npx workspai workspace eval initInitialize one live, versioned Workspace Intelligence evaluation run for a fixed task and retrieval strategy.
Workspace Intelligence command
workspai workspace eval
Measure model usage, agent activity, cost provenance, and verified task outcomes without storing prompt or response bodies.
- Category
- Workspace Intelligence
- Workspace Intelligence
- Yes
- Produced artifacts
- 2
- Related contracts
- 4
- Registration
- contract-declared
- Registered options
- 0
Operational role
This purpose is versioned in the CLI runtime command surface. It is not inferred from page copy or a generic command template.
Measure model usage, agent activity, cost provenance, and verified task outcomes without storing prompt or response bodies.
Canonical invocation
This invocation is derived from the runtime command surface. Commands with a registered intelligence execution include their canonical evidence-producing flags.
npx workspai workspace eval status --jsonOperational variants
These flag or action variants are separately described by the runtime contract but remain part of this canonical command rather than separate command pages.
npx workspai workspace eval record --jsonValidate one privacy-bounded usage or outcome event from stdin and append it to the live evaluation artifact.
npx workspai workspace eval report --jsonFinalize the live evaluation into an authoritative last-run report for IDE, CI, and comparison consumers.
Runtime signature
Arguments, aliases, and options come from the versioned snapshot of the live Commander tree emitted by the CLI. Contract-declared delegated commands intentionally expose only their canonical invocation.
This command is declared by the cross-runtime capability contract and delegated outside the npm Commander tree. Its invocation remains canonical, but runtime-specific flags are documented by the selected project adapter.
Required input
Input transport and validation are declared by the CLI runtime contract; consumers should not infer them from terminal examples.
- Transport
stdin- Media type
application/json- Required
- No
- Schema version
model-usage-event.v1
contracts/workspace-intelligence/model-usage-event.v1.jsonOutput modes
Output formats are contract-declared. Raw renderer text and structured JSON are separate integration boundaries.
| Selector | Format | Media type |
|---|---|---|
| init <task> [strategy] --json | json | application/json |
| record --json | json | application/json |
| status --json | json | application/json |
| report --json | json | application/json |
| compare --from <report> --json | json | application/json |
Contract relationships
Capability contracts govern whether the command exists; artifact contracts govern the structured output it produces.
Produced artifacts
Artifacts are durable machine-readable outputs. Their schema, path, and producer relationship must remain stable across CLI, CI, IDE, MCP, and agent consumers.
| Artifact | Schema version | Contract |
|---|---|---|
| .workspai/reports/workspace-intelligence-evaluation-last-run.json | workspace-intelligence-evaluation.v1 | contracts/workspace-intelligence/workspace-intelligence-evaluation.v1.json |
| .workspai/reports/workspace-intelligence-evaluation-live.json | workspace-intelligence-evaluation.v1 | contracts/workspace-intelligence/workspace-intelligence-evaluation.v1.json |
Operational interpretation
Run this command inside the scope documented by its category. Treat declared structured output, raw renderer modes, and generated artifacts according to their published media type; human terminal prose must not be parsed as a stable contract. Consumers must preserve schema version, provenance, freshness, and verdict semantics when forwarding the result.