Vehicle Diagnostic Assistant with Semantic Search for OBD-II Codes

How-it-works

DEMO

Problem-Statement-1:

I am experiencing issues with my clutch switch on a 2012 Honda Civic. The car doesn’t start unless I press the clutch, even though it used to start without pressing the clutch pedal (OBD Codes: P0704)

Solution:

This project is a vehicle diagnostic assistant powered by semantic search techniques to process OBD-II (On-Board Diagnostics) codes. The assistant helps users troubleshoot automotive issues by analyzing their vehicle problems, matching them with relevant OBD-II codes, and providing detailed diagnostics, including possible causes and solutions. The system also integrates image recognition to analyze vehicle-related images for additional diagnostic insights.

Features

• Cosine Semantic Search: Uses cosine similarity to compare user inputs (text and image descriptions) with OBD-II code embeddings, providing the most relevant diagnostic information.
• Image Processing: Allows users to upload images related to vehicle issues, which are processed and analyzed to provide context-specific diagnostic recommendations.
• Natural Language Processing (NLP): Leverages OpenAI’s GPT model to generate human-readable diagnostic reports, including explanations, causes, symptoms, and recommended solutions.
• OBD-II Code Database: A database stores nearly 1000 OBD-II codes, along with their meanings, causes, symptoms, and embeddings, to facilitate fast and accurate semantic search.
• Azure AI Integration: Uses Azure’s AI services to generate text embeddings and process vehicle-related images for enhanced diagnostics.

Technologies Used

• Backend: Node.js, Express.js
• Database: MongoDB
• AI & NLP: OpenAI API, Azure AI Inference
• Image Processing: Multer (for handling file uploads)
• Cosine Similarity: Used for embedding comparison and semantic search
• Environment Variables: dotenv for managing environment variables

Installation

Follow these steps to set up the project locally:

1. Clone the Repository

git clone https://github.com/your-username/vehicle-diagnostic-assistant.git
cd vehicle-diagnostic-assistant

2. Install Dependencies Make sure you have Node.js installed. Then, install the required dependencies:

npm install
3. Set Up Environment Variables
Create a .env file in the root directory and define the following environment variables:

env

GITHUB_TOKEN=your_github_token
MONGODB_URI=your_mongodb_connection_string
AZURE_API_KEY=your_azure_api_key
PORT=3000
GITHUB_TOKEN: Your GitHub token for API access.
MONGODB_URI: Your MongoDB connection URI.
AZURE_API_KEY: Your Azure API key for embedding and AI inference.

PORT: The port the server will run on (default is 3000). 4. Run the Application Start the server:

npm start

The server will run at http://localhost:3000.

Endpoints

1. GET / Returns a status message indicating the server is running.

2. GET /run-batch Triggers a batch process to embed and insert OBD-II codes into the database.

3. POST /query-with-image Accepts an image file and user input to analyze the image and provide diagnostic recommendations based on both image content and text.