Accorda Legal Assistant Backend & Frontend

This project contains the backend API and the frontend web application for the Accorda AI legal assistant.

Tech Stack

• Containerization: Docker, Docker Compose
• Backend: Python, FastAPI
• Frontend: Svelte, Vite, Tailwind CSS
• AI: Google Gemini API

Running the Project

There are two ways to access the application:

1. Hosted Version (Recommended):

   • Access the live application at: https://accorda.kolvian.com
   • Please contact the development team for login credentials. (ivlub on Discord/ivan.lubyako@gmail.com)

2. Local Development Setup:

   • Prerequisites:
     • Docker
     • Docker Compose (v2)
   • Environment Setup:
     • Create a .env file in the project root directory (where docker-compose.yml is located).
     • Add the following line to the .env file, replacing XXXX with your actual key:

       GOOGLE_API_KEY=XXXX
       

     • You can obtain a Google API key from the Google AI Studio.
   • Build and Run:
     • Open your terminal in the project's root directory.
     • Run the command:

       docker compose up --build
       

     • The --build flag is crucial for the first run or after dependency changes. Subsequent runs can use docker compose up.
   • Accessing Local Services:
     • Frontend: http://localhost:5173
     • Backend: http://localhost:8000
   • API Documentation:
     • Swagger UI documentation for the backend API is available at: http://localhost:8000/docs

API Endpoint Overview

The backend exposes the following primary API endpoints under the /api/ prefix:

• GET /hello: Basic health check or root endpoint.
• POST /generate: General purpose AI text generation.
• POST /extract_text: Extracts text content from uploaded documents (e.g., PDFs).
• POST /compare: Compares two contracts and identifies differences.
• POST /analyze/categorize: Categorizes the type of contract.
• POST /analyze/summary: Generates a summary of a contract.
• POST /analyze/check-requirements: Checks a contract against general requirements.
• POST /analyze/check-category-requirements: Checks a contract against requirements specific to its category.
• POST /explain-diff: Explains the differences found between two contracts.
• POST /analyze-diff-impact: Analyzes the potential impact of the differences between two contracts.

Note: For detailed request/response schemas and parameters, please refer to the Swagger documentation at /docs.

Data Handling & Privacy (GDPR Compliance)

This application prioritizes data privacy. Uploaded documents are not stored permanently on our servers. They are processed solely in temporary memory or temporary files during the analysis or comparison task. Once the processing is complete and the results are returned to the user, these temporary files are immediately and securely deleted. This ensures that your sensitive contract data is not retained, aligning with GDPR principles.

AI Contract Analysis Flow

The core contract analysis process, typically triggered via the frontend, involves a sequence of calls to the backend API:

1. Summary Generation (/api/analyze/summary): The AI first generates a concise summary of the entire contract.
2. General Requirements Check (/api/analyze/check-requirements): The AI then checks the contract against a predefined set of general legal and structural requirements applicable to most contracts.
3. Categorization & Category-Specific Check (/api/analyze/categorize & /api/analyze/check-category-requirements):
   • The AI attempts to categorize the contract (e.g., NDA, Service Agreement).
   • Based on the identified category, the backend retrieves the corresponding compliance matrix from backend/prompts/category_requirements_schemas.json.
   • The AI analyzes the contract against this specific matrix, extracting clauses or noting missing items for each criterion.

Schema File (category_requirements_schemas.json): This file defines the structured criteria (compliance matrix) for each supported contract category. It includes the specific points the AI should look for during the category-specific check.

Current Status: As a proof of concept, only the "Non-Disclosure Agreement" category is currently defined and implemented within the category_requirements_schemas.json file. Integration of other categories is very straightforward - just add those to the JSON file following the existing structure.

Development Notes

• Backend: Changes to Python files in the /backend directory trigger an automatic reload of the FastAPI server using uvicorn --reload.
• Frontend: Changes to Svelte/TS/CSS files in the /frontend/src directory trigger Vite's Hot Module Replacement (HMR) for instant updates in the browser.

Stopping the Local Application

1. Press Ctrl+C in the terminal where docker compose up is running.
2. To remove the containers and associated networks/volumes, run:

   docker compose down
   

Frontend Structure & Components

The frontend application, located in /frontend, is a Single Page Application (SPA) built with Svelte and TypeScript, using Vite for development and builds, and Tailwind CSS for styling.

Key Directories & Files:

• frontend/src/: Core application code.
  • main.ts: Application entry point.
  • App.svelte: Root component orchestrating the UI.
  • app.css: Global styles.
  • components/: Contains reusable UI components:
    • Login.svelte: Handles user authentication.
    • Sidebar.svelte: Navigation panel, likely for selecting analysis types or managing documents.
    • TopBar.svelte: Application header, possibly including user info or global actions.
    • Welcome.svelte: Initial view or dashboard.
    • ContractAnalysis.svelte, ContractComparison.svelte, ContractSummary.svelte: Components dedicated to specific contract interaction features.
    • SettingsView.svelte: Interface for user/application settings.
    • Modal.svelte: Reusable modal dialog component.
    • LoadingSpinner.svelte: Visual indicator for ongoing operations.
    • HelpTutorial.svelte: Component providing guidance or tutorials.
  • stores/: Svelte stores for state management.
    • settingsStore.ts: Manages application-wide settings.
• frontend/public/: Static assets.
• frontend/index.html: Main HTML entry point.
• Configuration Files: vite.config.ts, svelte.config.cjs, tailwind.config.cjs, tsconfig.json, etc.
• frontend/Dockerfile: Defines the build process for the frontend container.

Exporting Results

After performing a contract analysis using the ContractAnalysis component in the frontend, the generated results (including the summary, general requirements check, and category-specific analysis) can be exported to a professionally formatted PDF document. This functionality is handled within the ContractAnalysis.svelte component itself, using the jsPDF library. Same applies for the ContractComparison component too.

Future Enhancements

• Expand the number of supported contract categories in category_requirements_schemas.json.
• Implement more sophisticated analysis options (e.g., risk assessment).
• Refine the UI/UX based on user feedback.
• Integrate OCR for scanned documents
• Implement database for request history