Generate an AI analysis report
After a run, QuOptuna can generate a governance-ready report summarizing performance, SHAP interpretation, and recommendations. The report is produced by an analyst + reviewer agent pair, built on the OpenAI Agents SDK + LiteLLM.
Requirements
Section titled “Requirements”- Internet access.
- A provider API key, entered in the web UI’s Settings and sent with each report request (the
api_keyfield). Requests without a key are rejected with 400.
Providers
Section titled “Providers”Set llm_provider to one of:
llm_provider | Provider |
|---|---|
openai | OpenAI |
google (default) | Google Gemini |
anthropic | Anthropic Claude |
Pick the model with model_name (default gpt-4o; choose one that matches your provider). Requests are routed through LiteLLM as <provider>/<model_name>.
Generate from the web UI
Section titled “Generate from the web UI”The report is the final wizard step, Report. Run through the wizard, then produce the report from that step.
Generate via API
Section titled “Generate via API”POST /api/v1/analysis/reportCall this after a study and its analysis snapshot have completed. Required body fields: optimization_id, analysis_snapshot_id, analysis_revision, api_key. Optional: trial_number, llm_provider, model_name, dataset_description, sensitive_feature, analyst_instructions/reviewer_instructions (prompt overrides), and enable_review (default true; runs the reviewer pass).