How shuiii129/agora-rag-ml-assessment works
This is a local RAG application for asking grounded questions about AI-governance documents, through either a web chat or the CLI. The query path embeds a question, retrieves relevant chunks from ChromaDB, and asks Gemini to produce a grounded answer with source references. A separate ingestion path reads document text and CSV metadata, cleans and chunks the text, embeds it, and stores it for later retrieval. The README documents the major capabilities; details of some web and status handling are based on partial excerpts.
Main entry points: ingest.py Ingestion Pipeline · main.py CLI and API · rag.py RAG Query · Data AGORA Dataset
Components 10
Application
- CLI and API
main.py - Web Chat UI
api/static_assets.py
Document Ingestion
- AGORA Dataset
Data - Metadata Loader
src/ingest.py - Clean and Chunk
src/ingest.py - Ingestion Pipeline
src/ingest.py
Retrieval and Storage
- RAG Query
src/rag.py - ChromaDB Index
chroma_db
External Integrations
- Gemini API
Other components
- User
Connections 13
- User → CLI and API: submits query
- CLI and API → Web Chat UI: serves UI
- CLI and API → Ingestion Pipeline: runs ingestion
- CLI and API → RAG Query: runs query
- AGORA Dataset → Ingestion Pipeline: provides documents
- AGORA Dataset → Metadata Loader: provides CSV
- Ingestion Pipeline → Metadata Loader: loads metadata
- Ingestion Pipeline → Clean and Chunk: cleans and chunks
- Ingestion Pipeline → Gemini API: requests embeddings
- Ingestion Pipeline → ChromaDB Index: stores chunks
- RAG Query → Gemini API: embeds and generates
- RAG Query → ChromaDB Index: retrieves chunks
- CLI and API → User: returns answer and references