/LLM-Demo-modeling

FastAPI API Documentation

This document provides instructions on setting up and using the FastAPI-based API endpoints for processing CSV files and performing question answering.

Data Engineering

Raw Data

The raw data is sourced from the text_segments.csv file, containing information about document segments including pagenum, docname, and text.

Data Engineering Process

  1. Text Aggregation: Combine text segments from the same page of the same document to enhance document coherency.
  2. JSON Serialization: Serialize aggregated text segments into JSON files, one for each unique document, to preserve metadata.

Challenges Faced

  1. Consistency in Document Processing: Ensuring uniformity in document structure to facilitate effective embedding for question-answering.
  2. Document File Format Selection: Choosing between file formats; JSON was selected for metadata retention.

Proposed Improvements

  1. Custom Prompts for QA: Enhance question-answering with custom prompts for more polite and nuanced responses.

  2. Guardrails Implementation: Strengthen input validation and error handling mechanisms.

  3. Chat History Integration: Enable continuous conversations by adding support for chat history.

  4. User-Specific Document Support: Allow users to upload and analyze their own documents.

Project Structure

  • app/: Contains the FastAPI application code.
  • data/: Holds processed data, including serialized JSON files.

Prerequisites

  1. Python Environment:

    • Ensure you have Python installed on your system. You can download Python from python.org.

  2. Virtual Environment:

    • Create a virtual environment to isolate dependencies. 
      python -m venv venv

  3. Activate Virtual Environment:

    • Activate the virtual environment.
      • On Windows: 
        .\venv\Scripts\activate
      • On Unix or MacOS: 
        source venv/bin/activate

  4. Install Requirements:

    • Install the required Python packages. 
      pip install -r requirements.txt

  5. Create .env file:

    • Create a .env file in the project root based on the provided env.example file. This file should contain any necessary environment variables.

Running the API Server

  1. Run the FastAPI Server:

    • Execute the following command to start the FastAPI server. 
      uvicorn app.main:app --reload
    • The --reload flag enables automatic reloading of the server when code changes are detected.

  2. Access API Documentation:

    • Open your browser and go to http://127.0.0.1:8000/docs.
    • This will open the FastAPI Swagger UI, providing an interactive interface for exploring and testing the API endpoints.

Running the API Server with Docker Compose

  1. Build and Run Docker Containers:

    • Execute the following command to build the Docker image and start the FastAPI server along with any required services defined in the docker-compose.yml file. 
      docker-compose up --build -d
    • The --build flag ensures that the Docker images are rebuilt, and the -d flag runs the services in the background.

  2. Access API Documentation:

    • Open your browser and go to http://127.0.0.1:8000/docs.
    • This will open the FastAPI Swagger UI, providing an interactive interface for exploring and testing the API endpoints.

  3. Stopping Docker Compose:

    • To stop the running containers, use the following command: 
      docker-compose down
    • This will stop and remove the containers defined in the docker-compose.yml file.

API Endpoints

1. Process CSV File

  • Endpoint: /process_csv
  • Method: POST
  • Request Payload:
    • Upload a CSV file using the provided form.
  • Response:
    • Successful response: {"message": "File processed and stored"}
    • In case of an error, an appropriate error message will be returned.

2. Question Answering

  • Endpoint: /qa
  • Method: POST
  • Request Payload:
    • Provide a JSON object with the question using the key query. 
      { "query": "Your question goes here" }
  • Response:
    • Successful response: {"response": "Answer to the question"}
    • In case of an error, an appropriate error message will be returned.

Frontend Setup