Summary

• expose OpenAI/Codex models in OpenCode alongside Claude, Gemini, and OSS
• expose OSS models in Pi alongside Claude, OpenAI/Codex, and Gemini
• route Pi OSS models through the MLflow Responses API so GLM and Kimi complete with valid stream termination
• exclude incompatible gpt-oss model services from Codex discovery instead of exposing options that fail in Pi and OpenCode
• extend focused and end-to-end coverage for discovery, provider config, selectors, defaults, and live OSS launches

Testing

• .venv/bin/ruff check .
• .venv/bin/pytest tests/test_agent_pi.py tests/test_agent_opencode.py tests/test_agents_init.py tests/test_databricks.py tests/test_cli.py tests/test_lint.py -q (471 passed)
• live validation with a Databricks profile across representative Claude, OpenAI/Codex, Gemini, GLM, and Kimi models in both clients
• exhaustive live per-model E2E for OpenCode and Pi (2 passed in 239.58s)