@kien-le/selector-pipeline
v0.1.0
Published
Local selector pipeline service for Semantic DOM browser extension clients
Maintainers
Readme
Selector Pipeline
Local selector pipeline service for Semantic DOM browser extension clients.
The service listens on 127.0.0.1:4317 by default. The extension sends page snapshots over WebSocket, the service runs the multi-stage selector pipeline, and the extension verifies candidate selectors in the browser.
Quick Start
npx @kien-le/selector-pipelineThis also accepts an explicit serve subcommand:
npx @kien-le/selector-pipeline serveThe CLI prompts for missing model configuration in an interactive terminal:
endpointdeploymentsubscription_keyapi_version
You can also provide them as environment variables:
endpoint="https://your-resource.openai.azure.com" \
deployment="your-deployment" \
subscription_key="your-key" \
api_version="2024-02-01" \
npx @kien-le/selector-pipelineOptions
selector-pipeline --host 127.0.0.1 --port 4317 --checkpoints ./checkpoints--hostchanges the host to bind. Defaults to127.0.0.1.--portchanges the HTTP/WebSocket port. Defaults to4317.--checkpointschanges where run artifacts are written. Defaults to./checkpoints.
Runtime Shortcuts
When running in an interactive terminal:
eedits model configuration without restarting the service.pprints the current model configuration, maskingsubscription_key.qstops the service.
Extension Integration
Start this service before using the extension's local pipeline mode. The extension connects to:
ws://127.0.0.1:4317The health endpoint is available at:
http://127.0.0.1:4317/health