Feat/v0.5.0 streaming tenancy

[soneylegal]

This pull request introduces major enhancements for multi-tenancy, real-time streaming, and in-memory document ingestion, along with developer experience improvements for Docker-based workflows and documentation. The most significant changes are the addition of a streaming chat API, a document upload endpoint supporting .md and .pdf files with tenant isolation, and updates to the vector store and retrieval logic to support tenant-level partitioning. The Docker and Makefile workflows are also refactored to better support local and containerized development.

Major Feature Additions

• Added a real-time Server-Sent Events (SSE) streaming chat API at POST /api/v1/chat/stream, enabling token-by-token streaming of LLM output and final structured metadata, with support for tenant and session isolation. (src/app/api/v1/chat.py)
• Introduced a document ingestion API at POST /api/v1/documents to upload .md and .pdf files directly into tenant-partitioned ChromaDB vector stores, including PDF parsing and chunking logic. (src/app/api/v1/documents.py)

Multi-Tenancy and Vector Store Enhancements

• All vector store operations and semantic cache lookups are now partitioned by tenant_id, ensuring strict isolation between tenants for both retrieval and ingestion. (src/app/api/v1/chat.py, src/app/graph/nodes.py) [1] [2]
• Retrieval and grading nodes in the LangGraph workflow are now asynchronous and support tenant-aware document search. (src/app/graph/nodes.py) [1] [2]

Developer Experience and Docker Improvements

• Refactored the Dockerfile to use multi-stage builds for separate development and production images, and updated docker-compose.yml to mount source, test, and config files for live development. (Dockerfile, docker-compose.yml) [1] [2]
• Enhanced the Makefile with Docker-based quality check commands (docker-lint, docker-format, docker-typecheck, docker-test, docker-ci) for consistent local and CI workflows. (Makefile)

Documentation Updates

• Expanded the README.md and docs/developer-guide.md to document the new streaming and ingestion APIs, multi-tenancy, and Docker-based development workflows. [1] [2] [3]

Dependency Updates

• Added langchain-text-splitters, pypdf, and python-multipart to support PDF parsing and multipart uploads for document ingestion. (pyproject.toml) [1] [2]