v0.2.0

v0.2.0 (2024-11-25)

Chores

• Switch to fastapi-cli for development (8a1bc38)

• Add fastapi-cli to dev dependencies

• Update run-chatbot command to use fastapi CLI

• Fix test watch path to match new project structure

• Enhance Makefile and hooks for better semantic-release integration; update README for eMush RAG Chatbot repo rebranding (692ffc5)

Documentation

• Improve README and development setup (2e76ffc)

• Add watch API terminal to VSCode configuration

• Update README:

  • Clarify data sources (wikis, tutorials, forums)
  • Add section about improving RAG performance
  • Remove manual env file creation (now automated)
  • Update example query

• Fix vector store import in index_documents script

Features

• Add a streamlit app (886509c)

Refactoring

• Flatten project structure (138fa9b)

• Move core modules out of src/ directory:

  • Move chat_api.py, llm.py, rag_chain.py, and vector_store.py to root package
  • Update imports in tests and scripts

• Update paths in Makefile for scripts

• Improve README:

  • Fix typos and formatting
  • Add evaluation section
  • Make database and data links more readable

• Improve project structure and development setup (b8dc551)

• Reorganize code with Protocol-based vector store abstraction

• Add FakeVectorStore for testing purposes

• Implement lazy initialization for RAG chain

• Update development tooling:

  • Configure Ruff for import sorting
  • Add pytest-asyncio for better test support
  • Set up automatic env file creation
  • Improve test watching configuration

v0.1.0

v0.1.0 (2024-11-24)

Bug Fixes

• Make document indexing asynchronous with await for vector store (75ac781)

Documentation

• Update README with correct command to run the application (0c89bf9)

• Update README to use uv for running Python scripts (e4a1c3c)

• Update README with detailed project description and usage instructions (874be74)

Features

• Add initial dummy test in test_main.py for basic validation of test structure (1ba149a)

• Refactor chat API to remove unnecessary filter_metadata; update vector store and dependencies for better performance and stability (7bf039f)

• Update CHAT_MODEL to "gpt-4o-mini" and switch evaluation dataset to "test_set_v3.csv" for improved chatbot performance (6e536bc)

• Update chat model to "gpt-4o" and amend evaluation results for consistency in eMush chatbot configuration and responses (20f3645)

• Add prompt template for version V6 to enhance context handling and reasoning in eMush chatbot responses (50e36e8)

• Remove redundant import of app in main.py for cleaner entry point structure (659ecde)

• Update prompt templates in RAGChain for versions V4 and V5 with enhanced reasoning requirements (a0b68f1)

• Enhance RAGChain to utilize versioned prompt templates for improved context handling (9aa4b98)

• Allow configuring top_k for document retrieval in RAGChain (6d9d827)

• Save evaluation results to JSON with metadata and nested structure (1a81f37)

• Add question reformulation step to RAG chain for improved context retrieval (990008d)

• Enhance evaluation results storage with UUID, timestamp, and CSV export (068135d)

• Add new Makefile targets for RAG evaluation, document indexing, and chatbot execution scripts (e1a761d)

• Enhance RAG evaluation script with improved response evaluation and add test data for comprehensive assessments (905beab)

• Update DocumentLoader to allow larger chunk sizes and improved document splitting logic for long texts (8239850)

• Add RAG evaluation script with LLM-as-a-judge functionality (f471855)

• Modify chat API to return source documents with response (ab52c94)

• Add batch processing and tqdm progress bar for document indexing (de3a224)

• Restore batch processing for document indexing (5cccc32)

• Add RecursiveCharacterTextSplitter to handle long document chunks (01261d9)

• Update document loader to show loading progress and change batch size from 10 to 8 (d272ad1)

• Return response with source documents in RAG chain (de0351a)

• Add document batching to avoid OpenAI API rate limits (56bf522)

• Add title to document metadata in document loader (ad07daa)

• Add script to index documents from Mushpedia JSON into vector store (78dd265)

• Implement RAG chatbot for eMush game with vector store and FastAPI endpoint (102a65a)

This commit adds the core implementation of the RAG-powered chatbot for the eMush game, including:

• Vector store management with Chroma
• RAG chain for context-aware responses
• FastAPI endpoint for chat interactions
• Logging and error handling
• Modular architecture with separate components for vector store, RAG chain, and API

The implementation supports:

• Semantic document search

• Chat history context

• Metadata filtering

• Async processing

• Configurable LLM and embedding models

• Implement vector store for document indexing and similarity search using Chroma and OpenAI embeddings (388230a)

Refactoring

• Modify evaluation script to append results to existing CSV file (b4e8747)

• Remove reformulation step from RAG chain (2f98805)

• Enhance question reformulation strategy for RAG retrieval (c6692ac)

• Remove batch processing from document loading and indexing (a76d130)

• Remove batch processing in index_documents.py (4c7e51a)

• Modify document loading to remove batch return parameter (25e587f)

• Remove batch size logic from document loader (503d418)

• Adjust text splitter chunk overlap and remove custom separators (5d22ca9)

• Simplify RAG response to return only text response (51a1845)