models.dev-rss

Minimal server that converts https://models.dev/api.json into RSS on demand. Built with Hono so the same app runs on Node and Cloudflare Workers.

Node runtime can optionally emit OpenTelemetry traces and metrics. The Docker image in this repo is for the Node server only and publishes linux/amd64 and linux/arm64 variants.

Run

npm run start
# to test
npm run test

Local check URLs (Node server):

• http://localhost:3000/
• http://localhost:3000/health
• http://localhost:3000/rss

Optional environment variables:

• PORT (default: 3000)
• MAX_ITEMS (default: 1000) limits RSS item count
• FEED_BASE_URL (optional) overrides feed links base URL
• OTEL_ENABLED set to true to enable Node-only OpenTelemetry
• OTEL_EXPORTER_OTLP_ENDPOINT shared OTLP backend endpoint for traces and metrics
• OTEL_EXPORTER_OTLP_TRACES_ENDPOINT optional trace endpoint when not using shared endpoint
• OTEL_EXPORTER_OTLP_METRICS_ENDPOINT optional metrics endpoint when not using shared endpoint
• OTEL_SERVICE_NAME (optional) overrides the default service name models.dev-rss

Docker (linux/amd64 + linux/arm64)

Build for your local platform:

docker build -t models.dev-rss:local .

Build a specific target architecture:

docker build --platform linux/amd64 -t models.dev-rss:amd64 .
docker build --platform linux/arm64 -t models.dev-rss:arm64 .

Run it:

docker run --rm \
  -p 3000:3000 \
  models.dev-rss:local

All Node env vars work the same in Docker:

docker run --rm \
  -p 8080:8080 \
  -e PORT=8080 \
  -e MAX_ITEMS=250 \
  -e FEED_BASE_URL=https://rss.example.com \
  -e OTEL_ENABLED=true \
  -e OTEL_EXPORTER_OTLP_ENDPOINT=https://otel.example/v1 \
  -e OTEL_SERVICE_NAME=models-dev-rss \
  models.dev-rss:local

If you change PORT, match the published port mapping to the container port.

On a different host architecture, pass --platform linux/amd64 or --platform linux/arm64 explicitly.

GitHub Container Registry (linux/amd64 + linux/arm64)

GitHub Actions publishes a multi-arch image to ghcr.io/flexdinesh/models.dev-rss.

Workflow:

• Publish Docker image (amd64 + arm64)

Tags:

• sha-<full-commit-sha> on every main push
• latest on the default branch

Pull and run the published image:

docker pull ghcr.io/flexdinesh/models.dev-rss:latest

docker run --rm \
  -p 3000:3000 \
  ghcr.io/flexdinesh/models.dev-rss:latest

If you need a specific published target, pull or run with --platform linux/amd64 or --platform linux/arm64.

OpenTelemetry

• Node server only. Worker runtime stays uninstrumented.
• Disabled by default.
• Uses OpenTelemetry auto-instrumentation.
• When OTEL_ENABLED=true, you must set either:
  • OTEL_EXPORTER_OTLP_ENDPOINT, or
  • both OTEL_EXPORTER_OTLP_TRACES_ENDPOINT and OTEL_EXPORTER_OTLP_METRICS_ENDPOINT
• If OTel config is invalid or SDK init fails, the server logs a warning and keeps serving traffic.

Example:

OTEL_ENABLED=true \
OTEL_EXPORTER_OTLP_ENDPOINT=https://otel.example/v1 \
OTEL_SERVICE_NAME=models-dev-rss \
npm run start

Endpoints

• GET / plain text usage hint
• GET /health returns JSON liveness status without calling upstream
• GET /rss fetches api.json, converts to RSS 2.0, returns application/rss+xml
• GET /rss?providerId=openai&providerId=openrouter filters the feed to matching upstream provider ids and appends the matched provider names to the channel title

Cloudflare Worker

npm run dev
npm run deploy

Local check URLs (Wrangler dev):

• http://localhost:8787/
• http://localhost:8787/health
• http://localhost:8787/rss

For Worker MAX_ITEMS, add it as an env var in wrangler.toml:

[vars]
MAX_ITEMS = "1000"
FEED_BASE_URL = "https://your-domain.example"