Integrate with Grid
Choose the supported public boundary for model lifecycle, values, realtime events, training, publishing, and connector data.
Integrate with Grid
Grid's public HTTP boundary is the Node API. It owns authentication, HTTP connections, realtime fan-out, and connector ingress. Model state and execution remain owned by the Rust Runtime Host behind it.
Start here
- Run the API quickstart to deploy a protected model, write one input, and resolve one output.
- Use example verification contracts for reproducible models, fixtures, assertions, and execution receipts.
- Use the coding corpus and coverage registries for model training, retrieval, and private evaluation planning.
- Read API conventions and authentication before building retries or error handling.
- Use the model lifecycle reference for source, input, symbol, range, rename, and delete operations.
- Add realtime updates when a client needs changing values without polling.
- Use connector ingress for declared source bindings, not as an arbitrary cell-write route.
- Train and integrate predictive models when you need durable model versions, explicit pins, or calibrated upper endpoints.
- Publish Grid artifacts with exact signed graphs and the 0.65 repository ownership boundaries.
Public and private boundaries
Applications send requests to the Node API, normally port 3000 in local development. Do not connect an application directly to Runtime Host's /internal/* routes. Those routes are deployment-internal contracts and require a separate private credential.
These pages show their own version status. Canonical 0.65.0 candidate pages and locally reverified site guides can appear beside execution receipts and films retained on an older release. The canonical Grid language and product contracts remain the Grid Core documentation linked throughout this site.