Ask Your Docs

A RAG web app to chat with your PDFs. Upload a document, ask questions, and get answers grounded only in that document, streamed token by token.

Built with FastAPI, LangChain, ChromaDB, Jina embeddings and Groq.

Features

• PDF upload with per-session isolation: each browser session gets its own Chroma collection
• Retrieval-augmented answers that refuse to guess ("I don't know" when the context has no answer)
• Streaming responses over HTTP
• Minimal HTML/CSS/JS frontend served by the same FastAPI app

How it works

Upload   PDF -> load -> split into chunks -> embed (Jina) -> store in Chroma (collection per session)

Query    question -> embed -> top-k similar chunks -> prompt + context -> LLM (Groq) -> streamed answer

Tech stack

Layer	Choice
Backend	FastAPI
Orchestration	LangChain (LCEL)
LLM	Groq
Embeddings	Jina embeddings (API)
Vector store	ChromaDB (persistent, local)
Frontend	HTML, CSS, JavaScript

Project structure

admin/
  main.py              FastAPI app and routes
scripts/               ingestion pipeline
  content_fetcher.py   PDF loading
  chunking.py          text splitting
  vectorize.py         Jina embeddings wrapper
  chroma_db.py         vector store
  ingestion.py         load -> split -> embed -> store
services/
  query_service.py     retrieval and answer generation
static/
  index.html           chat UI

Getting started

Prerequisites: Python 3.10+, a Groq API key, a Jina API key.

git clone https://github.com/saimdev7/ask-your-docs.git
cd ask-your-docs

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

cp env.example .env      # then add your API keys
uvicorn admin.main:app --reload

Open http://127.0.0.1:8000, upload a PDF, and start asking.

Configuration

Set in .env:

Variable	Description
JINA_API_KEY	Jina API key
JINA_MODEL	Embedding model, e.g. jina-embeddings-v5-text-small
GROQ_API_KEY	Groq API key
GROQ_MODEL	Chat model, e.g. openai/gpt-oss-20b

Model availability on free tiers changes. If you get a model_not_found error, pick another model from your Groq console.

API

Method	Endpoint	Description
GET	/	Chat UI
GET	/health	Health check
POST	/upload	Form data: session_id, file (PDF). Ingests the document
POST	/chat	JSON: {"question", "session_id"}. Streams the answer as plain text

Interactive docs are available at /docs.

Limitations

• No authentication; sessions are identified by a client-generated ID
• Uploading a second PDF in the same session adds to the existing collection instead of replacing it
• Storage is local (chroma_db/, uploads/), so a deployment needs a persistent disk