[bong-water-water-bong]

coder-llm on the Strix Halo box serves Qwen3-Coder-30B-A3B Q4_K_M under the alias qwen3-coder, the house model (owner's call, 2026-09-28). The workflow, .pr_agent.toml and CONTRIBUTING still named zaya1-8b. The server answers any model name, so this is naming only: the readiness probe, the review log and the docs now say what actually runs.

No permission, trigger or runner change; the workflow diff is the model name and its comment.

Context: #183's review failed with a connection error while coder-llm was restarting during the model switch; it was rerun once the server was up.

🤖 Generated with Claude Code

[context7]

Docs7 for 1bit-monster/engine

Result	Status	Action
Deployment	➖ Not used	—
Content review	➖ Did not run. This site has no agent runs available this month. Wait for the monthly reset or check your Docs7 plan.	—

Commit fa35397

[github-actions]

PR Reviewer Guide 🔍

(Review updated until commit fa35397)

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶 183 - Partially compliant Compliant requirements: Implementation of --dflash flag and integration with llama-server's draft-dflash Exclusive use with --mtp enforced Performance improvements measured and documented Smoke tests pass Non-compliant requirements: Docs update in same PR (though docs/serve.md is not in diff, it was mentioned in the ticket) Requires further human verification: Verification that the performance claims are accurate on the hardware Confirmation that the docs/serve.md was updated in the same PR (not visible in diff)

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Model Name Mismatch The PR-Agent workflow configuration is updated to use qwen3-coder model, but the actual model name used in the readiness check and configuration file is openai/qwen3-coder. This could lead to confusion or misconfiguration if the server expects a different model identifier. CONFIG__MODEL: openai/qwen3-coder Model Name Mismatch The PR-Agent configuration file is updated to use openai/qwen3-coder as the model name, but the workflow file uses qwen3-coder. This inconsistency could cause issues if the server or PR-Agent expects a specific model identifier format. model = "openai/qwen3-coder"

[github-actions]

Persistent review updated to latest commit fa35397