[dezren39]

Pre-filed by the quota-check job (go-plan-expiry-watch) on 2026-10-31, per plan v1 audible A3 (#2173). Do not merge on filing - merging/activation is a separate decision for the Nov 2 trough; this PR exists so the change is ready and reviewable ahead of time.

What it does

• workhorse-primary: parks the DeepSeek Flash leg (removed from targets/weights) and makes the Meta door (muse-direct/muse-spark-1.3-contributor - the ONE Meta Muse API High Usage pool) the largest seat (weight 3). Gemini Flash keeps weight 2, Sonnet weight 1.
• Nothing else changes. Revert steps are written in the route comment (restore the DeepSeek leg at weight 1 and the Meta weight to 1, on Nov 5 as DeepSeek seats return - audible A4).

Why (quota figures at filing, laptop gatherer 2026-10-31T07:15Z)

• DeepSeek plan doors are closed through the trough: OpenCode Go Lab monthly 100%, Go Personal monthly 100% (Personal's plan ends Nov 2 21:02 CT, 0 pts stranded), ClinePass monthly 100% until Nov 5 19:22 CT.
• Meta weekly 94.49% pre-reset; the pool resets Nov 2 00:00Z with a fresh 100 pts and is the trough's main carrier alongside Codex (weekly 74%, resets Nov 4 06:35Z; A3's operational part queues Luna/Sol-fit work into Nov 2-4 separately).
• Leaving a walled DeepSeek leg in the primary round-robin costs a wall + failover on every pick through Nov 2-4.

Filed by quota-check run 2026-10-31T07:24Z; recorded in the job state file and plan v10.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 48 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 194379a6-7735-4e96-ae5d-4d99e3077e64

Commits

Reviewing files that changed from the base of the PR and between cffd524 and a6e46f8.

Files selected for processing: 1

• .opencode/opencode.jsonc

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-31T07:36:50.850715Z	a6e46f8	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[dezren39]

Merge pass 2026-11-01: parking this PR rather than resolving its conflict. (1) Its own comment says merge is a separate decision for the Nov 2-4 trough; (2) it now conflicts in .opencode/opencode.jsonc; (3) the live config changed this morning — the provider_routes collapse landed (any resolution must preserve experimental.provider_routes and must not resurrect the explicit openai-plan-* routes) and a muse-direct disable landed (c6bf933), while this PR seats muse-direct as workhorse primary at weight 3. Resolution task for the #2173 A3 owner: #3815

[dezren39]

Closing as superseded (queue task 7377, tracker #3815). Not merged: current main already contradicts the premise, and the lines it edits no longer exist in that shape.

What this PR does. Removes opencode-route/deepseek-flash#high from workhorse-primary and lifts muse-direct/muse-spark-1.3-contributor to weight 3, on the Oct-31 premise that the DeepSeek plan doors were 100% walled through the Nov 2-4 trough and Meta was the trough's carrier.

What main says now (.opencode/opencode.jsonc at 8e76c6fa5, 2026-11-03, inside experimental.provider_routes ~line 4353):

"targets": [
  "openai-plan/gpt-6-luna#xhigh",
  "muse-direct/muse-spark-1.3-contributor#xhigh",
  "opencode-route/deepseek-flash-plan#high"
],
"weights": { "openai-plan/gpt-6-luna": 2,
             "muse-direct/muse-spark-1.3-contributor": 2,
             "opencode-route/deepseek-flash-plan": 2 }

The route's own 2026-11-02 (Drew) note records the reversal: "DeepSeek Flash plan moves INTO primary … three-way round-robin", with the side benefit that "a third of primary picks leave the Meta pool, which is pacing to exhaust days before its Nov 9 reset". Separately, the disabled_providers note says Meta was re-enabled on 2026-11-02 ("REVERSED … The Meta pool reset … reads fresh 0% with 100 pts headroom").

Why merge is wrong, not just conflicted. Landing this would re-park a DeepSeek leg Drew deliberately re-seated two days after filing, and raise Meta to weight 3 exactly while the route note says the Meta pool is pacing to exhaust. The file also moved under the provider_routes collapse, so the diff's context is gone — same reason the 2026-11-01 merge pass parked it.

Not deleted, not rewritten. Reopen or re-file if the trough still needs a different mix for the remaining window — but the Nov 2-4 premise this was pre-filed for has itself been reversed, and the scheduled revert was Nov 5 (A4).