feat(providers): add copilot_cli agent-CLI LLM provider - #679
NamigGadir wants to merge 1 commit into
Conversation
4dfe9d5 to
e2fb909
Compare
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: Namig Gadirov <ngadirov@tbcbank.com.ge>
e2fb909 to
af4acaf
Compare
|
Hi @NamigGadir, thanks for working on this — always good to see independent interest in a Copilot provider. Flagging that #572 has been under active maintainer review for a while and already covers this ground with the controls the review has required: exact version pin with per-completion preflight, home and XDG migration-source hook audits, env scrubbing, adversarial transport tests, and live verification on 1.0.88 and 1.0.89. Your auth check (existing ~/.copilot directory counts as authenticated) and the missing version gate would both block on the bar set in those rounds. Suggest closing this in favor of #572 — happy to collaborate there. Two ideas of yours I'd like to evaluate as follow-ups on |
|
Thanks @Yoseph-Zuskin for flagging #572! I didn't realize there was already an active PR in review for copilot_cli. Makes total sense regarding the strict version gate and preflight requirements. |
Summary
This PR adds a new
copilot_cliprovider for local GitHub Copilot CLI-backed LLM access.It integrates with the project's existing agent-CLI provider pattern and enables:
copilotbinary without exposing untrusted prompt content in argvSKILLSPECTOR_MODELWhy
Some users want to run model inference through the local GitHub Copilot CLI instead of a remote API. This keeps the provider consistent with the existing
claude_cli,codex_cli,gemini_cli, andopencode_cliintegrations while preserving the project's hardened security model.Changes
copilot_cliprovider registrationValidation
pytest tests/unit/test_agent_cli.py