How dataforgoodfr/13_ia_financement works
This repository contains a Streamlit assistant that helps associations pre-fill funding applications from project and organization documents. The main workflow is documented as uploading source material and a blank application, retrieving relevant context with vector or graph RAG, and viewing or downloading a completed form. The README also describes V2 question extraction, table/diagram extraction, and RAG-selection work, but says these were not integrated. The V1 wiring is supported by the README and sampled imports; unsampled V2 implementation links are left unverified.
Main entry points: rag_pipelines.py Document ingestion · app.py Streamlit interface · PathRAG.py PathRAG engine · read_answer_aap.py Form extraction and filling
Components 18
Application
- Streamlit interface
streamlit_users/app.py - Introduction page
streamlit_users/page1_intro.py - Document ingestion
streamlit_users/rag_pipelines.py - Question answering
streamlit_users/rag_pipelines.py - Form extraction and filling
streamlit_users/read_answer_aap.py
Retrieval
- Hybrid retrieval
streamlit_users/hybrid_retriever.py - Graph retrieval
streamlit_users/graphrag_retriever.py
Graph RAG
- PathRAG engine
streamlit_users/PathRAG/PathRAG.py - Graph operations
streamlit_users/PathRAG/operate.py - Graph and vector storage
streamlit_users/PathRAG/storage.py - LLM adapters
streamlit_users/PathRAG/llm.py
Planned V2
- Visual question extraction
notebooks - Table and diagram extraction
notebooks - Question classifiers
notebooks/questions_classifier.ipynb
Other components
- Association user
- Source documents
- Blank application
- Completed application
Connections 22
- Association user → Streamlit interface: uses
README.md - Association user → Source documents: uploads
README.md - Association user → Blank application: uploads
README.md - Streamlit interface → Introduction page: routes to
streamlit_users/app.py - Streamlit interface → Document ingestion: processes documents
streamlit_users/app.py - Streamlit interface → Question answering: requests answers
streamlit_users/app.py - Streamlit interface → Form extraction and filling: uses
streamlit_users/app.py - Source documents → Document ingestion: provides content
README.md - Blank application → Form extraction and filling: provides template
README.md - Document ingestion → Hybrid retrieval: builds retrieval
streamlit_users/rag_pipelines.py - Document ingestion → Graph retrieval: builds retrieval
streamlit_users/rag_pipelines.py - Question answering → Hybrid retrieval: queries
streamlit_users/rag_pipelines.py - Question answering → Graph retrieval: queries
streamlit_users/rag_pipelines.py - Graph retrieval → PathRAG engine: creates and queries
streamlit_users/graphrag_retriever.py - PathRAG engine → Graph operations: delegates to
streamlit_users/PathRAG/PathRAG.py - PathRAG engine → Graph and vector storage: uses
streamlit_users/PathRAG/PathRAG.py - PathRAG engine → LLM adapters: uses
streamlit_users/PathRAG/PathRAG.py - Form extraction and filling → Completed application: produces
README.md - Association user → Completed application: downloads
README.md - Visual question extraction → Document ingestion: planned integration
notebooks/streamlit_MM/app.py - Table and diagram extraction → Document ingestion: planned integration
notebooks/streamlit_MM/app.py - Question classifiers → Question answering: planned routing