How mithun50/ragbot works
RagBot lets users upload documents and ask grounded questions with passage citations. Uploads are parsed page by page, chunked, indexed, and summarized; questions are retrieved and ranked against that index, then answered by a streaming model call or refused when evidence is insufficient. The API also exposes document management and question analytics, presented through a single-page frontend. The README documents retrieval details and API capabilities; some frontend and retrieval internals were not sampled, so their links rely on those documented relationships.
Main entry points: main.py FastAPI Routes · ingest.py Background Ingestion · qa.py Grounded QA · index.py BM25 Document Store
Components 13
User Interface
- Web Frontend
static/app.js - FastAPI Routes
app/main.py
Document Ingestion
- Background Ingestion
app/ingest.py - Document Parsing
app/parsing.py - Page Chunking
app/chunking.py - Document Summaries
app/summarize.py
Question Answering
- Grounded QA
app/qa.py - Grounding Prompts
app/prompts.py
Index and Analytics
- BM25 Document Store
app/index.py - Question Analytics
app/analytics.py
Model Integration
- Model Client
app/llm.py
Other components
- User
- OpenRouter
Connections 18
- User → Web Frontend: uses
README.md - Web Frontend → FastAPI Routes: calls API
README.md - FastAPI Routes → Background Ingestion: starts ingestion
app/main.py - FastAPI Routes → BM25 Document Store: manages documents
app/main.py - FastAPI Routes → Grounded QA: asks question
app/main.py - FastAPI Routes → Question Analytics: serves analytics
app/main.py - Background Ingestion → Document Parsing: iterates pages
app/ingest.py - Background Ingestion → Page Chunking: chunks pages
app/ingest.py - Background Ingestion → BM25 Document Store: adds chunks
app/ingest.py - Background Ingestion → Document Summaries: builds summary
app/ingest.py - Document Summaries → BM25 Document Store: reads chunks
app/summarize.py - Document Summaries → Model Client: requests summary
app/summarize.py - Grounded QA → BM25 Document Store: retrieves passages
app/qa.py - Grounded QA → Grounding Prompts: formats context
app/qa.py - Grounded QA → Model Client: streams answer
app/qa.py - Grounded QA → Question Analytics: records question
app/qa.py - Question Analytics → BM25 Document Store: reads document stats
app/analytics.py - Model Client → OpenRouter: sends model requests
README.md