anilch963/research-intelligence-platform

How anilch963/research-intelligence-platform works

Research Intelligence is a document-grounded RAG platform: users submit a question, choose vector or hybrid retrieval, and receive an answer with inline citations and source snippets. Documents are loaded and chunked into a persistent Chroma collection; query handling checks index availability, retrieves context, and calls the chat model. The repository also includes a RAGAS evaluation workflow and draft-testset generation. The README documents these capabilities, but evaluation internals beyond the sampled runner and draft generator are not represented in detail.

Main entry points: retrievers.py Retriever Selection · rag.py RAG Answering · run_ragas.py RAGAS Runner · ingest.py Document Ingestion

Components 19

Question Interface

Research UI frontend/src/App.jsx
API Client frontend/src/api.js

Query Answering

FastAPI API backend/app/main.py
RAG Answering backend/app/rag.py
Retriever Selection backend/app/retrievers.py
BM25 Search backend/app/retrievers.py
Vector Search backend/app/retrievers.py
OpenAI Chat Model
Runtime Settings backend/app/config.py

Document Indexing

Chroma Chunk Store backend/app/retrievers.py
OpenAI Embeddings backend/app/retrievers.py
Document Ingestion backend/app/ingest.py
Source Documents backend/data/docs

RAG Evaluation

RAGAS Runner backend/eval/run_ragas.py
Testset Drafter backend/eval/draft_testset.py
Evaluation Testset backend/eval/testset.json
Evaluation Results backend/eval/results
Results Comparison backend/eval/compare.py

Other components

Researcher

Connections 24

Researcher → Research UI: asks questions frontend/src/App.jsx
Research UI → API Client: calls frontend/src/App.jsx
API Client → FastAPI API: sends query README.md
FastAPI API → RAG Answering: answers with backend/app/main.py
FastAPI API → Retriever Selection: checks chunks backend/app/main.py
RAG Answering → Retriever Selection: retrieves context backend/app/rag.py
Retriever Selection → BM25 Search: selects hybrid backend/app/retrievers.py
Retriever Selection → Vector Search: selects mode backend/app/retrievers.py
BM25 Search → Chroma Chunk Store: reads chunks backend/app/retrievers.py
Vector Search → Chroma Chunk Store: searches vectors README.md
Chroma Chunk Store → OpenAI Embeddings: embeds queries backend/app/retrievers.py
RAG Answering → OpenAI Chat Model: generates answer backend/app/rag.py
Document Ingestion → Source Documents: loads files backend/app/ingest.py
Document Ingestion → Chroma Chunk Store: writes chunks backend/app/ingest.py
Document Ingestion → Retriever Selection: resets caches backend/app/ingest.py
Runtime Settings → Document Ingestion: configures backend/app/ingest.py
Runtime Settings → Retriever Selection: configures backend/app/retrievers.py
Runtime Settings → RAG Answering: configures backend/app/rag.py
RAGAS Runner → Evaluation Testset: reads cases
RAGAS Runner → RAG Answering: runs questions backend/eval/run_ragas.py
RAGAS Runner → Retriever Selection: checks index backend/eval/run_ragas.py
RAGAS Runner → Evaluation Results: writes report backend/eval/run_ragas.py
Testset Drafter → Retriever Selection: loads chunks backend/eval/draft_testset.py
Results Comparison → Evaluation Results: compares reports

Drawn by GitDiagram on November 11, 2026 from the repository's file tree, README and source files. Read as Markdown ·