sonuramashishnpm/NPM-Rag-API-Framework
A powerful FastAPI-based multi-modal ingestion system that processes PDFs, scanned documents, images, videos, YouTube links, and text files — then optionally performs semantic retrieval using FAISS + HuggingFace embeddings and refines answers using an LLM (Ollama via NPMAI).
What's novel
A powerful FastAPI-based multi-modal ingestion system that processes PDFs, scanned documents, images, videos, YouTube links, and text files — then optionally performs semantic retrieval using FAISS + HuggingFace embeddings and refines answers using an LLM (Ollama via NPMAI).
Code Analysis
0 files read · 4 roundsA monolithic FastAPI script that attempts to unify PDF OCR, video transcription, and LLM-based RAG retrieval into a single endpoint without proper separation of concerns.
Strengths
The README provides a clear and accurate architectural overview of the intended workflow (Ingestion -> Extraction -> Retrieval -> Refinement). The project aims to solve a real problem by combining multiple modalities into one API.
Weaknesses
The code is severely limited in substance due to inaccessible source files, but based on the monolithic structure implied, it likely suffers from poor modularity, lack of error handling for diverse extraction failures, and potential security risks with FAISS deserialization. No tests are present.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
17
Contributors
1
Files
6
Active weeks
4
Repository
Language
Python
Stars
1
Forks
0
License
MIT