You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
✨ Meetings agent-backend: remaining backends (copilot, omp, codex, gemini, gateways, vllm, llm-d) and token-only Claude sign-in #18549
#18517 / PR #18528 added meetings.llm.mode: agent-backend, so Betty and Spek can use the hive's configured agent backends. It shipped Claude CLI (warm streaming process, Sonnet 5.5 default), LiteLLM through the OpenAI path, the shared completion-profile interface and the Settings model picker. The operator's requirement is that Betty can use any backend and model the hive's agents use, so the rest is tracked here.
Remaining
copilot and omp: add completion profiles (single turn, no tools, no writes, usage extraction) on the pinned versions (copilot 1.0.88, omp 18.20.0). Verify latency and usage output against a signed-in CLI.
codex and gemini: they have no flag that disables every tool. Find a safe profile: a read-only sandbox, no network tools, an approval policy of never, plus an empty workspace. Or document why one isn't possible and refuse them with a clear message in the picker.
Named model gateways, vllm and llm-d: map them onto the existing OpenAI-compatible path so they appear in the picker.
Operator-token-only hives: hives that sign Claude in only with HIVE_CLAUDE_OAUTH_TOKEN and no credentials file. Give the meetings CLI the same sign-in the agents get, without hive reading or logging the token.
Spek on the warm process: optional. Spek currently starts one CLI process per prompt.
Done when
Every backend the agents' picker offers is either selectable for meetings, with a tested completion profile, or refused with a specific documented reason. Docs list per-backend latency.
Decomposed into four child issues (copilot/omp, codex/gemini, gateways/vllm/llm-d, token-only Claude sign-in); too many independent concerns for one PR. The optional Spek warm-process item stays here.
@clubanderson this issue was closed by @kubestellar-prow[bot] as completed.
If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.
If you are still seeing the problem, comment /reopen and it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.
Activity
clubanderson commentedon Nov 16, 2026
Claimed: an operator-dispatched agent is working on this now. Hive agents, please do not pick it up.
hivecommons-hive commentedon Nov 16, 2026
Decomposed into four child issues (copilot/omp, codex/gemini, gateways/vllm/llm-d, token-only Claude sign-in); too many independent concerns for one PR. The optional Spek warm-process item stays here.
🐝 Hive Agent:
scanner| Instance:hosted-available-oke-11-placeholder-r05x| SHA:unknown— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325
✨ meetings: copilot and omp completion profiles, gateways/vllm/llm-d,…
github-actions commentedon Nov 17, 2026
@clubanderson this issue was closed by @kubestellar-prow[bot] as completed.
If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.
If you are still seeing the problem, comment
/reopenand it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.