Repository navigation
feat(skillgen): run the Harness on GLM 5.3 through OpenCode Go - #529
Merged
Merged
Conversation
Skillgen runs up to 96 model calls per job. GLM 5.3 on OpenCode Go lists at $1.40 per million input tokens and $4.40 per million output tokens, well under Sonnet 4.6, so each run costs less. The gateway gains an opencode-go route with the same guarantees as the Anthropic and Google routes: Worker-held credential, forced model, 8,192 output token cap, hosted execution blocked, and the job call limit. Extra choices and service tiers are removed because both raise the price. The pinned OpenCode 1.18.31 bootstrap lists opencode-go/glm-5.3 from its bundled snapshot with model fetches disabled and no network, so the runner needs no explicit provider declaration.
harlan-zw
added
the
harlan-agent-review
harlan-github-agent
Bot
added
harlan-agent-review-required
This was referenced Oct 7, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
🔗 Linked issue
Related to #528
❓ Type of change
📚 Description
Skillgen should run OpenCode on GLM 5.3 instead of Claude Sonnet. The Worker gateway only proxied Anthropic and Google, so the container could not reach OpenCode Go. The gateway now routes
opencode-gotohttps://opencode.ai/zen/go/v1/chat/completionswith the Worker's key. It forces the model, caps output at 8,192 tokens, allows only function tools, and counts each call against the job limit.wrangler.jsoncmakesopencode-go/glm-5.3the default. The Anthropic and Google paths are unchanged.The pinned OpenCode bootstrap (
opencode-ai1.18.31) already resolvesopencode-go/glm-5.3with model fetching disabled, so the runner needs no change.📝 Migration
Set the secret before this merges. A merge deploys the Worker, and without the key every Skillgen job fails.
cd workers/skill-harness && pnpm exec wrangler secret put OPENCODE_API_KEYgh workflow run deploy-cloudflare.yml --ref feat/skillgen-glm-5-3 --field target=harness-proofdocs/runbooks/skill-harness-proof.mddescribes, and check the model call count and review findings.