chore(ai-agents): bump the controller+agent pair to v0.10.8 for the GeminiAgent vendor - #1711
Merged
Merged
Conversation
…eminiAgent vendor The platform's first gemini review (2026-09-02) proved every layer but this one: the credential reached agent-secrets, the render mounted GEMINI_API_KEY, the sync applied the CRs, dispatch qualified — and the run failed "Not logged in · Please run /login", because the deployed v0.10.6 controller predates gemini routing and fell through to the Claude adapter. GeminiAgent (gemini-* → @google/gemini-cli) shipped in v0.10.8. Both digests move together, as the pair contract requires: the GeminiSetup installer lives in the agent image's init container, so a new controller with the old agent image would route to a CLI that was never installed. Digests fetched from ghcr for the v0.10.8 tags of both images. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
|
Lore is reviewing this PR — 8e132b62-82e4-4587-8dba-94d803cbd0cf. Reply to any review comment to discuss or approve a fix (e.g. "ok, fix it"). Comment |
🔍 Lore Spec ImpactExamined 1 changed file(s): 0 had graph data (no coupling found), 1 had none — no ingested test run covers them, so this check cannot speak for them. Deterministic · graph @ |
Contributor
|
Lore code-review run failed (failed — node "review" failed: Not logged in · Please run /login — Unrecognized failure — see the Event Timeline metadata and agent pod logs.) — 8e132b62-82e4-4587-8dba-94d803cbd0cf. Comment |
gedaiu
added a commit
that referenced
this pull request
Sep 1, 2026
…#1712) Second half of the gemini vendor chain. v0.10.8 (deployed in #1711) added the `GeminiAgent` routing, but its installer curls `https://dl.google.com/gemini/install.sh` — a URL that does not exist, so every gemini run died at init with `curl: (22)` before the agent container ever started. v0.10.9 carries [ai-agent-subsystem#210](re-cinq/ai-agent-subsystem#210): the installer now fetches the `@google/gemini-cli` npm tarball — a dependency-free self-contained bundle — and extracts it with curl+tar, which is all the debian-slim init container has, dropping a node wrapper the main container (the one with node) executes. The fix was verified in real `debian:bookworm-slim` and `node:22-bookworm` containers before it shipped. Controller and agent digests move together as the pair contract requires. Render check passes. After the `deploy-ai-agents` rollout I'll re-trigger the #1699 review as the end-to-end gemini proof. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The platform's first Gemini review run (tonight, on #1699) proved every layer except this one: the key reached
agent-secrets, the catalog render mountedGEMINI_API_KEY, the sync loop applied the qualified CRs, the status table flipped to applied, and dispatch routed to them — and then the run failed withNot logged in · Please run /login. That is the Claude CLI's message: the deployed controller is pinned at v0.10.6, which predates Gemini support, so agemini-*model fell through the vendor routing to the Claude adapter. TheGeminiAgentvendor (gemini-*→@google/gemini-cli) shipped in v0.10.8.Both digests move together, as the chart's pair contract requires: the
GeminiSetupinstaller lives in the agent image's init container, so a new controller with the old agent image would route to a CLI that was never installed in the pod.Digests fetched from ghcr for the v0.10.8 tags of both images; the render check passes.
After merge this needs one deploy of the umbrella (a
terraform applypicks it up); I'll re-trigger the #1699 review as the end-to-end proof.🤖 Generated with Claude Code