Skip to content

ci: open-code-review uses high reasoning effort, not max #212

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We鈥檒l occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ojungo69 merged 2 commits into from
Sep 28, 2026
Merged

ci: open-code-review uses high reasoning effort, not max #212

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3a4f55d
ci: open-code-review asks its model for high reasoning effort, not max
ojungo69 Sep 28, 2026
21ba8ab
docs: open-code-review's reasoning effort is high (review on #212)
ojungo69 Sep 28, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions .github/workflows/open-code-review.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -108,7 +108,7 @@ jobs:
llm_use_anthropic: ${{ vars.OCR_LLM_USE_ANTHROPIC || 'false' }}
# The upstream default contains a provider-specific "thinking" field.
llm_extra_body: '{}'
llm_reasoning_effort: max
llm_reasoning_effort: high
Comment thread
ojungo69 marked this conversation as resolved.
llm_timeout: '600'
review_task_timeout: '5'
effort: high
Expand Down Expand Up @@ -139,7 +139,7 @@ jobs:
llm_model: ${{ vars.OCR_LLM_FALLBACK_MODEL }}
llm_use_anthropic: ${{ vars.OCR_LLM_USE_ANTHROPIC || 'false' }}
llm_extra_body: '{}'
llm_reasoning_effort: max
llm_reasoning_effort: high
llm_timeout: '600'
review_task_timeout: '5'
effort: high
Expand Down
9 changes: 5 additions & 4 deletions docs/open-code-review.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,9 +11,10 @@ It supplements the existing Rust CI and SonarCloud checks; it does not approve o
This repository uses NVIDIA NIM: `z-ai/glm-5.3` is the primary model and
`moonshotai/kimi-k3` is the fallback. GLM is the starting choice for text-only
code review, not a claim that it outperforms Kimi on every repository. Both use
the same NIM account, OpenAI-compatible endpoint, and maximum `max` reasoning
effort. OCR's separate review effort is also at its maximum, `high` (three
review rounds); OCR does not accept `max` for that setting.
the same NIM account, OpenAI-compatible endpoint, and `high` reasoning effort
(the owner lowered it from `max` on 2026-09-28). OCR's separate review effort is
at its maximum, `high` (three review rounds); OCR does not accept `max` for that
setting.

In the repository's **Settings > Secrets and variables > Actions**, configure:

Expand Down Expand Up @@ -79,7 +80,7 @@ The alternate model is tried only once, even for these failures.
Manual reviews use the current head and still reject closed/draft PRs.
- Each job has a 45-minute timeout. Each attempt reviews one file group at a time,
with a 600-second LLM request timeout and a 500,000-token budget. The longer request
and job limits leave room for maximum reasoning and the fallback attempt.
and job limits leave room for long reasoning and the fallback attempt.
The CLI multiplies the five-minute task setting by the three rounds of explicit
`high` review effort, giving each file group a fifteen-minute deadline. The job
deadline still applies across all groups and both models. Fallback can consume
Expand Down
Loading