AcademicRag

Overview

AcademicRag is a Dockerized Retrieval-Augmented Generation (RAG) system.
It combines:

• A custom LLM served by Ollama (built from a Modelfile)
• Chroma for document embeddings
• MongoDB for logging queries
• A Flask REST API for interaction

Features

• Query the LLM with context from your documents
• Automatic citation extraction ([Document: ..., Section: ...])
• Persistent query logs with timestamp filters
• Fully containerized with Docker Compose

Directory Structure

slava project/
├─ app/                  # Flask application
│  ├─ api/               # REST endpoints (e.g. /logs)
│  └─ rag/               # LLM output parser
├─ modelfile/
│  └─ Modelfile          # Ollama model definition
├─ Ollama.Dockerfile     # Builds custom Ollama model image
├─ docker-compose.yml    # Multi-container setup
├─ Dockerfile            # Flask service build
├─ requirements.txt      # Python deps
└─ .gitignore            # Ignored files/folders

Prerequisites

• Docker & Docker Compose
• Docker Hub account (optional, for pushing images)

Getting Started

1. Clone the repo

   git clone https://github.com/<your-username>/AcademicRag.git
   cd "slava project"

2. Configure environment
   Copy .env.example to .env and set:

   MONGO_DB_NAME=academic_rag
   LOGS_COLLECTION=query_logs
   

3. Build & run containers
   This command also builds the Ollama image using Ollama.Dockerfile to bake in your custom model:

   docker-compose up --build -d

4. Access services

   • Flask API: http://localhost:5000
   • Swagger UI (interactive API docs): http://localhost:5000/apidocs
   • Mongo-Express: http://localhost:8081
   • Ollama LLM: use ollama CLI on host or container port 11434
   • Chroma HTTP API: http://localhost:8000

API Usage

• Interactive docs
  Browse and test endpoints in Swagger UI:

• POST /ask

  { "query": "Your question here" }

  Response:

  {
    "answer": "Generated answer…",
    "citations": ["[Document: file.pdf, Section: …]"]
  }

• GET /logs?start_time=2025-07-01T00:00:00&end_time=2025-07-15T23:59:59
  Returns filtered query logs.

Managing Data

• To clear MongoDB data, stop services and delete the volume folder:

  docker-compose down
  Remove-Item -Recurse -Force mongo_data
  docker-compose up -d

• Similarly remove chroma_data/ or ollama_data/ to reset embeddings or models.

Customizing the Model

• Edit modelfile/Modelfile as needed:

  FROM llama2
  PARAMETER temperature 0.25
  

• Rebuild and load your custom model image using Docker Compose:

  docker-compose up -d --build ollama

  This uses Ollama.Dockerfile to bake your modelfile/Modelfile into the guyyagil/ollama-custom:latest image.

Pushing Images to Docker Hub

1. Build all images:

   docker-compose build

2. Push to Docker Hub:

   docker login
   docker push guyyagil/flask_app:latest
   docker push guyyagil/ollama-custom:latest

Contributing

1. Fork the repo
2. Create a feature branch
3. Submit a pull request

License

This project is licensed under the MIT