Skip to content

[M7] Test fake HTTP coding integration end to end - #8

Merged
subrosa5 merged 1 commit into
mainfrom
kimi/m7-fake-inference-integration
Nov 9, 2026
Merged

subrosa5 merged 1 commit into
mainfrom
kimi/m7-fake-inference-integration

Conversation

@subrosa5

@subrosa5 subrosa5 commented Nov 9, 2026

Copy link
Copy Markdown
Owner

Summary

  • Add a test-only scripted OpenAI-compatible HTTP endpoint with no external AI calls.
  • Cover HTTP provider → ModelPatchProvider → diff validation → Git worktree → sandbox → bounded repair/audit.
  • Add a real Docker compileall/unittest variant using the same HTTP/model path.
  • Keep the fixed diff explicitly marked as MOCK_NOTIONAL_DIFF; benchmark and prepared/golden patches are untouched.
  • Document the limits of integration simulation; production M7 code is unchanged.

Local validation

  • Python 3.14.7: 81 passed, 23 skipped (PostgreSQL/Docker unavailable on this Mac).
  • ruff check, ruff format --check, and git diff --check passed.
  • Python 3.11/3.13 full gates are required in CI, including Docker.

Limits

This proves integration and guardrails only. It does not prove model generation quality, performance, GPU, offline operation, or real-model behavior. Embeddings remain fake. Hardware/model validation on the RTX 4070 Ti remains a separate mandatory M7 gate.

@subrosa5
subrosa5 merged commit e6321cd into main Nov 9, 2026
4 checks passed
@subrosa5
subrosa5 deleted the kimi/m7-fake-inference-integration branch November 9, 2026 03:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant