ci: use MiniMax M2.7 in integration default matrix#3383
Conversation
Co-authored-by: openhands <openhands@all-hands.dev>
Python API breakage checks — ✅ PASSEDResult: ✅ PASSED |
REST API breakage checks (OpenAPI) — ✅ PASSEDResult: ✅ PASSED |
all-hands-bot
left a comment
all-hands-bot
left a comment
There was a problem hiding this comment.
🟢 Taste Rating: Good taste. This is a straightforward workflow default-matrix update, and the replacement model ID is already present in resolve_model_config.py.
[RISK ASSESSMENT]
- [Overall PR]
⚠️ Risk Assessment: 🟢 LOW
CI-only config change; no production code, dependency, or API behavior changes.
VERDICT: ✅ Worth merging.
Reviewed by an AI agent (OpenHands) on behalf of the user.
Was this automated review useful? React with 👍 or 👎 to this review to help us measure review quality.
Workflow run: https://github.com/OpenHands/software-agent-sdk/actions/runs/26413263405
all-hands-bot
left a comment
all-hands-bot
left a comment
There was a problem hiding this comment.
✅ QA Report: PASS
The integration workflow default matrix now resolves MiniMax M2.7 in place of Kimi K2.6 while preserving the other default models.
Does this PR achieve its stated goal?
Yes. I exercised the setup-matrix model resolution path the workflow uses when no model_ids input is supplied: origin/main resolved kimi-k2.6, while the PR branch resolved minimax-m2.7 and still emitted gpt-5.5, deepseek-v4-flash, and gemini-3.1-pro. The PR branch also successfully wrote a matrix entry to GITHUB_OUTPUT, confirming the new default model ID is recognized by the workflow's model configuration.
| Phase | Result |
|---|---|
| Environment Setup | ✅ Python 3.13.13 and jq 1.7 were available; no dependency install was needed for this workflow step. |
| CI Status | 🟡 Latest check snapshot: 19 success, 3 skipped, 10 still in progress, no completed failures. |
| Functional Verification | ✅ Ran the workflow matrix resolution before/after and parsed the workflow YAML successfully. |
Functional Verification
Test 1: Default integration model matrix resolution
Step 1 — Establish baseline without the PR change:
Checked out origin/main and ran the same model-resolution logic used by the workflow's setup-matrix job with MODEL_IDS_INPUT empty:
--- Baseline: origin/main without PR change ---
=== origin/main baseline (27d6dcc1) ===
No model_ids specified, using defaults: gpt-5.5,deepseek-v4-flash,kimi-k2.6,gemini-3.1-pro
Resolved models: GPT-5.5,DeepSeek V4 Flash Kimi K2.6,Gemini 3.1 Pro
Matrix IDs: ["gpt-5.5","deepseek-v4-flash","kimi-k2.6","gemini-3.1-pro"]
Run suffixes: "gpt_5_5_run","deepseek_v4_flash_run","kimi_k2_6_run","gemini_3_1_pro_run"
GITHUB_OUTPUT contains matrix: 1 line
This confirms the previous default scheduled/label-triggered matrix included kimi-k2.6 and that the setup step could resolve it into a matrix.
Step 2 — Apply the PR's changes:
Checked out replace-kimi-with-minimax-m27 at 9e27789b.
Step 3 — Re-run with the PR in place:
Ran the same workflow matrix resolution with MODEL_IDS_INPUT empty:
--- PR branch: replace-kimi-with-minimax-m27 ---
=== PR branch (9e27789b) ===
No model_ids specified, using defaults: gpt-5.5,deepseek-v4-flash,minimax-m2.7,gemini-3.1-pro
Resolved models: GPT-5.5,DeepSeek V4 Flash MiniMax M2.7,Gemini 3.1 Pro
Matrix IDs: ["gpt-5.5","deepseek-v4-flash","minimax-m2.7","gemini-3.1-pro"]
Run suffixes: "gpt_5_5_run","deepseek_v4_flash_run","minimax_m2_7_run","gemini_3_1_pro_run"
GITHUB_OUTPUT contains matrix: 1 line
This shows the changed workflow default now selects minimax-m2.7, the new ID resolves through resolve_model_config.py, and the other three defaults remain unchanged.
Test 2: Workflow YAML loads
Ran a direct YAML load of .github/workflows/integration-runner.yml on the PR branch:
Loaded .github/workflows/integration-runner.yml as YAML
DEFAULT_MODEL_IDS: gpt-5.5,deepseek-v4-flash,minimax-m2.7,gemini-3.1-pro
This confirms the changed workflow file remains parseable and exposes the expected default model list.
Issues Found
None.
AI-generated QA review posted by OpenHands on behalf of the requester.
3 participants
Summary
kimi-k2.6withminimax-m2.7in the integration workflow default model matrix.Validation
uv run pre-commit run --files .github/workflows/integration-runner.ymlThis PR was created by an AI agent (OpenHands) on behalf of the maintainer.
@enyst can click here to continue refining the PR
Agent Server images for this PR
• GHCR package: https://github.com/OpenHands/agent-sdk/pkgs/container/agent-server
Variants & Base Images
eclipse-temurin:17-jdknikolaik/python-nodejs:python3.13-nodejs22-slimgolang:1.21-bookwormPull (multi-arch manifest)
# Each variant is a multi-arch manifest supporting both amd64 and arm64 docker pull ghcr.io/openhands/agent-server:9e27789-pythonRun
All tags pushed for this build
About Multi-Architecture Support
9e27789-python) is a multi-arch manifest supporting both amd64 and arm649e27789-python-amd64) are also available if needed