tony
ed64712525
Grok now drives the trip rendering in real time instead of dumping
the full result after ~90 seconds.
Backend
- GrokHeadlessClient gains a chatStream() async generator that spawns
grok with --output-format streaming-json (NDJSON of {type,data}
events), buffers the "text" tokens, and emits partial events as the
buffer becomes parseable.
- tryPartialJsonParse — lenient JSON repair: walks the buffer once,
closes structures in stack order, drops in-progress strings and
dangling keys, returns whatever object is currently consistent.
Hard-tested with progressive slicing of a multi-stop itinerary.
- New SSE endpoint POST /api/chat/stream with events: open / thinking
/ partial / done / error. Uses res.on('close') + writableEnded as a
reliable client-disconnect signal (req.on('close') fires in Express
5 once the body is consumed, which was killing the grok child).
Frontend
- sendMessage swaps to fetch+ReadableStream against /api/chat/stream
and parses SSE blocks. Each partial event runs a fast synchronous
normalizePartialItinerary (no Nominatim — drops stops missing
lat/lng so partial render doesn't block on geocoding).
- The done event runs the full async normalizer for the final pass
and caches the result per variant.
- Stops, day cards, map markers, polylines, the variant strip, and
the trip summary all update progressively as Grok writes each stop.
Verified with a London → Edinburgh prompt: 6 partial events landed
across the 76-second stream, with the rail filling in
"Baldock Services" → "+Grantham A1" → "+Premier Inn Newcastle"
→ "+Fort Kinnaird" before the final done event.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Tesla Roadtrip — Grok-Powered Planner (UK & Europe First)
Beautiful split-pane Tesla trip planner.
Left side: Natural language chat with Grok Drive (powered by the real xAI Grok API)
Right side: Interactive Leaflet map + live itinerary that updates as you talk
First Release Focus
- United Kingdom
- Europe (France, Germany, Benelux, Switzerland, Spain)
Key Features
- Real Grok intelligence via the xAI API
- Excellent Supercharger coverage awareness across the UK & Europe
- Metric units (km, Wh/km)
- Proactive route planning — just say where you want to go
- Safe, restricted tool use
Tech Stack
- Vite + React + Tailwind + Leaflet (client)
- Express + Prisma + pino (server)
- Real xAI Grok API (with local
grokCLI as optional path) - Heavy structured logging for fast debugging
Local Development
1. Install dependencies
npm install
cd client && npm install && cd ..
2. Set up environment variables
cp .env.example .env
Then edit .env and add your xAI API key:
XAI_API_KEY=xai-YourKeyHere
You can get a key from: https://console.x.ai
3. Start the development environment
You have two options:
Option A (Recommended) – Better developer experience:
./scripts/dev.sh
Option B – Using npm:
npm run dev
Both commands will start:
- Backend: http://localhost:3000 (with detailed logging)
- Frontend: http://localhost:5173
4. Open the app
Visit: http://localhost:5173
Useful Commands
| Command | Description |
|---|---|
npm run dev |
Start both frontend and backend |
./scripts/dev.sh |
Recommended way to start (better output) |
npm run dev:server |
Start only the backend |
npm run dev:client |
Start only the frontend |
Security & Philosophy
This project follows a strict security policy:
- No packages with known active critical or high vulnerabilities
react-markdown/ remark ecosystem avoided (LLM output sanitization via DOMPurify)- All AI tool use is heavily restricted
- Maximum logging so we can debug the "chat → route → stops" flow easily
Deployment Target
Will be deployed via Dokku on Hetzner with Gitea CI.
First launch target: United Kingdom & Europe.
Need Help?
- Make sure your
XAI_API_KEYis set (otherwise you'll get very basic responses) - Check the backend terminal — it has extremely detailed logs
- The app is designed so you can iterate quickly by watching the logs
Development & Iteration Workflow (Autonomous Loop)
This project is designed for fast, autonomous iteration using Playwright.
One-Command Iteration Loop
The recommended way to test and iterate is:
./scripts/iterate.sh
This script will:
- Ensure both backend and frontend are running (via
./scripts/dev.sh) - Run the fast backend diagnostic test (
smoke test) - Run the full E2E Playwright test in headed mode
- Open the Playwright HTML report
- Show the latest screenshots and video
- Print the most relevant backend log lines from the test window
After the script finishes, review the artifacts and tell me what to fix. Then just run ./scripts/iterate.sh again.
Fast Smoke Test (Backend Only)
When you only want to quickly test if Grok is responding (without waiting for the full UI flow):
./scripts/smoke.sh
This runs in ~30–90 seconds and is perfect for prompt tuning or backend debugging.
Manual Commands
| Command | Description |
|---|---|
./scripts/dev.sh |
Start both servers (recommended) |
npm run dev |
Start both servers (via npm) |
./scripts/iterate.sh |
Full autonomous test + report loop |
./scripts/smoke.sh |
Fast backend-only Grok test |
npx playwright test |
Run all Playwright tests manually |
Test Reports & Artifacts
- Playwright HTML report:
npx playwright show-report - Screenshots & videos:
test-results/ - Backend logs: Look in your terminal or
/tmp/tesla-roadtrip-backend.log(if you enabled logging to file)
Workflow Summary
- Make a code or prompt change
- Run
./scripts/iterate.sh - Review the report + screenshots + backend logs
- Tell me what to fix
- Repeat
This loop lets me drive most of the testing and debugging with minimal manual work from you.