Reference

Versions

Verified against the Monocle source. OKAHU_MAX_FACTS and the setup_test_cases discovery window arrived in 0.8.15 (the current release). The three scenario-test variables (marked † below) ship with scenario testing, which is on Monocle main and not yet in a published release — see Create test scenarios. The lg-simple-agent example pins 0.8.9; run pip install -U monocle_apptrace monocle_test_tools for the released features.

Environment variables

Variable	Used by	Purpose
OKAHU_API_KEY	Monocle, test tools, skills	Okahu credential
MONOCLE_EXPORTER	Monocle	Where traces go: okahu, console, otlp, … (comma-separated)
MONOCLE_WORKFLOW_NAME	Monocle (zero-code)	Workflow name
MONOCLE_TEST_WORKFLOW_NAME	Test tools	Workflow name on test traces (default: git repo name)
MONOCLE_EXPORT_FAILED_TESTS_ONLY	Test tools	Export only failing tests' traces
MONOCLE_TEST_AGENT_MODEL †	Scenario tests	Model that plays the user, e.g. openai:gpt-4.1-mini. Required (no default — raises if unset)
MONOCLE_TEST_JUDGE_MODEL †	Scenario tests	Model that judges each reply (default: same as MONOCLE_TEST_AGENT_MODEL)
MONOCLE_TRACE_TEST_AGENT †	Scenario tests	Also trace the simulated user and judge (default false)
MONOCLE_EVAL_MATRIX	Test tools	Write a per-test eval results file
OKAHU_MAX_FACTS	Test tools	Max facts a setup_test_cases discovery window may hold before it refuses. Precedence: max_facts argument → this → 1000 (since 0.8.15)

† Scenario testing is on Monocle main, not yet in a published release. | Provider keys (OPENAI_API_KEY, …) | Scenario tests | For the models named above |

API quick reference

Auth on every call: x-api-key: okh_...

Purpose	Method and path
List evals	GET https://eval.okahu.co/api/v1/eval/templates (add ?kind=custom for your own)
Save or version a custom eval	POST / PUT https://eval.okahu.co/api/v1/eval/templates[/{id}]
Run an eval	POST https://eval.okahu.co/api/v1/eval/jobs?workflow_name=…&fact_name=…&(trace_id=… \| start_time=…&end_time=…)
Eval job status	GET https://eval.okahu.co/api/v1/eval/jobs/{job_id}/status
Record a label	POST https://eval.okahu.co/api/v1/eval/label?trace_id=…&fact_name=…&workflow_name=…
Train an eval from labels	POST https://eval.okahu.co/api/v1/eval/training/jobs?… then GET …/training/jobs/{job_id}
Eval results report	POST https://api.okahu.co/api/v1/workflows/{name}/evals/report
SLAs	PUT https://api.okahu.co/api/v1/apps/{app}
Measures	GET https://api.okahu.co/api/v1/insights/apps/{app}/facts/{fact}
Error groups	GET https://ml-analysis.okahu.co/api/v1/insights/apps/{app}/error_groups
Schedules	POST / GET https://api.okahu.co/api/v1/schedules
Usage	POST https://api.okahu.co/api/v1/usage/query
MCP server	https://mcp.okahu.co/mcp