- litellm had dependency conflicts with pydantic-ai - tiktoken is lighter and already required by pydantic-ai - Updated documentation (README.md, architecture.md) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- SQLAlchemy async database layer (SQLite dev, PostgreSQL prod) - Conversation and Message models with UUID primary keys - Token counting utilities using litellm - Context summarization at 80% token threshold - REST API endpoints for multi-turn conversations - 19 conversation tests, 6 token tests (176 total passing) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>