solstone — the journal #

a memory your agents can work from. sol — the app on your devices — experiences your day with you and keeps it all in your journal. your journal is always private, only yours.

this repo is the journal: the memory that holds everything, plus the thin sol access client. it's the python core of the solstone platform — the sol apps on your devices pair with a journal running on a computer you choose. AI agents transcribe, extract entities, detect meetings, build knowledge graphs, and surface daily insights — all without any manual input. everything stays in daily journal directories on your machine. open source, local-first, no cloud required.

Python 3.11+, Linux + macOS, AGPL-3.0-only, maintained by sol pbc.

Daily dashboard — goal, todos, upcoming events, and detected entities, all generated from observations. Facet tabs organize your life by project or context.

what you get #

a system of intelligence, not just storage.

• automatic transcription — sol hears what you hear and keeps every conversation in your journal, transcribed with speaker identification and searchable.
• people and projects — extracted from your conversations and remembered across time.
• knowledge graphs — relationships between entities mapped automatically. who works with whom, which projects connect to which people.
• meeting detection — meetings identified, summarized, and linked. meeting prep that surfaces what you discussed last time and personal context you'd forget.
• commitments — todos extracted from natural conversation. no manual entry.
• facet organization — group everything by project or context (work, personal, client-name) with scoped views across all apps.
• AI chat — talk to your journal. ask anything about your digital life and get answers grounded in your actual data.
• full-text search — find anything you've ever seen or heard.
• 30 AI agents — configurable workflows for activities, scheduling, research, media analysis, and more. extensible via the agent skill framework.
• local-first — all data in daily journal directories on your filesystem. configurable AI providers (Google Gemini, OpenAI, Anthropic). no cloud dependency.

Transcript viewer — dual-timeline navigation, speaker-diarized dialogue, audio playback, screen analysis. every conversation browsable by time.

People and projects — automatically extracted and remembered across your journal with mention counts and relationship data.

architecture #

  +---------+       +----------------+       +---------+
  | observe | ----> |    journal     | ----> |  think  |
  | inputs  |       | YYYYMMDD/ dirs |       | process |
  +---------+       | media, jsonl,  |       | index   |
                    | entities       |       +----+----+
                    +-------+--------+            |
                            ^                     |
                            |  agent outputs      |
                       +----+----+                |
                       | cortex  | <--------------+
                       | agents  |
                       +---------+
                            |
  ==== callosum (event bus) | ==========================
                            |
                     +------+------+
                     |   convey    |
                     | web UI      |
                     +-------------+

• observe — receives audio and screen observations from standalone observers (solstone-linux, solstone-tmux, solstone-macos) via observer ingest. processes FLAC audio, WebM screen media, and timestamped metadata.
• think — transcribes audio (faster-whisper), analyzes screen observations, surfaces entities, detects meetings, and indexes everything into SQLite. runs 30 configurable agent/generator templates from solstone/talent/.
• cortex — orchestrates agent execution. receives events, dispatches agents, writes results back to the journal.
• callosum — async message bus connecting all services. enables event-driven coordination between observe, think, cortex, and convey.
• convey — Flask-based web interface with 17 pluggable apps for navigating journal data.
• journal — journal/YYYYMMDD/ daily directories. the single source of truth — transcripts, media, entities, agent outputs, and the SQLite index all live here.

quick start #

run a journal here — the full host:

uv tool install solstone-journal && uv tool install solstone
journal setup

pip and pipx equivalents, followed by journal setup:

pip install solstone-journal
pipx install solstone-journal && pipx install solstone

A pip install solstone-journal puts journal and mlx-vlm-server on PATH natively; uv tool and pipx expose each tool's own commands, so the journal install is two commands — the journal tool plus the thin solstone tool. For GPU transcription, install solstone-journal-cuda instead.

want only the thin sol client — to talk to a journal running elsewhere? uv tool install solstone (no extras), or uvx solstone for an ephemeral one-shot.

not sure a computer is up to running the journal? uvx solstone check gives a one-shot readiness verdict — GPU, memory, and disk — before you install anything.

then open http://localhost:5015 in a browser; the first-run wizard sets up your identity and gets sol thinking — locally by default, or on your own provider key if you'd rather use a cloud lane.

see INSTALL.md for prerequisites, observer install, and troubleshooting; see CONTRIBUTING.md if you want to develop on solstone from a source checkout.

CLI #

solstone is operated through sol for day-to-day journal access and journal for host operations.

sol                    # Status overview and command list
journal supervisor         # Start the full stack (observe + processing + web)
sol chat               # Interactive AI chat from the terminal
journal transcribe <file>  # Transcribe an audio file
journal indexer            # Rebuild the search index

Run sol help for the full command reference.

documentation #

development #

See AGENTS.md for development guidelines, coding standards, and testing instructions.

Use make dev to run the full stack against test fixtures and make ci for pre-commit checks.

feedback #

Questions, feedback, or a bug? Follow and tag @solstone.app on Bluesky for discussion and updates, open an issue at github.com/solpbc/solstone-journal/issues for bugs, or reach support at support.solstone.app. You don't need to know anyone — those are the front doors.

contributing #

See CONTRIBUTING.md for contribution terms.

license #

AGPL-3.0-only. See LICENSE for details. Bundled third-party model notices: solstone/THIRD_PARTY_NOTICES.md. Maintained by sol pbc.