A NestJS application that provides an API for analyzing document images using AI with LangChain and the Gemma-3-12b-it model via LMStudio.

This project provides an API endpoint for analyzing document images. It uses LangChain to integrate with the Gemma-3-12b-it model running on LMStudio to perform AI-powered image analysis. The API returns a detailed quality assessment of the uploaded document image, including scores and explanations for image sharpness, blurriness, text visibility, and overall text quality.

• Node.js (v16 or higher)
• npm or yarn
• LMStudio installed locally with Gemma-3-12b-it model

$ npm install

1. Download and install LMStudio
2. Download the Gemma-3-12b-it model in LMStudio
3. Start the local server in LMStudio with the Gemma-3-12b-it model
4. Ensure the server is running on http://localhost:1234

# development
$ npm run start

# watch mode
$ npm run start:dev

# production mode
$ npm run start:prod

The API will be available at http://localhost:3000.

Analyzes a document image and returns quality assessment.

• URL: /document-identifier/analyze

• Method: POST

• Content-Type: application/json

• Request Body:

  {
    "image": "base64_encoded_image_data",
    "contentType": "image/jpeg"
  }

• Response:

  {
    "confident": "0.95",
    "conclusion": "The image appears to be a scanned ID card and the text is generally readable...",
    "reason": {
      "image_sharpness": {
        "score": "0.85",
        "reason": "The image has decent sharpness, though some areas show slight blurring."
      },
      "blurred": {
        "score": "0.75",
        "reason": "There's a noticeable blur around the edges and signature, which could impact OCR accuracy in those regions."
      },
      "obscured": {
        "score": "1",
        "reason": "The text is not obscured by any major objects or reflections."
      },
      "text_quality": {
        "score": "0.9",
        "reason": "The text quality is good overall, but some minor artifacts and shadowing are present due to the scanning process."
      }
    }
  }

The application is configured to connect to LMStudio running locally on port 1234. If your LMStudio instance is running on a different port or host, update the configuration in src/document-identifier/langchain.service.ts:

this.model = new ChatOpenAI({
  modelName: 'gemma-3-12b-it',
  temperature: 0.2,
  maxTokens: 1000,
  openAIApiKey: 'dummy-key',
  configuration: {
    baseURL: 'http://localhost:1234/v1', // Update this URL if needed
  },
});

# unit tests
$ npm run test

# e2e tests
$ npm run test:e2e

# test coverage
$ npm run test:cov

This project is MIT licensed.