Update Docker configuration and add new client and inference scripts master
- Updated `docker-compose.yml` to use the `microsoft/Phi-4` model for improved performance. - Introduced `modal_client.py`, a new script for connecting to the Comind Modal inference server, demonstrating usage of the OpenAI client library. - Added `modal_inference.py` to set up the Modal application with vLLM, including model configuration and server setup. - Created documentation for prompt best practices in `prompts.md`. - Enhanced `cominds` prompts to align with the new core perspective structure. - Implemented additional methods in `record_manager.py` and `comind.py` for better sphere management and context handling. - Added a new utility module `format.py` for robust string formatting in the Comind project.