PdfSearchAI

PdfSearchAI is a local RAG (Retrieval-Augmented Generation) service that indexes a PDF into Qdrant and answers user queries using Ollama.

Requirements

• Python 3.10+
• Docker (for Qdrant + Valkey/Redis)
• Ollama running locally

Project Files

• indexing-pdf.py: Loads Resume.pdf, splits text into chunks, builds embeddings, and stores vectors in Qdrant collection learning_rag.
• main.py: Starts the FastAPI application with Uvicorn on 0.0.0.0:8000.
• server.py: Defines API routes to enqueue chat jobs and check job status/results.
• queues/worker.py: RQ worker logic (process_query) that performs vector search and calls Ollama chat.
• clients/rq_client.py: Creates the Redis-backed RQ queue client.
• docker-compose.yml: Starts Qdrant (6333) and Valkey/Redis (6379).

Setup

1. Create and activate a virtual environment.

python3 -m venv .venv
source .venv/bin/activate

2. Install dependencies.

pip install -r requirements.txt

3. Start infrastructure services.

docker compose up -d

4. Pull required Ollama models.

ollama pull nomic-embed-text
ollama pull llama3.2:1b

Index Your PDF

Place your PDF at ./Resume.pdf (or update the path in indexing-pdf.py) and run:

python3 indexing-pdf.py

Run the Application

Start the API server:

python3 main.py

Start the RQ worker (in a separate terminal):

export OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES && rq worker --with-scheduler

API Endpoints

• GET / returns health message.
• POST /chat?query=... enqueues a background job and returns a job_id.
• GET /job?job_id=... returns processing status or final result.

How It Works

PDF -> chunking -> nomic-embed-text embeddings -> Qdrant (learning_rag)
User query -> queued job (RQ) -> similarity search -> context -> llama3.2:1b -> response

Learn RAG

• Documentation