISTS idx_posts_embedded_post;
-DROP INDEX IF EXISTS idx_post_likes_post_actor_id_post_rkey;
-DROP INDEX IF EXISTS idx_reposts_post_actor_id_post_rkey;
-
--- ============================================================================
--- Phase 4: Drop primary keys based on natural keys
--- ============================================================================
-
-ALTER TABLE posts DROP CONSTRAINT IF EXISTS posts_pkey;
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_pkey;
-ALTER TABLE post_aggregate_stats DROP CONSTRAINT IF EXISTS post_aggregate_stats_pkey;
-ALTER TABLE threadgates DROP CONSTRAINT IF EXISTS threadgates_pkey;
-ALTER TABLE postgates DROP CONSTRAINT IF EXISTS postgates_pkey;
-
--- Child tables
-ALTER TABLE threadgate_allowed_lists DROP CONSTRAINT IF EXISTS threadgate_allowed_lists_pkey;
-ALTER TABLE threadgate_hidden_replies DROP CONSTRAINT IF EXISTS threadgate_hidden_replies_pkey;
-ALTER TABLE postgate_detached DROP CONSTRAINT IF EXISTS postgate_detached_pkey;
-
--- ============================================================================
--- Phase 5: Restore id columns with BIGSERIAL (generates sequential integers)
--- ============================================================================
-
--- posts: Restore id and create composite primary key temporarily
-ALTER TABLE posts ADD COLUMN id BIGSERIAL;
-ALTER TABLE posts ADD PRIMARY KEY (id, rkey);  -- Composite PK includes rkey (partition key for hypertable)
-
--- reposts: Restore id
-ALTER TABLE reposts ADD COLUMN id BIGSERIAL PRIMARY KEY;
-
--- threadgates: Restore id
-ALTER TABLE threadgates ADD COLUMN id BIGSERIAL PRIMARY KEY;
-
--- postgates: Restore id
-ALTER TABLE postgates ADD COLUMN id BIGSERIAL PRIMARY KEY;
-
--- post_aggregate_stats: Restore post_id as primary key
-ALTER TABLE post_aggregate_stats ADD COLUMN post_id BIGINT PRIMARY KEY;
-
--- ============================================================================
--- Phase 6: Add back old foreign key columns
--- ============================================================================
-
--- posts self-references
-ALTER TABLE posts ADD COLUMN parent_post_id BIGINT;
-ALTER TABLE posts ADD COLUMN root_post_id BIGINT;
-ALTER TABLE posts ADD COLUMN embedded_post_id BIGINT;
-
--- post_likes
-ALTER TABLE post_likes ADD COLUMN post_id BIGINT;
-ALTER TABLE post_likes ADD COLUMN via_repost_id BIGINT;
-
--- reposts
-ALTER TABLE reposts ADD COLUMN post_id BIGINT;
-ALTER TABLE reposts ADD COLUMN via_repost_id BIGINT;
-
--- profiles
-ALTER TABLE profiles ADD COLUMN pinned_post_id BIGINT;
-
--- Child tables
-ALTER TABLE threadgate_allowed_lists ADD COLUMN threadgate_id BIGINT;
-ALTER TABLE threadgate_hidden_replies ADD COLUMN threadgate_id BIGINT;
-ALTER TABLE threadgate_hidden_replies ADD COLUMN post_id BIGINT;
-ALTER TABLE postgate_detached ADD COLUMN postgate_id BIGINT;
-ALTER TABLE postgate_detached ADD COLUMN detached_post_id BIGINT;
-
--- ============================================================================
--- Phase 7: Populate old columns from natural keys
--- ============================================================================
-
--- posts self-references
-UPDATE posts p SET
-    parent_post_id = parent.id
-FROM posts parent
-WHERE p.parent_post_actor_id = parent.actor_id
-    AND p.parent_post_rkey = parent.rkey;
-
-UPDATE posts p SET
-    root_post_id = root.id
-FROM posts root
-WHERE p.root_post_actor_id = root.actor_id
-    AND p.root_post_rkey = root.rkey;
-
-UPDATE posts p SET
-    embedded_post_id = embedded.id
-FROM posts embedded
-WHERE p.embedded_post_actor_id = embedded.actor_id
-    AND p.embedded_post_rkey = embedded.rkey;
-
--- post_likes (post reference)
-UPDATE post_likes pl SET
-    post_id = p.id
-FROM posts p
-WHERE pl.post_actor_id = p.actor_id
-    AND pl.post_rkey = p.rkey;
-
--- post_likes (via_repost reference)
-UPDATE post_likes pl SET
-    via_repost_id = r.id
-FROM reposts r
-WHERE pl.via_repost_actor_id = r.actor_id
-    AND pl.via_repost_rkey = r.rkey;
-
--- reposts (post reference)
-UPDATE reposts r SET
-    post_id = p.id
-FROM posts p
-WHERE r.post_actor_id = p.actor_id
-    AND r.post_rkey = p.rkey;
-
--- reposts (via_repost self-reference)
-UPDATE reposts r SET
-    via_repost_id = vr.id
-FROM reposts vr
-WHERE r.via_repost_actor_id = vr.actor_id
-    AND r.via_repost_rkey = vr.rkey;
-
--- post_aggregate_stats
-UPDATE post_aggregate_stats pas SET
-    post_id = p.id
-FROM posts p
-WHERE pas.post_actor_id = p.actor_id
-    AND pas.post_rkey = p.rkey;
-
--- profiles (only the actor's own posts)
-UPDATE profiles pr SET
-    pinned_post_id = p.id
-FROM posts p
-WHERE pr.pinned_post_rkey = p.rkey
-    AND pr.actor_id = p.actor_id;
-
--- threadgate_allowed_lists
-UPDATE threadgate_allowed_lists tal SET
-    threadgate_id = tg.id
-FROM threadgates tg
-WHERE tal.post_actor_id = tg.post_actor_id
-    AND tal.post_rkey = tg.post_rkey;
-
--- threadgate_hidden_replies
-UPDATE threadgate_hidden_replies thr SET
-    threadgate_id = tg.id,
-    post_id = p.id
-FROM threadgates tg, posts p
-WHERE thr.post_actor_id = tg.post_actor_id
-    AND thr.post_rkey = tg.post_rkey
-    AND thr.hidden_post_actor_id = p.actor_id
-    AND thr.hidden_post_rkey = p.rkey;
-
--- postgate_detached
-UPDATE postgate_detached pd SET
-    postgate_id = pg.id,
-    detached_post_id = p.id
-FROM postgates pg, posts p
-WHERE pd.post_actor_id = pg.post_actor_id
-    AND pd.post_rkey = pg.post_rkey
-    AND pd.detached_post_actor_id = p.actor_id
-    AND pd.detached_post_rkey = p.rkey;
-
--- ============================================================================
--- Phase 8: Drop natural key columns
--- ============================================================================
-
--- posts
-ALTER TABLE posts DROP COLUMN parent_post_actor_id;
-ALTER TABLE posts DROP COLUMN parent_post_rkey;
-ALTER TABLE posts DROP COLUMN root_post_actor_id;
-ALTER TABLE posts DROP COLUMN root_post_rkey;
-ALTER TABLE posts DROP COLUMN embedded_post_actor_id;
-ALTER TABLE posts DROP COLUMN embedded_post_rkey;
-
--- post_likes
-ALTER TABLE post_likes DROP COLUMN post_actor_id;
-ALTER TABLE post_likes DROP COLUMN post_rkey;
-ALTER TABLE post_likes DROP COLUMN via_repost_actor_id;
-ALTER TABLE post_likes DROP COLUMN via_repost_rkey;
-
--- reposts
-ALTER TABLE reposts DROP COLUMN post_actor_id;
-ALTER TABLE reposts DROP COLUMN post_rkey;
-ALTER TABLE reposts DROP COLUMN via_repost_actor_id;
-ALTER TABLE reposts DROP COLUMN via_repost_rkey;
-
--- post_aggregate_stats
-ALTER TABLE post_aggregate_stats DROP COLUMN post_actor_id;
-ALTER TABLE post_aggregate_stats DROP COLUMN post_rkey;
-
--- threadgates (keep post_actor_id for other use, but drop post_rkey)
--- Actually, these were added for this migration, so drop both
-ALTER TABLE threadgates DROP COLUMN post_actor_id;
-ALTER TABLE threadgates DROP COLUMN post_rkey;
-
--- postgates
-ALTER TABLE postgates DROP COLUMN post_actor_id;
-ALTER TABLE postgates DROP COLUMN post_rkey;
-
--- profiles
-ALTER TABLE profiles DROP COLUMN pinned_post_rkey;
-
--- Child tables
-ALTER TABLE threadgate_allowed_lists DROP COLUMN post_actor_id;
-ALTER TABLE threadgate_allowed_lists DROP COLUMN post_rkey;
-
-ALTER TABLE threadgate_hidden_replies DROP COLUMN post_actor_id;
-ALTER TABLE threadgate_hidden_replies DROP COLUMN post_rkey;
-ALTER TABLE threadgate_hidden_replies DROP COLUMN hidden_post_actor_id;
-ALTER TABLE threadgate_hidden_replies DROP COLUMN hidden_post_rkey;
-
-ALTER TABLE postgate_detached DROP COLUMN post_actor_id;
-ALTER TABLE postgate_detached DROP COLUMN post_rkey;
-ALTER TABLE postgate_detached DROP COLUMN detached_post_actor_id;
-ALTER TABLE postgate_detached DROP COLUMN detached_post_rkey;
-
--- ============================================================================
--- Phase 9: Restore original primary keys for child tables
--- ============================================================================
-
-ALTER TABLE threadgate_allowed_lists ADD PRIMARY KEY (threadgate_id, list_id);
-ALTER TABLE threadgate_hidden_replies ADD PRIMARY KEY (threadgate_id, post_id);
-ALTER TABLE postgate_detached ADD PRIMARY KEY (postgate_id, detached_post_id);
-
--- ============================================================================
--- Phase 10: Restore unique constraints
--- ============================================================================
-
-ALTER TABLE post_likes ADD CONSTRAINT post_likes_actor_id_post_id_key
-    UNIQUE (actor_id, post_id);
-
-ALTER TABLE reposts ADD CONSTRAINT reposts_actor_id_post_id_key
-    UNIQUE (actor_id, post_id);
-
--- ============================================================================
--- Phase 11: Restore foreign key constraints
--- ============================================================================
-
--- posts self-references
-ALTER TABLE posts ADD CONSTRAINT posts_parent_post_id_fkey
-    FOREIGN KEY (parent_post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE posts ADD CONSTRAINT posts_root_post_id_fkey
-    FOREIGN KEY (root_post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE posts ADD CONSTRAINT posts_embedded_post_id_fkey
-    FOREIGN KEY (embedded_post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
--- post_likes
-ALTER TABLE post_likes ADD CONSTRAINT post_likes_post_id_fkey
-    FOREIGN KEY (post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE post_likes ADD CONSTRAINT post_likes_via_repost_id_fkey
-    FOREIGN KEY (via_repost_id)
-    REFERENCES reposts(id)
-    ON DELETE CASCADE;
-
--- reposts
-ALTER TABLE reposts ADD CONSTRAINT reposts_post_id_fkey
-    FOREIGN KEY (post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE reposts ADD CONSTRAINT reposts_via_repost_id_fkey
-    FOREIGN KEY (via_repost_id)
-    REFERENCES reposts(id)
-    ON DELETE CASCADE;
-
--- post_aggregate_stats
-ALTER TABLE post_aggregate_stats ADD CONSTRAINT post_aggregate_stats_post_id_fkey
-    FOREIGN KEY (post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
--- threadgates (needs post_id column first)
-ALTER TABLE threadgates ADD COLUMN post_id BIGINT;
-UPDATE threadgates tg SET post_id = p.id
-FROM posts p
-WHERE tg.post_actor_id = p.actor_id AND tg.post_rkey = p.rkey;
-
-ALTER TABLE threadgates ADD CONSTRAINT threadgates_post_id_fkey
-    FOREIGN KEY (post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
--- postgates (needs post_id column first)
-ALTER TABLE postgates ADD COLUMN post_id BIGINT;
-UPDATE postgates pg SET post_id = p.id
-FROM posts p
-WHERE pg.post_actor_id = p.actor_id AND pg.post_rkey = p.rkey;
-
-ALTER TABLE postgates ADD CONSTRAINT postgates_post_id_fkey
-    FOREIGN KEY (post_id)
-    REFERENCES posts(id)
-    ON DELETE CASCADE;
-
--- profiles
-ALTER TABLE profiles ADD CONSTRAINT profiles_pinned_post_id_fkey
-    FOREIGN KEY (pinned_post_id)
-    REFERENCES posts(id)
-    ON DELETE SET NULL;
-
--- Child tables
-ALTER TABLE threadgate_allowed_lists ADD CONSTRAINT threadgate_allowed_lists_threadgate_id_fkey
-    FOREIGN KEY (threadgate_id)
-    REFERENCES threadgates(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE threadgate_allowed_lists ADD CONSTRAINT threadgate_allowed_lists_list_id_fkey
-    FOREIGN KEY (list_id)
-    REFERENCES lists(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE threadgate_hidden_replies ADD CONSTRAINT threadgate_hidden_replies_threadgate_id_fkey
-    FOREIGN KEY (threadgate_id)
-    REFERENCES threadgates(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE postgate_detached ADD CONSTRAINT postgate_detached_postgate_id_fkey
-    FOREIGN KEY (postgate_id)
-    REFERENCES postgates(id)
-    ON DELETE CASCADE;
-
--- ============================================================================
--- Phase 12: Restore original indexes
--- ============================================================================
-
-CREATE INDEX idx_posts_parent
-    ON posts(parent_post_id)
-    WHERE parent_post_id IS NOT NULL;
-
-CREATE INDEX idx_posts_root
-    ON posts(root_post_id)
-    WHERE root_post_id IS NOT NULL;
-
-CREATE INDEX idx_posts_embedded_post_id
-    ON posts(embedded_post_id)
-    WHERE embedded_post_id IS NOT NULL;
-
-CREATE INDEX idx_post_likes_post_id
-    ON post_likes(post_id);
-
-CREATE INDEX idx_reposts_post_id
-    ON reposts(post_id);
diff --git a/migrations/2025-11-18-191434_remove_post_id_natural_keys/up.sql b/migrations/2025-11-18-191434_remove_post_id_natural_keys/up.sql
deleted file mode 100644
index f35e5b72..00000000
--- a/migrations/2025-11-18-191434_remove_post_id_natural_keys/up.sql
+++ /dev/null
@@ -1,391 +0,0 @@
--- Remove post_id synthetic integer keys, use natural keys (actor_id, rkey) instead
--- This migration is designed for parakeet_test (empty database) initially
--- Production migration for parakeet database will be handled separately
-
--- IMPORTANT: TimescaleDB Hypertable Constraints (as of recent TimescaleDB versions)
--- - Hypertable → Hypertable FKs: NOT SUPPORTED (e.g., posts → posts, post_likes → posts)
--- - Regular → Hypertable FKs: SUPPORTED (e.g., reposts → posts)
--- - Hypertable → Regular FKs: SUPPORTED (e.g., posts → actors)
---
--- posts and post_likes are hypertables:
--- - Cannot have FK constraints to each other (hypertable → hypertable)
--- - Cannot have self-referencing FKs (posts → posts)
--- - Regular tables CAN have FKs to them (reposts → posts works!)
---
--- NOTE: reposts may become a hypertable in the future, which would break post_likes → reposts FK
--- if post_likes.via_repost_id needs to reference it. This is already handled by using natural keys
--- without database FK constraints.
-
--- ============================================================================
--- Phase 1: Add new natural key columns to all tables (before dropping posts.id)
--- ============================================================================
-
--- posts table: Add natural key columns for parent/root/embedded references
-ALTER TABLE posts ADD COLUMN parent_post_actor_id INTEGER;
-ALTER TABLE posts ADD COLUMN parent_post_rkey BIGINT;
-ALTER TABLE posts ADD COLUMN root_post_actor_id INTEGER;
-ALTER TABLE posts ADD COLUMN root_post_rkey BIGINT;
-ALTER TABLE posts ADD COLUMN embedded_post_actor_id INTEGER;
-ALTER TABLE posts ADD COLUMN embedded_post_rkey BIGINT;
-
--- post_likes table: Add natural key columns for post reference and via_repost
-ALTER TABLE post_likes ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE post_likes ADD COLUMN post_rkey BIGINT;
-ALTER TABLE post_likes ADD COLUMN via_repost_actor_id INTEGER;
-ALTER TABLE post_likes ADD COLUMN via_repost_rkey BIGINT;
-
--- reposts table: Add natural key columns for post reference and via_repost
-ALTER TABLE reposts ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE reposts ADD COLUMN post_rkey BIGINT;
-ALTER TABLE reposts ADD COLUMN via_repost_actor_id INTEGER;
-ALTER TABLE reposts ADD COLUMN via_repost_rkey BIGINT;
-
--- post_aggregate_stats table: Add natural key columns
-ALTER TABLE post_aggregate_stats ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE post_aggregate_stats ADD COLUMN post_rkey BIGINT;
-
--- threadgates table: Add natural key columns
-ALTER TABLE threadgates ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE threadgates ADD COLUMN post_rkey BIGINT;
-
--- postgates table: Add natural key columns
-ALTER TABLE postgates ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE postgates ADD COLUMN post_rkey BIGINT;
-
--- profiles table: Add rkey-only column (actor_id is implicit)
-ALTER TABLE profiles ADD COLUMN pinned_post_rkey BIGINT;
-
--- statuses table: Add natural key columns for embed_post
-ALTER TABLE statuses ADD COLUMN embed_post_actor_id INTEGER;
-ALTER TABLE statuses ADD COLUMN embed_post_rkey BIGINT;
-
--- Child tables: threadgate_allowed_lists
-ALTER TABLE threadgate_allowed_lists ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE threadgate_allowed_lists ADD COLUMN post_rkey BIGINT;
-
--- Child tables: threadgate_hidden_replies
-ALTER TABLE threadgate_hidden_replies ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE threadgate_hidden_replies ADD COLUMN post_rkey BIGINT;
-ALTER TABLE threadgate_hidden_replies ADD COLUMN hidden_post_actor_id INTEGER;
-ALTER TABLE threadgate_hidden_replies ADD COLUMN hidden_post_rkey BIGINT;
-
--- Child tables: postgate_detached
-ALTER TABLE postgate_detached ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE postgate_detached ADD COLUMN post_rkey BIGINT;
-ALTER TABLE postgate_detached ADD COLUMN detached_post_actor_id INTEGER;
-ALTER TABLE postgate_detached ADD COLUMN detached_post_rkey BIGINT;
-
--- ============================================================================
--- Phase 2: Populate new columns from existing foreign keys (BEFORE dropping posts.id!)
--- ============================================================================
-
--- Populate posts self-references
-UPDATE posts p SET
-    parent_post_actor_id = parent.actor_id,
-    parent_post_rkey = parent.rkey
-FROM posts parent
-WHERE p.parent_post_id = parent.id;
-
-UPDATE posts p SET
-    root_post_actor_id = root.actor_id,
-    root_post_rkey = root.rkey
-FROM posts root
-WHERE p.root_post_id = root.id;
-
-UPDATE posts p SET
-    embedded_post_actor_id = embedded.actor_id,
-    embedded_post_rkey = embedded.rkey
-FROM posts embedded
-WHERE p.embedded_post_id = embedded.id;
-
--- Populate post_likes (post reference)
-UPDATE post_likes pl SET
-    post_actor_id = p.actor_id,
-    post_rkey = p.rkey
-FROM posts p
-WHERE pl.post_id = p.id;
-
--- Populate post_likes (via_repost reference)
-UPDATE post_likes pl SET
-    via_repost_actor_id = r.actor_id,
-    via_repost_rkey = r.rkey
-FROM reposts r
-WHERE pl.via_repost_id = r.id;
-
--- Populate reposts (post reference)
-UPDATE reposts r SET
-    post_actor_id = p.actor_id,
-    post_rkey = p.rkey
-FROM posts p
-WHERE r.post_id = p.id;
-
--- Populate reposts (via_repost self-reference)
-UPDATE reposts r SET
-    via_repost_actor_id = vr.actor_id,
-    via_repost_rkey = vr.rkey
-FROM reposts vr
-WHERE r.via_repost_id = vr.id;
-
--- Populate post_aggregate_stats
-UPDATE post_aggregate_stats pas SET
-    post_actor_id = p.actor_id,
-    post_rkey = p.rkey
-FROM posts p
-WHERE pas.post_id = p.id;
-
--- Populate threadgates
-UPDATE threadgates tg SET
-    post_actor_id = p.actor_id,
-    post_rkey = p.rkey
-FROM posts p
-WHERE tg.post_id = p.id;
-
--- Populate postgates
-UPDATE postgates pg SET
-    post_actor_id = p.actor_id,
-    post_rkey = p.rkey
-FROM posts p
-WHERE pg.post_id = p.id;
-
--- Populate profiles (only the actor's own posts)
-UPDATE profiles pr SET
-    pinned_post_rkey = p.rkey
-FROM posts p
-WHERE pr.pinned_post_id = p.id
-    AND p.actor_id = pr.actor_id;
-
--- Populate statuses (embed_post reference)
-UPDATE statuses s SET
-    embed_post_actor_id = p.actor_id,
-    embed_post_rkey = p.rkey
-FROM posts p
-WHERE s.embed_post_id = p.id;
-
--- Populate threadgate_allowed_lists
-UPDATE threadgate_allowed_lists tal SET
-    post_actor_id = tg.post_actor_id,
-    post_rkey = tg.post_rkey
-FROM threadgates tg
-WHERE tal.threadgate_id = tg.id;
-
--- Populate threadgate_hidden_replies
-UPDATE threadgate_hidden_replies thr SET
-    post_actor_id = tg.post_actor_id,
-    post_rkey = tg.post_rkey,
-    hidden_post_actor_id = p.actor_id,
-    hidden_post_rkey = p.rkey
-FROM threadgates tg, posts p
-WHERE thr.threadgate_id = tg.id
-    AND thr.post_id = p.id;
-
--- Populate postgate_detached
-UPDATE postgate_detached pd SET
-    post_actor_id = pg.post_actor_id,
-    post_rkey = pg.post_rkey,
-    detached_post_actor_id = p.actor_id,
-    detached_post_rkey = p.rkey
-FROM postgates pg, posts p
-WHERE pd.postgate_id = pg.id
-    AND pd.detached_post_id = p.id;
-
--- ============================================================================
--- Phase 3: Drop all foreign key constraints
--- ============================================================================
-
--- Child tables first (depend on parent tables)
-ALTER TABLE threadgate_allowed_lists DROP CONSTRAINT IF EXISTS threadgate_allowed_lists_threadgate_id_fkey;
-ALTER TABLE threadgate_allowed_lists DROP CONSTRAINT IF EXISTS threadgate_allowed_lists_list_id_fkey;
-ALTER TABLE threadgate_hidden_replies DROP CONSTRAINT IF EXISTS threadgate_hidden_replies_threadgate_id_fkey;
-ALTER TABLE postgate_detached DROP CONSTRAINT IF EXISTS postgate_detached_postgate_id_fkey;
-
--- Main tables
-ALTER TABLE post_likes DROP CONSTRAINT IF EXISTS post_likes_post_id_fkey;
-ALTER TABLE post_likes DROP CONSTRAINT IF EXISTS post_likes_via_repost_id_fkey;
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_post_id_fkey;
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_via_repost_id_fkey;
-ALTER TABLE post_aggregate_stats DROP CONSTRAINT IF EXISTS post_aggregate_stats_post_id_fkey;
-ALTER TABLE threadgates DROP CONSTRAINT IF EXISTS threadgates_post_id_fkey;
-ALTER TABLE postgates DROP CONSTRAINT IF EXISTS postgates_post_id_fkey;
-ALTER TABLE profiles DROP CONSTRAINT IF EXISTS profiles_pinned_post_id_fkey;
-ALTER TABLE statuses DROP CONSTRAINT IF EXISTS statuses_embed_post_id_fkey;
-ALTER TABLE posts DROP CONSTRAINT IF EXISTS posts_parent_post_id_fkey;
-ALTER TABLE posts DROP CONSTRAINT IF EXISTS posts_root_post_id_fkey;
-ALTER TABLE posts DROP CONSTRAINT IF EXISTS posts_embedded_post_id_fkey;
-
--- ============================================================================
--- Phase 4: Drop old columns and primary keys
--- ============================================================================
-
--- Drop old primary keys that will be replaced
-ALTER TABLE threadgates DROP CONSTRAINT IF EXISTS threadgates_pkey;
-ALTER TABLE postgates DROP CONSTRAINT IF EXISTS postgates_pkey;
-ALTER TABLE post_aggregate_stats DROP CONSTRAINT IF EXISTS post_aggregate_stats_pkey;
-ALTER TABLE posts DROP CONSTRAINT posts_pkey;
-
--- Drop primary keys from child tables
-ALTER TABLE threadgate_allowed_lists DROP CONSTRAINT IF EXISTS threadgate_allowed_lists_pkey;
-ALTER TABLE threadgate_hidden_replies DROP CONSTRAINT IF EXISTS threadgate_hidden_replies_pkey;
-ALTER TABLE postgate_detached DROP CONSTRAINT IF EXISTS postgate_detached_pkey;
-
--- Drop unique constraints that will change
-ALTER TABLE post_likes DROP CONSTRAINT IF EXISTS post_likes_actor_id_post_id_key;
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_actor_id_post_id_key;
-
--- Drop old columns (ORDER MATTERS - children before parents)
-ALTER TABLE threadgate_allowed_lists DROP COLUMN threadgate_id;
-ALTER TABLE threadgate_hidden_replies DROP COLUMN threadgate_id;
-ALTER TABLE threadgate_hidden_replies DROP COLUMN post_id;
-ALTER TABLE postgate_detached DROP COLUMN postgate_id;
-ALTER TABLE postgate_detached DROP COLUMN detached_post_id;
-
-ALTER TABLE posts DROP COLUMN parent_post_id;
-ALTER TABLE posts DROP COLUMN root_post_id;
-ALTER TABLE posts DROP COLUMN embedded_post_id;
-ALTER TABLE posts DROP COLUMN id;  -- Drop synthetic id
-
-ALTER TABLE post_likes DROP COLUMN post_id;
-ALTER TABLE post_likes DROP COLUMN via_repost_id;
-
-ALTER TABLE reposts DROP COLUMN post_id;
-ALTER TABLE reposts DROP COLUMN via_repost_id;
-ALTER TABLE reposts DROP COLUMN id;  -- Can now drop synthetic id!
-
-ALTER TABLE post_aggregate_stats DROP COLUMN post_id;
-ALTER TABLE threadgates DROP COLUMN id;
-ALTER TABLE threadgates DROP COLUMN post_id;
-ALTER TABLE postgates DROP COLUMN id;
-ALTER TABLE postgates DROP COLUMN post_id;
-ALTER TABLE profiles DROP COLUMN pinned_post_id;
-ALTER TABLE statuses DROP COLUMN embed_post_id;
-
--- ============================================================================
--- Phase 5: Create new primary keys
--- ============================================================================
-
--- Main tables
-ALTER TABLE posts ADD PRIMARY KEY (actor_id, rkey);
-ALTER TABLE reposts ADD PRIMARY KEY (actor_id, rkey);  -- NEW! Was id before
-ALTER TABLE post_aggregate_stats ADD PRIMARY KEY (post_actor_id, post_rkey);
-ALTER TABLE threadgates ADD PRIMARY KEY (post_actor_id, post_rkey);
-ALTER TABLE postgates ADD PRIMARY KEY (post_actor_id, post_rkey);
-
--- Child tables
-ALTER TABLE threadgate_allowed_lists ADD PRIMARY KEY (post_actor_id, post_rkey, list_id);
-ALTER TABLE threadgate_hidden_replies ADD PRIMARY KEY (post_actor_id, post_rkey, hidden_post_actor_id, hidden_post_rkey);
-ALTER TABLE postgate_detached ADD PRIMARY KEY (post_actor_id, post_rkey, detached_post_actor_id, detached_post_rkey);
-
--- ============================================================================
--- Phase 6: Create indexes on new natural key columns
--- ============================================================================
-
--- Posts table self-reference indexes
-DROP INDEX IF EXISTS idx_posts_parent;
-CREATE INDEX idx_posts_parent
-    ON posts(parent_post_actor_id, parent_post_rkey)
-    WHERE parent_post_actor_id IS NOT NULL;
-
-DROP INDEX IF EXISTS idx_posts_root;
-CREATE INDEX idx_posts_root
-    ON posts(root_post_actor_id, root_post_rkey)
-    WHERE root_post_actor_id IS NOT NULL;
-
-DROP INDEX IF EXISTS idx_posts_embedded_post_id;
-CREATE INDEX idx_posts_embedded_post
-    ON posts(embedded_post_actor_id, embedded_post_rkey)
-    WHERE embedded_post_actor_id IS NOT NULL;
-
--- post_likes indexes
-DROP INDEX IF EXISTS idx_post_likes_post_id;
-CREATE INDEX idx_post_likes_post_actor_id_post_rkey
-    ON post_likes(post_actor_id, post_rkey);
-
--- reposts indexes
-DROP INDEX IF EXISTS idx_reposts_post_id;
-CREATE INDEX idx_reposts_post_actor_id_post_rkey
-    ON reposts(post_actor_id, post_rkey);
-
--- ============================================================================
--- Phase 7: Recreate unique constraints
--- ============================================================================
-
--- reposts: One repost per actor per post
-ALTER TABLE reposts ADD CONSTRAINT reposts_actor_id_post_actor_id_post_rkey_key
-    UNIQUE (actor_id, post_actor_id, post_rkey);
-
--- post_likes: TimescaleDB hypertables require partition key in ALL unique constraints
--- post_likes is partitioned by rkey, so we CANNOT add UNIQUE(actor_id, post_actor_id, post_rkey)
--- Application must ensure one actor doesn't like the same post twice
--- (Removed: UNIQUE constraint not possible)
-
--- ============================================================================
--- Phase 8: Add foreign key constraints (where allowed)
--- ============================================================================
-
--- NOTE: TimescaleDB FK constraint rules (recent versions):
--- - Hypertable → Hypertable: NOT ALLOWED (posts → posts, post_likes → posts)
--- - Regular → Hypertable: ALLOWED (reposts → posts, threadgates → posts)
--- - Hypertable → Regular: ALLOWED (posts → actors, post_likes → actors)
-
--- Regular tables CAN have FK constraints to posts (a hypertable)
--- This is explicitly supported in recent TimescaleDB versions
-ALTER TABLE reposts ADD CONSTRAINT reposts_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE CASCADE;
-
--- reposts self-reference for via_repost (quote-repost-of-repost)
-ALTER TABLE reposts ADD CONSTRAINT reposts_via_repost_fkey
-    FOREIGN KEY (via_repost_actor_id, via_repost_rkey)
-    REFERENCES reposts(actor_id, rkey)
-    ON DELETE CASCADE;
-
-ALTER TABLE post_aggregate_stats ADD CONSTRAINT post_aggregate_stats_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE CASCADE;
-
-ALTER TABLE threadgates ADD CONSTRAINT threadgates_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE CASCADE;
-
-ALTER TABLE postgates ADD CONSTRAINT postgates_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE CASCADE;
-
-ALTER TABLE profiles ADD CONSTRAINT profiles_pinned_post_fkey
-    FOREIGN KEY (actor_id, pinned_post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE SET NULL;
-
--- Child tables reference their parent gate tables
-ALTER TABLE threadgate_allowed_lists ADD CONSTRAINT threadgate_allowed_lists_threadgate_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES threadgates(post_actor_id, post_rkey)
-    ON DELETE CASCADE;
-
--- Restore FK to lists table
-ALTER TABLE threadgate_allowed_lists ADD CONSTRAINT threadgate_allowed_lists_list_id_fkey
-    FOREIGN KEY (list_id)
-    REFERENCES lists(id)
-    ON DELETE CASCADE;
-
-ALTER TABLE threadgate_hidden_replies ADD CONSTRAINT threadgate_hidden_replies_threadgate_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES threadgates(post_actor_id, post_rkey)
-    ON DELETE CASCADE;
-
-ALTER TABLE postgate_detached ADD CONSTRAINT postgate_detached_postgate_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES postgates(post_actor_id, post_rkey)
-    ON DELETE CASCADE;
-
--- Referential integrity WITHOUT database FKs (application-level enforcement required):
--- 1. posts.parent_post_{actor_id,rkey} -> posts(actor_id,rkey) - NO FK (hypertable → hypertable not allowed)
--- 2. posts.root_post_{actor_id,rkey} -> posts(actor_id,rkey) - NO FK (hypertable → hypertable not allowed)
--- 3. posts.embedded_post_{actor_id,rkey} -> posts(actor_id,rkey) - NO FK (hypertable → hypertable not allowed)
--- 4. post_likes.{post_actor_id,post_rkey} -> posts(actor_id,rkey) - NO FK (hypertable → hypertable not allowed)
--- 5. post_likes.{via_repost_actor_id,via_repost_rkey} -> reposts(actor_id,rkey) - NO FK (would break if reposts becomes hypertable)
--- 6. threadgate_hidden_replies.{hidden_post_*} -> posts - NO FK (could be added, but optional/not critical)
--- 7. postgate_detached.{detached_post_*} -> posts - NO FK (could be added, but optional/not critical)
diff --git a/migrations/2025-11-18-213307_update_post_aggregate_stats_to_natural_keys/down.sql b/migrations/2025-11-18-213307_update_post_aggregate_stats_to_natural_keys/down.sql
deleted file mode 100644
index 22bffa30..00000000
--- a/migrations/2025-11-18-213307_update_post_aggregate_stats_to_natural_keys/down.sql
+++ /dev/null
@@ -1,18 +0,0 @@
--- Revert post_aggregate_stats back to using synthetic post_id
-
-DROP TABLE IF EXISTS post_aggregate_stats;
-
-CREATE TABLE post_aggregate_stats (
-    post_id BIGINT NOT NULL REFERENCES posts(id) ON DELETE CASCADE,
-    stat_type post_stat_type NOT NULL,
-    value INTEGER NOT NULL,
-    updated_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT now(),
-
-    PRIMARY KEY (post_id, stat_type)
-);
-
-CREATE INDEX idx_post_aggregate_stats_updated_at ON post_aggregate_stats(updated_at DESC)
-    WHERE value > 0;
-
-CREATE INDEX idx_post_aggregate_stats_hot_content ON post_aggregate_stats(stat_type, updated_at DESC, value DESC)
-    WHERE value > 10;
diff --git a/migrations/2025-11-18-213307_update_post_aggregate_stats_to_natural_keys/up.sql b/migrations/2025-11-18-213307_update_post_aggregate_stats_to_natural_keys/up.sql
deleted file mode 100644
index 2c951e8c..00000000
--- a/migrations/2025-11-18-213307_update_post_aggregate_stats_to_natural_keys/up.sql
+++ /dev/null
@@ -1,37 +0,0 @@
--- Update post_aggregate_stats to use natural keys (actor_id, rkey) instead of synthetic post_id
---
--- This migration changes the primary key from post_id to (actor_id, rkey) to align with
--- the natural key migration across the codebase.
-
--- Step 1: Drop the existing table (since it references posts(id) which will change)
--- This is safe because parakeet-index loads stats into memory on startup
-DROP TABLE IF EXISTS post_aggregate_stats;
-
--- Step 2: Recreate with natural key structure
-CREATE TABLE post_aggregate_stats (
-    actor_id INTEGER NOT NULL,
-    rkey BIGINT NOT NULL,
-    stat_type post_stat_type NOT NULL,
-    value INTEGER NOT NULL,
-    updated_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT now(),
-
-    PRIMARY KEY (actor_id, rkey, stat_type),
-
-    -- Foreign key constraint to posts using natural key
-    FOREIGN KEY (actor_id, rkey) REFERENCES posts(actor_id, rkey) ON DELETE CASCADE
-);
-
--- Step 3: Recreate indexes for query performance
-CREATE INDEX idx_post_aggregate_stats_updated_at ON post_aggregate_stats(updated_at DESC)
-    WHERE value > 0;
-
--- Index for hot content queries (recently updated high-value stats)
-CREATE INDEX idx_post_aggregate_stats_hot_content ON post_aggregate_stats(stat_type, updated_at DESC, value DESC)
-    WHERE value > 10;
-
-COMMENT ON TABLE post_aggregate_stats IS 'Post aggregate statistics using natural keys (actor_id, rkey)';
-COMMENT ON COLUMN post_aggregate_stats.actor_id IS 'Actor who created the post (natural key part 1)';
-COMMENT ON COLUMN post_aggregate_stats.rkey IS 'Record key as i64 (TID converted) - natural key part 2';
-COMMENT ON COLUMN post_aggregate_stats.stat_type IS 'Type of statistic (like, reply, repost, quote)';
-COMMENT ON COLUMN post_aggregate_stats.value IS 'Current count value';
-COMMENT ON COLUMN post_aggregate_stats.updated_at IS 'Last time value changed (NOT last write time)';
diff --git a/migrations/2025-11-18-225351_migrate_thread_mutes_to_natural_keys/down.sql b/migrations/2025-11-18-225351_migrate_thread_mutes_to_natural_keys/down.sql
deleted file mode 100644
index 498bd9c1..00000000
--- a/migrations/2025-11-18-225351_migrate_thread_mutes_to_natural_keys/down.sql
+++ /dev/null
@@ -1,18 +0,0 @@
--- Revert thread_mutes migration back to synthetic root_post_id
--- Note: This is a destructive rollback - existing thread mute data will be lost
-
--- Drop the foreign key constraint
-ALTER TABLE thread_mutes DROP CONSTRAINT IF EXISTS thread_mutes_root_post_fkey;
-
--- Drop the new primary key
-ALTER TABLE thread_mutes DROP CONSTRAINT thread_mutes_pkey;
-
--- Add back the old root_post_id column
-ALTER TABLE thread_mutes ADD COLUMN root_post_id BIGINT NOT NULL DEFAULT 0;
-
--- Drop the natural key columns
-ALTER TABLE thread_mutes DROP COLUMN root_post_actor_id;
-ALTER TABLE thread_mutes DROP COLUMN root_post_rkey;
-
--- Restore old primary key
-ALTER TABLE thread_mutes ADD PRIMARY KEY (actor_id, root_post_id);
diff --git a/migrations/2025-11-18-225351_migrate_thread_mutes_to_natural_keys/up.sql b/migrations/2025-11-18-225351_migrate_thread_mutes_to_natural_keys/up.sql
deleted file mode 100644
index 0aad3c79..00000000
--- a/migrations/2025-11-18-225351_migrate_thread_mutes_to_natural_keys/up.sql
+++ /dev/null
@@ -1,29 +0,0 @@
--- Migrate thread_mutes table from synthetic root_post_id to natural keys (root_post_actor_id, root_post_rkey)
--- This aligns with the posts table migration to natural keys
-
--- Add new natural key columns
-ALTER TABLE thread_mutes ADD COLUMN root_post_actor_id INTEGER;
-ALTER TABLE thread_mutes ADD COLUMN root_post_rkey BIGINT;
-
--- Populate new columns from existing root_post_id (if any data exists)
--- Note: Since posts table no longer has 'id' column, this migration assumes thread_mutes is currently empty
--- or that root_post_id values are no longer valid references
-
--- Drop the old primary key first (before dropping the column it depends on)
-ALTER TABLE thread_mutes DROP CONSTRAINT thread_mutes_pkey;
-
--- Drop the old root_post_id column
-ALTER TABLE thread_mutes DROP COLUMN root_post_id;
-
--- Make the new columns NOT NULL
-ALTER TABLE thread_mutes ALTER COLUMN root_post_actor_id SET NOT NULL;
-ALTER TABLE thread_mutes ALTER COLUMN root_post_rkey SET NOT NULL;
-
--- Create new primary key with natural keys
-ALTER TABLE thread_mutes ADD PRIMARY KEY (actor_id, root_post_actor_id, root_post_rkey);
-
--- Add foreign key constraint to posts table using natural keys
-ALTER TABLE thread_mutes ADD CONSTRAINT thread_mutes_root_post_fkey
-    FOREIGN KEY (root_post_actor_id, root_post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE CASCADE;
diff --git a/migrations/2025-11-18-232413_migrate_bookmarks_to_natural_keys/down.sql b/migrations/2025-11-18-232413_migrate_bookmarks_to_natural_keys/down.sql
deleted file mode 100644
index df3fb26d..00000000
--- a/migrations/2025-11-18-232413_migrate_bookmarks_to_natural_keys/down.sql
+++ /dev/null
@@ -1,10 +0,0 @@
--- Reverse the migration
-
-DROP INDEX IF EXISTS idx_bookmarks_post;
-ALTER TABLE bookmarks DROP CONSTRAINT IF EXISTS bookmarks_post_fkey;
-ALTER TABLE bookmarks DROP COLUMN IF EXISTS post_rkey;
-ALTER TABLE bookmarks DROP COLUMN IF EXISTS post_actor_id;
-
--- Restore old discriminated union columns
-ALTER TABLE bookmarks ADD COLUMN subject_type bookmark_subject_type;
-ALTER TABLE bookmarks ADD COLUMN subject_id BIGINT;
diff --git a/migrations/2025-11-18-232413_migrate_bookmarks_to_natural_keys/up.sql b/migrations/2025-11-18-232413_migrate_bookmarks_to_natural_keys/up.sql
deleted file mode 100644
index 6cfb8416..00000000
--- a/migrations/2025-11-18-232413_migrate_bookmarks_to_natural_keys/up.sql
+++ /dev/null
@@ -1,22 +0,0 @@
--- Migrate bookmarks to use natural keys for posts
--- Bookmarks can only reference posts (not feedgens, lists, etc.)
-
--- Add new natural key columns
-ALTER TABLE bookmarks ADD COLUMN post_actor_id INTEGER;
-ALTER TABLE bookmarks ADD COLUMN post_rkey BIGINT;
-
--- Drop old discriminated union columns (subject_type, subject_id)
-ALTER TABLE bookmarks DROP COLUMN subject_type;
-ALTER TABLE bookmarks DROP COLUMN subject_id;
-
--- Make new columns NOT NULL (all bookmarks must reference a post)
-ALTER TABLE bookmarks ALTER COLUMN post_actor_id SET NOT NULL;
-ALTER TABLE bookmarks ALTER COLUMN post_rkey SET NOT NULL;
-
--- Add foreign key constraint to posts using natural keys
-ALTER TABLE bookmarks ADD CONSTRAINT bookmarks_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES posts(actor_id, rkey) ON DELETE CASCADE;
-
--- Create index for better join performance
-CREATE INDEX idx_bookmarks_post ON bookmarks(post_actor_id, post_rkey);
diff --git a/migrations/2025-11-19-011902_drop_natural_key_fk_constraints/down.sql b/migrations/2025-11-19-011902_drop_natural_key_fk_constraints/down.sql
deleted file mode 100644
index 6f3cf69d..00000000
--- a/migrations/2025-11-19-011902_drop_natural_key_fk_constraints/down.sql
+++ /dev/null
@@ -1,14 +0,0 @@
--- This file should undo anything in `up.sql`
--- Re-add FK constraints (though this may fail if there are orphaned references)
-
--- Re-add FK constraint on reposts -> posts
-ALTER TABLE reposts ADD CONSTRAINT reposts_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey)
-    REFERENCES posts(actor_id, rkey)
-    ON DELETE CASCADE;
-
--- Re-add self-referencing FK constraint on reposts -> reposts (via_repost)
-ALTER TABLE reposts ADD CONSTRAINT reposts_via_repost_fkey
-    FOREIGN KEY (via_repost_actor_id, via_repost_rkey)
-    REFERENCES reposts(actor_id, rkey)
-    ON DELETE CASCADE;
diff --git a/migrations/2025-11-19-011902_drop_natural_key_fk_constraints/up.sql b/migrations/2025-11-19-011902_drop_natural_key_fk_constraints/up.sql
deleted file mode 100644
index ee2f026e..00000000
--- a/migrations/2025-11-19-011902_drop_natural_key_fk_constraints/up.sql
+++ /dev/null
@@ -1,14 +0,0 @@
--- Drop FK constraints on natural key references
--- With natural keys (actor_id, rkey), we don't need FK constraints since:
--- 1. We can reference posts/reposts by their natural key even if they don't exist yet
--- 2. TimescaleDB hypertables don't support compound FK constraints efficiently
--- 3. This eliminates the need for stub creation during indexing
-
--- Drop FK constraint on reposts -> posts
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_post_fkey;
-
--- Drop self-referencing FK constraint on reposts -> reposts (via_repost)
-ALTER TABLE reposts DROP CONSTRAINT IF EXISTS reposts_via_repost_fkey;
-
--- Note: post_likes already has no FK constraint on (post_actor_id, post_rkey)
--- Note: We keep actor_id FK constraints since actors still use synthetic IDs
diff --git a/migrations/2025-11-19-015513_drop_post_fk_constraints/down.sql b/migrations/2025-11-19-015513_drop_post_fk_constraints/down.sql
deleted file mode 100644
index 4ca58752..00000000
--- a/migrations/2025-11-19-015513_drop_post_fk_constraints/down.sql
+++ /dev/null
@@ -1,17 +0,0 @@
--- Restore FK constraints (down migration)
--- Note: This will fail if there is data that violates the constraints
-
-ALTER TABLE bookmarks ADD CONSTRAINT bookmarks_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey) REFERENCES posts(actor_id, rkey) ON DELETE CASCADE;
-
-ALTER TABLE postgates ADD CONSTRAINT postgates_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey) REFERENCES posts(actor_id, rkey) ON DELETE CASCADE;
-
-ALTER TABLE threadgates ADD CONSTRAINT threadgates_post_fkey
-    FOREIGN KEY (post_actor_id, post_rkey) REFERENCES posts(actor_id, rkey) ON DELETE CASCADE;
-
-ALTER TABLE profiles ADD CONSTRAINT profiles_pinned_post_fkey
-    FOREIGN KEY (actor_id, pinned_post_rkey) REFERENCES posts(actor_id, rkey) ON DELETE SET NULL;
-
-ALTER TABLE thread_mutes ADD CONSTRAINT thread_mutes_root_post_fkey
-    FOREIGN KEY (root_post_actor_id, root_post_rkey) REFERENCES posts(actor_id, rkey) ON DELETE CASCADE;
diff --git a/migrations/2025-11-19-015513_drop_post_fk_constraints/up.sql b/migrations/2025-11-19-015513_drop_post_fk_constraints/up.sql
deleted file mode 100644
index cb3f77f6..00000000
--- a/migrations/2025-11-19-015513_drop_post_fk_constraints/up.sql
+++ /dev/null
@@ -1,13 +0,0 @@
--- Drop FK constraints that reference posts table
--- These constraints prevent backfilling when posts from other actors don't exist yet
--- With natural keys (actor_id, rkey), we can trust the data without FK validation
-
--- Drop main table FK constraints
-ALTER TABLE bookmarks DROP CONSTRAINT IF EXISTS bookmarks_post_fkey;
-ALTER TABLE postgates DROP CONSTRAINT IF EXISTS postgates_post_fkey;
-ALTER TABLE threadgates DROP CONSTRAINT IF EXISTS threadgates_post_fkey;
-ALTER TABLE profiles DROP CONSTRAINT IF EXISTS profiles_pinned_post_fkey;
-ALTER TABLE thread_mutes DROP CONSTRAINT IF EXISTS thread_mutes_root_post_fkey;
-
--- Note: TimescaleDB hypertable chunk constraints are managed automatically
--- and will be dropped when the main table constraint is dropped
diff --git a/migrations/2025-11-19-023830_optimize_viewer_states_indexes_and_constraints/down.sql b/migrations/2025-11-19-023830_optimize_viewer_states_indexes_and_constraints/down.sql
deleted file mode 100644
index 4fa3c931..00000000
--- a/migrations/2025-11-19-023830_optimize_viewer_states_indexes_and_constraints/down.sql
+++ /dev/null
@@ -1,15 +0,0 @@
--- Revert viewer states optimizations
-
--- Drop the viewer-first composite indexes
-DROP INDEX IF EXISTS idx_bookmarks_viewer_post;
-DROP INDEX IF EXISTS idx_reposts_viewer_post;
-DROP INDEX IF EXISTS idx_post_likes_viewer_post;
-
--- Revert NOT NULL constraints on reposts
--- Note: This doesn't restore NULL values, just removes the constraint
-ALTER TABLE reposts ALTER COLUMN post_rkey DROP NOT NULL;
-ALTER TABLE reposts ALTER COLUMN post_actor_id DROP NOT NULL;
-
--- Revert NOT NULL constraints on post_likes
-ALTER TABLE post_likes ALTER COLUMN post_rkey DROP NOT NULL;
-ALTER TABLE post_likes ALTER COLUMN post_actor_id DROP NOT NULL;
diff --git a/migrations/2025-11-19-023830_optimize_viewer_states_indexes_and_constraints/up.sql b/migrations/2025-11-19-023830_optimize_viewer_states_indexes_and_constraints/up.sql
deleted file mode 100644
index 2d0f2ed2..00000000
--- a/migrations/2025-11-19-023830_optimize_viewer_states_indexes_and_constraints/up.sql
+++ /dev/null
@@ -1,45 +0,0 @@
--- Optimize viewer states queries with NOT NULL constraints and viewer-first composite indexes
---
--- Changes:
--- 1. Set post_actor_id and post_rkey to NOT NULL on post_likes and reposts
---    (bookmarks already has them as NOT NULL)
--- 2. Add composite indexes with viewer_id first for efficient viewer-specific lookups
--- 3. These enable PostgreSQL to:
---    - Skip NULL checks in index scans
---    - Use smaller, more efficient indexes
---    - Seek directly to viewer's records before filtering by post
-
--- Step 1: Set NOT NULL constraints on post_likes
--- These should never be NULL since they reference the post being liked
-ALTER TABLE post_likes ALTER COLUMN post_actor_id SET NOT NULL;
-ALTER TABLE post_likes ALTER COLUMN post_rkey SET NOT NULL;
-
--- Step 2: Set NOT NULL constraints on reposts
--- These should never be NULL since they reference the post being reposted
-ALTER TABLE reposts ALTER COLUMN post_actor_id SET NOT NULL;
-ALTER TABLE reposts ALTER COLUMN post_rkey SET NOT NULL;
-
--- Step 3: Add viewer-first composite indexes for efficient viewer state lookups
--- Pattern: (viewer_actor_id, post_actor_id, post_rkey)
--- This allows PostgreSQL to:
---   1. Seek directly to the viewer's records using actor_id
---   2. Filter by specific posts within that viewer's subset
--- Much faster than: seeking by post first, then filtering by viewer
---
--- Note: TimescaleDB hypertables don't support CONCURRENTLY, but index creation
--- is generally fast for hypertables as it operates on chunks
-
--- Post likes: viewer lookups are very common in getAuthorFeed
-CREATE INDEX idx_post_likes_viewer_post
-    ON post_likes (actor_id, post_actor_id, post_rkey);
-
--- Reposts: same access pattern (not a hypertable, but removing CONCURRENTLY for consistency)
-CREATE INDEX idx_reposts_viewer_post
-    ON reposts (actor_id, post_actor_id, post_rkey);
-
--- Bookmarks: same access pattern (not a hypertable)
-CREATE INDEX idx_bookmarks_viewer_post
-    ON bookmarks (actor_id, post_actor_id, post_rkey);
-
--- Note: postgates already has (post_actor_id, post_rkey) as PK which is sufficient
--- since we don't filter by viewer on postgates (they're post-level, not viewer-specific)
diff --git a/migrations/2025-11-19-025200_enable_chunk_skipping_for_timeline_queries/down.sql b/migrations/2025-11-19-025200_enable_chunk_skipping_for_timeline_queries/down.sql
deleted file mode 100644
index 19c5478c..00000000
--- a/migrations/2025-11-19-025200_enable_chunk_skipping_for_timeline_queries/down.sql
+++ /dev/null
@@ -1,11 +0,0 @@
--- Disable chunk skipping for timeline queries
-
--- Disable chunk skipping on specific columns
-SELECT disable_chunk_skipping('post_likes', 'actor_id');
-SELECT disable_chunk_skipping('posts', 'actor_id');
-
--- Disable chunk skipping globally
-ALTER SYSTEM SET timescaledb.enable_chunk_skipping = off;
-
--- Reload configuration to apply the setting
-SELECT pg_reload_conf();
diff --git a/migrations/2025-11-19-025200_enable_chunk_skipping_for_timeline_queries/up.sql b/migrations/2025-11-19-025200_enable_chunk_skipping_for_timeline_queries/up.sql
deleted file mode 100644
index 8f5ba7d4..00000000
--- a/migrations/2025-11-19-025200_enable_chunk_skipping_for_timeline_queries/up.sql
+++ /dev/null
@@ -1,38 +0,0 @@
--- Enable TimescaleDB chunk skipping for timeline queries
---
--- Chunk skipping tracks min/max values of actor_id per chunk, allowing PostgreSQL
--- to skip entire chunks when querying for a specific actor's timeline.
---
--- This is superior to a traditional index on (actor_id, rkey DESC) because:
--- 1. No large index overhead (uses lightweight per-chunk statistics)
--- 2. Automatically excludes chunks where actor_id doesn't exist
--- 3. Works with the existing time-based partitioning
--- 4. Perfect for correlated columns (actor_id correlates with rkey/time)
---
--- Expected performance: 50-90% reduction in chunks scanned for timeline queries
-
--- Step 1: Enable chunk skipping globally in postgresql.conf
--- This persists across PostgreSQL restarts
-ALTER SYSTEM SET timescaledb.enable_chunk_skipping = on;
-
--- Step 2: Reload configuration to apply the setting immediately
-SELECT pg_reload_conf();
-
--- Step 3: Enable chunk skipping on actor_id for posts table
--- This allows queries like "SELECT * FROM posts WHERE actor_id = X ORDER BY rkey DESC"
--- to skip chunks that don't contain that actor's posts
-SELECT enable_chunk_skipping('posts', 'actor_id');
-
--- Step 4: Enable chunk skipping on actor_id for post_likes table
--- This allows viewer state queries to skip chunks without the viewer's likes
-SELECT enable_chunk_skipping('post_likes', 'actor_id');
-
--- Note: Chunk skipping only applies to NEW chunks compressed AFTER enabling this feature.
--- Existing chunks won't benefit until they're recompressed.
---
--- To verify chunk skipping is working, check the chunk_column_stats catalog:
---   SELECT * FROM _timescaledb_catalog.chunk_column_stats
---   WHERE hypertable_id IN (
---     SELECT id FROM _timescaledb_catalog.hypertable
---     WHERE table_name IN ('posts', 'post_likes')
---   );
diff --git a/migrations/2025-11-19-025652_drop_post_aggregate_stats_fk_and_compress_posts/down.sql b/migrations/2025-11-19-025652_drop_post_aggregate_stats_fk_and_compress_posts/down.sql
deleted file mode 100644
index 610ef524..00000000
--- a/migrations/2025-11-19-025652_drop_post_aggregate_stats_fk_and_compress_posts/down.sql
+++ /dev/null
@@ -1,35 +0,0 @@
--- Revert FK drop and decompression
---
--- WARNING: This will decompress all posts chunks, which may take significant time
--- and storage space. Decompression is generally not recommended in production.
-
--- Step 1: Decompress all posts chunks
-DO $$
-DECLARE
-    chunk_record RECORD;
-    decompressed_count INTEGER := 0;
-BEGIN
-    FOR chunk_record IN
-        SELECT chunk_schema, chunk_name
-        FROM timescaledb_information.chunks
-        WHERE hypertable_name = 'posts'
-          AND is_compressed
-        ORDER BY chunk_name
-    LOOP
-        EXECUTE format('SELECT decompress_chunk(%L)',
-                      chunk_record.chunk_schema || '.' || chunk_record.chunk_name);
-        decompressed_count := decompressed_count + 1;
-
-        IF decompressed_count % 10 = 0 THEN
-            RAISE NOTICE 'Decompressed % posts chunks...', decompressed_count;
-        END IF;
-    END LOOP;
-
-    RAISE NOTICE 'Decompression complete: % posts chunks decompressed', decompressed_count;
-END $$;
-
--- Step 2: Re-add the foreign key constraint
--- Note: This will fail if there are orphaned rows in post_aggregate_stats
-ALTER TABLE post_aggregate_stats
-    ADD CONSTRAINT post_aggregate_stats_actor_id_rkey_fkey
-    FOREIGN KEY (actor_id, rkey) REFERENCES posts(actor_id, rkey);
diff --git a/migrations/2025-11-19-025652_drop_post_aggregate_stats_fk_and_compress_posts/up.sql b/migrations/2025-11-19-025652_drop_post_aggregate_stats_fk_and_compress_posts/up.sql
deleted file mode 100644
index 2e1ebb90..00000000
--- a/migrations/2025-11-19-025652_drop_post_aggregate_stats_fk_and_compress_posts/up.sql
+++ /dev/null
@@ -1,61 +0,0 @@
--- Drop FK constraint and compress posts chunks to enable chunk skipping
---
--- Background:
--- TimescaleDB cannot compress chunks that have foreign keys pointing into them.
--- The post_aggregate_stats table has an FK to posts, preventing compression.
---
--- Solution:
--- 1. Drop the FK constraint (app-level integrity is sufficient)
--- 2. Compress all posts chunks to enable chunk skipping on actor_id
---
--- Expected benefits:
--- - Timeline queries skip 90%+ of chunks (only scan chunks with that actor's posts)
--- - Stabilizes query time variance (currently 28-125ms → expected 10-40ms)
--- - Reduces I/O and buffer cache pressure
--- - Saves storage with compression
-
--- Step 1: Drop the foreign key constraint
-ALTER TABLE post_aggregate_stats
-    DROP CONSTRAINT IF EXISTS post_aggregate_stats_actor_id_rkey_fkey;
-
--- Step 2: Compress all uncompressed posts chunks
--- This populates chunk skipping statistics for actor_id
-DO $$
-DECLARE
-    chunk_record RECORD;
-    compressed_count INTEGER := 0;
-BEGIN
-    FOR chunk_record IN
-        SELECT chunk_schema, chunk_name
-        FROM timescaledb_information.chunks
-        WHERE hypertable_name = 'posts'
-          AND NOT is_compressed
-        ORDER BY chunk_name
-    LOOP
-        EXECUTE format('SELECT compress_chunk(%L)',
-                      chunk_record.chunk_schema || '.' || chunk_record.chunk_name);
-        compressed_count := compressed_count + 1;
-
-        -- Log progress every 10 chunks
-        IF compressed_count % 10 = 0 THEN
-            RAISE NOTICE 'Compressed % posts chunks...', compressed_count;
-        END IF;
-    END LOOP;
-
-    RAISE NOTICE 'Compression complete: % posts chunks compressed', compressed_count;
-END $$;
-
--- Step 3: Verify chunk skipping statistics were populated
-DO $$
-DECLARE
-    stats_count INTEGER;
-BEGIN
-    SELECT COUNT(*) INTO stats_count
-    FROM _timescaledb_catalog.chunk_column_stats
-    WHERE hypertable_id = (
-        SELECT id FROM _timescaledb_catalog.hypertable WHERE table_name = 'posts'
-    )
-    AND column_name = 'actor_id';
-
-    RAISE NOTICE 'Chunk skipping statistics: % entries for posts.actor_id', stats_count;
-END $$;
diff --git a/parakeet-db/src/schema.rs b/parakeet-db/src/schema.rs
index 4adb45cc..ce50edd3 100644
--- a/parakeet-db/src/schema.rs
+++ b/parakeet-db/src/schema.rs
@@ -505,13 +505,13 @@ diesel::table! {
     use diesel::sql_types::*;
     use super::sql_types::PostgateRule;
 
-    postgates (post_actor_id, post_rkey) {
+    postgates (actor_id, rkey) {
         actor_id -> Int4,
         rkey -> Int8,
         cid -> Bytea,
+        post_actor_id -> Nullable<Int4>,
+        post_rkey -> Nullable<Int8>,
         rules -> Array<Nullable<PostgateRule>>,
-        post_actor_id -> Int4,
-        post_rkey -> Int8,
     }
 }
 
@@ -521,9 +521,9 @@ diesel::table! {
     use super::sql_types::EmbedType;
     use super::sql_types::PostStatus;
     use super::sql_types::PostExtEmbed;
+    use super::sql_types::PostVideoEmbed;
     use super::sql_types::PostImageEmbed;
     use super::sql_types::PostFacetEmbed;
-    use super::sql_types::PostVideoEmbed;
 
     posts (actor_id, rkey) {
         actor_id -> Int4,
@@ -532,12 +532,19 @@ diesel::table! {
         content -> Nullable<Bytea>,
         langs -> Array<Nullable<LanguageCode>>,
         tags -> Array<Nullable<Text>>,
+        tokens -> Nullable<Array<Nullable<Text>>>,
+        parent_post_actor_id -> Nullable<Int4>,
+        parent_post_rkey -> Nullable<Int8>,
+        root_post_actor_id -> Nullable<Int4>,
+        root_post_rkey -> Nullable<Int8>,
         embed_type -> Nullable<EmbedType>,
         embed_subtype -> Nullable<EmbedType>,
         violates_threadgate -> Bool,
         status -> PostStatus,
-        tokens -> Nullable<Array<Nullable<Text>>>,
         ext_embed -> Nullable<PostExtEmbed>,
+        video_embed -> Nullable<PostVideoEmbed>,
+        embedded_post_actor_id -> Nullable<Int4>,
+        embedded_post_rkey -> Nullable<Int8>,
         record_detached -> Nullable<Bool>,
         image_1 -> Nullable<PostImageEmbed>,
         image_2 -> Nullable<PostImageEmbed>,
@@ -552,13 +559,6 @@ diesel::table! {
         facet_7 -> Nullable<PostFacetEmbed>,
         facet_8 -> Nullable<PostFacetEmbed>,
         mentions -> Nullable<Array<Nullable<Int4>>>,
-        video_embed -> Nullable<PostVideoEmbed>,
-        parent_post_actor_id -> Nullable<Int4>,
-        parent_post_rkey -> Nullable<Int8>,
-        root_post_actor_id -> Nullable<Int4>,
-        root_post_rkey -> Nullable<Int8>,
-        embedded_post_actor_id -> Nullable<Int4>,
-        embedded_post_rkey -> Nullable<Int8>,
     }
 }
 
@@ -574,11 +574,11 @@ diesel::table! {
         banner_cid -> Nullable<Bytea>,
         display_name -> Nullable<Text>,
         description -> Nullable<Text>,
+        pinned_post_rkey -> Nullable<Int8>,
         joined_sp_id -> Nullable<Int8>,
         pronouns -> Nullable<Text>,
         website -> Nullable<Text>,
         search_vector -> Nullable<Tsvector>,
-        pinned_post_rkey -> Nullable<Int8>,
     }
 }
 
@@ -589,12 +589,11 @@ diesel::table! {
     reposts (actor_id, rkey) {
         actor_id -> Int4,
         rkey -> Int8,
-        cid -> Bytea,
-        status -> RepostStatus,
         post_actor_id -> Int4,
         post_rkey -> Int8,
         via_repost_actor_id -> Nullable<Int4>,
         via_repost_rkey -> Nullable<Int8>,
+        status -> RepostStatus,
     }
 }
 
@@ -650,27 +649,27 @@ diesel::table! {
         created_at -> Timestamptz,
         status -> StatusType,
         duration -> Nullable<Int4>,
-        thumb_mime_type -> Nullable<ImageMimeType>,
-        thumb_cid -> Nullable<Bytea>,
         embed_post_actor_id -> Nullable<Int4>,
         embed_post_rkey -> Nullable<Int8>,
+        thumb_mime_type -> Nullable<ImageMimeType>,
+        thumb_cid -> Nullable<Bytea>,
     }
 }
 
 diesel::table! {
     thread_mutes (actor_id, root_post_actor_id, root_post_rkey) {
         actor_id -> Int4,
-        created_at -> Timestamptz,
         root_post_actor_id -> Int4,
         root_post_rkey -> Int8,
+        created_at -> Timestamptz,
     }
 }
 
 diesel::table! {
     threadgate_allowed_lists (post_actor_id, post_rkey, list_id) {
-        list_id -> Int8,
         post_actor_id -> Int4,
         post_rkey -> Int8,
+        list_id -> Int8,
     }
 }
 
@@ -687,13 +686,13 @@ diesel::table! {
     use diesel::sql_types::*;
     use super::sql_types::ThreadgateRule;
 
-    threadgates (post_actor_id, post_rkey) {
+    threadgates (actor_id, rkey) {
         actor_id -> Int4,
         rkey -> Int8,
         cid -> Bytea,
+        post_actor_id -> Nullable<Int4>,
+        post_rkey -> Nullable<Int8>,
         allow -> Nullable<Array<Nullable<ThreadgateRule>>>,
-        post_actor_id -> Int4,
-        post_rkey -> Int8,
     }
 }
 
@@ -722,7 +721,6 @@ diesel::joinable!(actor_aggregate_stats -> actors (actor_id));
 diesel::joinable!(bookmarks -> actors (actor_id));
 diesel::joinable!(chat_decls -> actors (actor_id));
 diesel::joinable!(feedgen_likes -> actors (actor_id));
-diesel::joinable!(feedgen_likes -> feedgens (feedgen_id));
 diesel::joinable!(labeler_defs -> actors (labeler_actor_id));
 diesel::joinable!(labelers -> actors (actor_id));
 diesel::joinable!(labels -> actors (labeler_actor_id));
@@ -744,6 +742,7 @@ diesel::joinable!(starterpack_feeds -> starterpacks (starterpack_id));
 diesel::joinable!(starterpacks -> lists (list_id));
 diesel::joinable!(statuses -> actors (actor_id));
 diesel::joinable!(thread_mutes -> actors (actor_id));
+diesel::joinable!(threadgate_allowed_lists -> lists (list_id));
 diesel::joinable!(threadgates -> actors (actor_id));
 
 diesel::allow_tables_to_appear_in_same_query!(