Enables distributed processing and resource sharing across the Covia network via grid operations. Provides access to remote venues, distributed job execution, and collaborative computing capabilities. Perfect for scaling computational tasks, leveraging distributed resources, and building resilient, distributed AI applications.
| Name: | grid |
| Class: | covia.adapter.GridAdapter |
This adapter provides 4 MCP tool(s) for AI agent integration:
| Tool Name | Description |
|---|---|
| grid_run | Execute an operation and return its result. The operation argument accepts any resolvable form: v/ops/json/merge (venue catalog), o/my-pipeline (your pin), a/<hash> (asset), or did:web:<venue>/v/ops/... (remote venue). Pass input as a JSON map matching the operation's schema. Blocks until complete — use grid_invoke for long-running work. |
| grid_invoke | Submit any operation on the local or a remote venue and return immediately with a job snapshot — does not block. Use grid_run if you want the result inline. The escape hatch for invoking operations not exposed as MCP tools: discover with covia_inspect path=v/ops, drill into a group, then invoke any path. The operation argument accepts every resolvable form: v/ops/json/merge (venue catalog), o/my-pipeline (your pin), a/<hash> (asset by content hash), or did:web:<host>/v/ops/... (remote venue). Returns a job snapshot — poll with grid_job_status or wait for the result with grid_job_result(id=<returned id>). |
| grid_job_result | Wait for a Covia grid job to finish and return its output. Use after agent_request (or any async invocation) to retrieve the result. With no timeout, blocks until the job completes. With timeout, errors if the job hasn't completed in time — callers wanting a result treat missing-result as failure. Retry (with a longer timeout) or fall back to alternative work if a timeout occurs. |
| grid_job_status | Fetch the current status of a Covia grid job by ID |