Nvidia Nim x AutoRAG¶

Setting Up the Environment¶

Installation¶

First, you need to have AutoRAG.

Install AutoRAG:

pip install autorag

And go to the NVIDIA NIM website, register, and select what models that you want use.

nvidia_nim

After select the right model, click “Build with this NIM” Button. And copy your api key!

nvidia_api

Using NVIDIA NIM with AutoRAG¶

For using NVIDIA NIM, you can use Llama Index LLm’s openailike at the AutoRAG config YAML file without any further configuration.

It is EASY!

Writing the Config YAML File¶

Here’s the modified YAML configuration using NVIDIA NIM:

nodes:
  - node_line_name: node_line_1
    nodes:
      - node_type: generator
        modules:
          - module_type: llama_index_llm
            llm: openailike
            model: nvidia/llama-3.1-nemotron-70b-instruct
            api_base: https://integrate.api.nvidia.com/v1
            api_key: your_api_key

For full YAML files, please see the sample_config folder in the AutoRAG repo at here.

Running AutoRAG¶

Before running AutoRAG, make sure you have your QA dataset and corpus dataset ready. If you want to know how to make it, visit here.

Run AutoRAG with the following command:

autorag evaluate \
 - qa_data_path ./path/to/qa.parquet \
 - corpus_data_path ./path/to/corpus.parquet \
 - project_dir ./path/to/project_dir \
 - config ./path/to/nim_config.yaml

AutoRAG will automatically experiment and optimize RAG.