This repository has two active tracks:
exploration/: ad hoc and deep-dive research scripts/reports.validation/: repeatable regression-style validation harness.
exploration/README.mdvalidation/README.md
- Install
uv - Azure & Foundry Setup: set up Azure account, create Foundry resource/project, and deploy models.
- Authenticate locally:
az loginuv sync --lockedcat > .env <<'EOF'
AZURE_SUBSCRIPTION_ID=<your-subscription-id>
AZURE_RESOURCE_GROUP=<your-resource-group>
FOUNDRY_RESOURCE_NAME=<your-foundry-resource-name>
FOUNDRY_PROJECT_NAME=<your-project-name>
AZURE_AI_PROJECT_ENDPOINT=https://<your-foundry-resource-name>.services.ai.azure.com/api/projects/<your-project-name>
AZURE_AI_MODEL_DEPLOYMENT_NAME=gpt-5-mini
AGENT_NAME_PREFIX=ValidationAgent
EOFuv run exploration/deep_dive/list_models.py
uv run exploration/deep_dive/trace_openai_requests.pybash validation/scripts/run_python_validation.sh --default-model-only --retries 1