CAcaveman

POST /projects/{projectId}/eval-plan

POST /api/v1/projects/{projectId}/eval-plan on the Caveman Cloud control plane. Requires the trace:read_metadata scope. Exposed to autonomous agents as the plan action of the caveman_quality tool (read access). Quality per workload: pass rates, issues, criteria and how far their judges are trusted, the setup Caveman proposes, judges and their calibration, CI suites, and verdicts with the cases that changed. Plan prices a comparison before anything runs. A criterion's trend reads its failure rate over time and across models from sampled live traffic; its comparisons are observed, not verified. The Review queue lists the units waiting for a person's label.

post/projects/{projectId}/eval-plan

Request

  • Base URL
    https://api.caveman.so/api/v1
  • URL
    https://api.caveman.so/api/v1/projects/{projectId}/eval-plan
  • Auth
    • OAuth 2
      Authorization code: /api/v1/oauth/authorize
      Token URL: /api/v1/oauth/token scopes: trace:read_metadata
    • or HTTP bearer
    • or API key in cookie cave_access
    • or API key in cookie __Secure-cave_identity.session_token

Request body

workload_idstring uuid required
metric'preference' | 'pass_rate'
marginnumber
claim'cost_and_quality' | 'quality_only'
criteriastring[]
cases_availableinteger
trialsinteger
powernumber

Response

Successful response. Exact status depends on synchronous or durable delivery outcome.

Changes

No recorded changes to this operation across all 1 revision of this API.