AI Subaru Tuning Platform

This project provides a FastAPI backend and React frontend for managing Subaru tuning files and datalogs. It archives data, analyzes logs and tunes, and guides users through safe engine tuning with help from built-in AI agents.

Installation

Backend

Use Python 3.11+ and install dependencies:

pip install -r backend/requirements.txt

Frontend

Install Node.js packages inside the frontend folder:

cd frontend
npm install

Running the applications

FastAPI server

Before starting the server, install the backend requirements:

pip install -r backend/requirements.txt

Then run the API with Uvicorn:

uvicorn backend.main:app --reload

This starts the API at http://localhost:8000. Set the JWT_SECRET environment variable to the shared secret used for verifying JWT bearer tokens.

JWT bearer tokens. During development you can bypass authentication entirely by setting DISABLE_JWT_AUTH=1 when launching the server.

Example (Linux/macOS shell):

export DISABLE_JWT_AUTH=1
uvicorn backend.main:app --reload

Example (Windows PowerShell):

$Env:DISABLE_JWT_AUTH=1
uvicorn backend.main:app --reload

React app

From the frontend folder execute:

npm start

The development server will open at http://localhost:3000 and proxy API requests to the FastAPI server.

Testing

Backend tests

Install the backend dependencies and run unit tests with:

pip install -r backend/requirements.txt
pytest

Frontend tests

In the frontend folder run:

npm test

Downloading Exports

After exporting tune changes you can download the file using:

GET /api/download/{session_id}/{filename}

This returns the exported JSON or CSV file for offline review.

AI agents

The application coordinates several specialized agents:

• tuning-ai – provides expert tuning recommendations using datalog and tune analysis while warning of unsafe changes.
• carberry-parser – interprets Carberry/ROMRaider definition and tune files, exposing structured data.
• datalog-analyzer – indexes and visualizes datalogs, comparing runs and flagging anomalies.
• ui-ux-pro – ensures the interface is clear and accessible for beginners and experts.
• db-archivist – manages storage, versioning and integrity of tune files and datalogs.
• supervisor – oversees all agents, delegating tasks and enforcing standards.

These agents collaborate through documented APIs and follow strict safety and accessibility requirements.

The UI now includes a Tune Info Panel that exposes ROM metadata such as file name, checksum, parser version and table counts for each session. Users can now browse the unmodified ROM tables right after uploading a tune. Tables expand into 2D or 3D grids with axis headers so you can inspect raw values before any changes are applied. A dedicated Debug view exposes full session details and API responses for troubleshooting.

Data Verification

See docs/data_validation.md for a checklist of metadata and change details exposed by the platform. For a sample explanation of a detected AFR issue and how a tune change is presented to the user, read docs/tune_change_example.md.