https://github.com/anasaber/mlflow_with_rag

Using MLflow to track a RAG pipeline, using LLamaIndex and Ollama/HuggingfaceLLMs
https://github.com/anasaber/mlflow_with_rag

cicd deployment evaluation-metrics llamaindex llamaindex-rag mlflow mlflow-deployement mlflow-projects mlflow-tracking mlflow-tracking-server mlflow-ui mlops mlops-project mlops-template rag rag-evaluation rag-pipeline

Last synced: 7 days ago
JSON representation

Using MLflow to track a RAG pipeline, using LLamaIndex and Ollama/HuggingfaceLLMs

• Host: GitHub
• URL: https://github.com/anasaber/mlflow_with_rag
• Owner: AnasAber
• Created: 2024-12-09T11:21:46.000Z (17 days ago)
• Default Branch: master
• Last Pushed: 2024-12-18T15:06:33.000Z (8 days ago)
• Last Synced: 2024-12-18T16:22:42.726Z (8 days ago)
• Topics: cicd, deployment, evaluation-metrics, llamaindex, llamaindex-rag, mlflow, mlflow-deployement, mlflow-projects, mlflow-tracking, mlflow-tracking-server, mlflow-ui, mlops, mlops-project, mlops-template, rag, rag-evaluation, rag-pipeline
• Language: Python
• Homepage:
• Size: 60.3 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
### MLflow Deployement of a RAG pipeline 🥀

This project is for people that want to deploy a RAG pipeline using MLflow.

The project uses:

- `LlamaIndex` and `langchain` as orchestrators

- `Ollama` and `HuggingfaceLLMs`

- `MLflow` as an MLOps framework for deploying and tracking

![Project Overview Diagram](images/mlflow_rag_schema.png)

### How to start

1. Clone the repository

```bash

git clone https://github.com/AnasAber/RAG_in_CPU.git

```

2. Install the dependencies

```bash

pip install -r requirements.txt

```

Make sure to put your api_keys into the `example.env`, and rename it to `.env`

3. Notebook Prep:

- Put your own data files in the data/ folder

- Go to the notebook, and replace "api_key_here" with your huggingface_api_key

- If you have GPU, you're fine, if not, run it on google colab, and make sure to download the json file output at the end of the run.

4. Go to `deployement` folder, and open two terminals:

```bash

python workflow.py

```

And after the run, go to your mlflow run, and pick the run ID:

![Run ID](images/run_id.png)

Place it into this command:

```bash

mlflow models serve -m runs://rag_deployement -p 5001

```

In the other terminal, make sure to run

```bash

app.py

```

5. Open another terminal, and move to the `frontend` folder, and run:

```bash

npm start

```

Now, you should be seeing a web interface, and the two terminals are running.

![Interface](images/interface.png)

If you got errors, try to see what's missing in the requirements.txt.

Enjoy!