mi/werkrevisor
1
0
Fork 0
Code Issues Pull Requests 3 Actions Packages Projects Releases Wiki Activity

the README explains why output limit and timeout depend on the model #49

Merged
ai-reviewer merged 6 commits from model-dependent-limits into main 2026-10-08 14:17:21 +02:00
Conversation 10 Commits 6 Files Changed 4
+80 -9
claude-code commented 2026-10-08 14:00:39 +02:00
Collaborator
Copy Link
Copy Source

See the PR document.

🤖 Generated with Claude Code

See [the PR document](docs/prs/2026-10-08-PR%2349-model-dependent-limits.md). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
claude-code added 1 commit 2026-10-08 14:00:39 +02:00
PR#49: PR document for model-dependent output limit and timeout
werkator build succeeded after 03:46
Details
werkator/werkrevisor-statistics build succeeded after 00:49
Details
70668a708c 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
claude-code marked the pull request as ready for review 2026-10-08 14:01:24 +02:00
mi added 1 commit 2026-10-08 14:01:24 +02:00
The README explains why output limit and timeout depend on the model
ai-review/assessment AI review passed; approved by werkrevisor; merge waits for 2 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit
Details
ai-review/gate build succeeded after 04:08
Details
werkator build succeeded after 00:15
Details
werkator/werkrevisor-statistics build succeeded after 00:25
Details
45c9404f2e 
config/werkrevisor.toml gets the values deployed on mih09:
max_output_tokens = 48000 and max_usd_per_review = 0.25.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ai-reviewer commented 2026-10-08 14:05:31 +02:00
Collaborator
Copy Link
Copy Source
🟢 werkrevisor GREEN for 45c9404, superseded by the result for f4a658d

🟢 werkrevisor: GREEN / PASS

Gate: approved by werkrevisor; merge waits for 2 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit.

Change kind: mixed.

Reasons

  • hint 1 changed file(s) that neither weaken the checks nor touch security according to Jev
  • hint the PR document may contradict the change summary according to Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19
  • hint 1 finding(s) of low severity or rated minor by Jev: README.md:324-333
  • hint model recommends merge with caution

Findings

  1. Inconsistent head count at README.md:324-333, severity unknown, counted as medium, hint, as a line comment

  2. [medium] Requirement conflicts with the change summary at docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19, hint, as a line comment

Classifier

4 of 4 units passed all questions (1 PR document(s)).

Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 6445, calls 1, tokens in 6282 / out 6096 (reasoning included in out), duration 64.5 s, cost 0.0092 USD (reported), classifier `jev-1.13.0` calls 4, tokens in 2819, cost 0.0001 USD, finding judge `jev-1.13.0` calls 1, tokens in 2231, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1492 / out 146, cost 0.0006 USD (reported), PR document check `jev-1.13.0` calls 4, tokens in 3655, cost 0.0002 USD, build file judge `jev-1.13.0` calls 1, tokens in 729, cost 0.0000 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).
Reviewed head `45c9404f2eb54a45daa71f071cd131c70bba9f9f` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0.
<!-- werkrevisor:45c9404f2eb54a45daa71f071cd131c70bba9f9f --> <!-- werkrevisor-verdict:GREEN --> <!-- werkrevisor-folded --> <details><summary>🟢 werkrevisor GREEN for <code>45c9404</code>, superseded by <a href="http://git.javagil.de/mi/werkrevisor/pulls/49#issuecomment-1971">the result for <code>f4a658d</code></a></summary> ## 🟢 werkrevisor: GREEN / PASS Gate: approved by werkrevisor; merge waits for 2 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit. Change kind: `mixed`. ### Reasons - **hint** 1 changed file(s) that neither weaken the checks nor touch security according to Jev - **hint** the PR document may contradict the change summary according to Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19 - **hint** 1 finding(s) of low severity or rated minor by Jev: README.md:324-333 - **hint** model recommends merge with caution ### Findings 1. **Inconsistent head count** at `README.md:324-333`, severity unknown, counted as medium, hint, as a line comment 2. **[medium] Requirement conflicts with the change summary** at `docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19`, hint, as a line comment ### Classifier 4 of 4 units passed all questions (1 PR document(s)). <sub> Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 6445, calls 1, tokens in 6282 / out 6096 (reasoning included in out), duration 64.5 s, cost 0.0092 USD (reported), classifier `jev-1.13.0` calls 4, tokens in 2819, cost 0.0001 USD, finding judge `jev-1.13.0` calls 1, tokens in 2231, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1492 / out 146, cost 0.0006 USD (reported), PR document check `jev-1.13.0` calls 4, tokens in 3655, cost 0.0002 USD, build file judge `jev-1.13.0` calls 1, tokens in 729, cost 0.0000 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).<br> Reviewed head `45c9404f2eb54a45daa71f071cd131c70bba9f9f` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0. </sub> </details>
ai-reviewer approved these changes 2026-10-08 14:05:31 +02:00
Dismissed
ai-reviewer left a comment
Collaborator
Copy Link
Copy Source

werkrevisor: GREEN / PASS for 45c9404f2eb54a45daa71f071cd131c70bba9f9f.
The result comment of werkrevisor holds the details.

werkrevisor: GREEN / PASS for `45c9404f2eb54a45daa71f071cd131c70bba9f9f`. The result comment of werkrevisor holds the details. <!-- werkrevisor-review:45c9404f2eb54a45daa71f071cd131c70bba9f9f -->
README.md Outdated
@@ -309,0 +321,4 @@
A higher output limit therefore lowers the input limit within the same `[budget] max_usd_per_review`;
raise the budget with the output limit, or the prompt gets smaller.
Values measured on 4 heads in October 2026, with the settings of the example configs:
ai-reviewer commented 2026-10-08 14:05:31 +02:00
Collaborator
Copy Link
Copy Source

Inconsistent head count, severity unknown, counted as medium (werkrevisor)

hint; Jev: real 0.81, severity 1.8 of 3

The README states "Values measured on 4 heads in October 2026" but the table immediately below contains 6 rows covering 5 distinct models (Claude Sonnet 5.5, Grok 4.7, MiniMax M3 with two reasoning efforts, GPT-6 Astra, Mistral Large 4). The "4 heads" claim is either wrong or ambiguous: it does not match the 6 rows, and the PR document itself only says the values were measured per model on 2026-10-06 and 2026-10-07 without mentioning a head count. Readers cannot tell whether "4 heads" means 4 models, 4 runs per model, or something else, which undermines the table's credibility.

**Inconsistent head count**, severity unknown, counted as medium (werkrevisor) hint; Jev: real 0.81, severity 1.8 of 3 The README states "Values measured on 4 heads in October 2026" but the table immediately below contains 6 rows covering 5 distinct models (Claude Sonnet 5.5, Grok 4.7, MiniMax M3 with two reasoning efforts, GPT-6 Astra, Mistral Large 4). The "4 heads" claim is either wrong or ambiguous: it does not match the 6 rows, and the PR document itself only says the values were measured per model on 2026-10-06 and 2026-10-07 without mentioning a head count. Readers cannot tell whether "4 heads" means 4 models, 4 runs per model, or something else, which undermines the table's credibility.
claude-code commented 2026-10-08 14:06:03 +02:00
Author
Collaborator
Copy Link
Copy Source

Fixed: the sentence now says that the values come from the model comparison of 2026-10-06 and 2026-10-07, where each model reviewed the same 4 PR heads; the rows are models and settings, not heads.

Fixed: the sentence now says that the values come from the model comparison of 2026-10-06 and 2026-10-07, where each model reviewed the same 4 PR heads; the rows are models and settings, not heads.
claude-code marked this conversation as resolved
docs/prs/2026-10-08-PR#49-model-dependent-limits.md
@@ -0,0 +16,4 @@
An operator who changes the model therefore keeps limits that fit the old one,
and finds out only when reviews fail or time out.
## Non-Goals
ai-reviewer commented 2026-10-08 14:05:31 +02:00
Collaborator
Copy Link
Copy Source

[medium] Requirement conflicts with the change summary (werkrevisor)

hint

Jev compared this section 'Non-Goals' with a summary that openrouter/minimax/minimax-m3 wrote from the diff alone: description contradicts the change summary 0.72, requirement conflicts with the change summary 0.86.
Fix the PR document or the code, or reply why the section holds, and resolve this comment.
The summary says:

  • Adds a new 'Model-dependent values' section with a table of measured durations and max_output_tokens per model, and notes that max_output_tokens, timeout_seconds, and reasoning_effort depend on the model.
  • Raises max_output_tokens from 32000 to 48000 and max_usd_per_review from 0.20 to 0.25, updating the comments to reflect a higher MiniMax M3 output cost.
**[medium] Requirement conflicts with the change summary** (werkrevisor) hint Jev compared this section 'Non-Goals' with a summary that openrouter/minimax/minimax-m3 wrote from the diff alone: description contradicts the change summary 0.72, requirement conflicts with the change summary 0.86. Fix the PR document or the code, or reply why the section holds, and resolve this comment. The summary says: - Adds a new 'Model-dependent values' section with a table of measured durations and `max_output_tokens` per model, and notes that `max_output_tokens`, `timeout_seconds`, and `reasoning_effort` depend on the model. - Raises `max_output_tokens` from 32000 to 48000 and `max_usd_per_review` from 0.20 to 0.25, updating the comments to reflect a higher MiniMax M3 output cost.
claude-code commented 2026-10-08 14:06:03 +02:00
Author
Collaborator
Copy Link
Copy Source

Fixed: the Non-Goal said "no change of the defaults", while the PR changes the example config. It now says that the built-in defaults stay and only config/werkrevisor.toml changes.

Fixed: the Non-Goal said "no change of the defaults", while the PR changes the example config. It now says that the built-in defaults stay and only config/werkrevisor.toml changes.
claude-code marked this conversation as resolved
mi added 1 commit 2026-10-08 14:06:03 +02:00
PR#49: name the 4 heads of the comparison; the example config changes, the built-in defaults do not
werkator build succeeded after 00:08
Details
werkator/werkrevisor-statistics build succeeded after 00:26
Details
ai-review/assessment AI review passed; approved by werkrevisor; merge waits for 2 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit
Details
ai-review/gate build succeeded after 01:29
Details
f4a658daa4 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mi dismissed ai-reviewer's review 2026-10-08 14:06:03 +02:00
Reason:

New commits pushed, approval review dismissed automatically according to repository settings

ai-reviewer commented 2026-10-08 14:07:36 +02:00
Collaborator
Copy Link
Copy Source
🟢 werkrevisor GREEN for f4a658d, superseded by the result for 7db83f6

🟢 werkrevisor: GREEN / PASS

Gate: approved by werkrevisor; merge waits for 2 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit.

Change kind: mixed.

Reasons

  • hint 1 changed file(s) that neither weaken the checks nor touch security according to Jev
  • hint the PR document may contradict the change summary according to Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19
  • hint 1 finding(s) of low severity or rated minor by Jev: config/werkrevisor.toml:4
  • hint model recommends changes before merge

Findings

  1. Header policy violated at config/werkrevisor.toml:4, severity unknown, counted as medium, hint, as a line comment

  2. [medium] Requirement conflicts with the change summary at docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19, hint, as a line comment

Classifier

4 of 4 units passed all questions (1 PR document(s)).

Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 6471, calls 1, tokens in 6306 / out 14340 (reasoning included in out), duration 86.8 s, cost 0.0191 USD (reported), classifier `jev-1.13.0` calls 4, tokens in 2834, cost 0.0001 USD, finding judge `jev-1.13.0` calls 1, tokens in 1097, cost 0.0000 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1505 / out 158, cost 0.0004 USD (reported), PR document check `jev-1.13.0` calls 4, tokens in 3766, cost 0.0002 USD, build file judge `jev-1.13.0` calls 1, tokens in 729, cost 0.0000 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).
Reviewed head `f4a658daa4ebb23b809a5390e31bddc38550306e` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0.
<!-- werkrevisor:f4a658daa4ebb23b809a5390e31bddc38550306e --> <!-- werkrevisor-verdict:GREEN --> <!-- werkrevisor-folded --> <details><summary>🟢 werkrevisor GREEN for <code>f4a658d</code>, superseded by <a href="http://git.javagil.de/mi/werkrevisor/pulls/49#issuecomment-1979">the result for <code>7db83f6</code></a></summary> ## 🟢 werkrevisor: GREEN / PASS Gate: approved by werkrevisor; merge waits for 2 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit. Change kind: `mixed`. ### Reasons - **hint** 1 changed file(s) that neither weaken the checks nor touch security according to Jev - **hint** the PR document may contradict the change summary according to Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19 - **hint** 1 finding(s) of low severity or rated minor by Jev: config/werkrevisor.toml:4 - **hint** model recommends changes before merge ### Findings 1. **Header policy violated** at `config/werkrevisor.toml:4`, severity unknown, counted as medium, hint, as a line comment 2. **[medium] Requirement conflicts with the change summary** at `docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19`, hint, as a line comment ### Classifier 4 of 4 units passed all questions (1 PR document(s)). <sub> Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 6471, calls 1, tokens in 6306 / out 14340 (reasoning included in out), duration 86.8 s, cost 0.0191 USD (reported), classifier `jev-1.13.0` calls 4, tokens in 2834, cost 0.0001 USD, finding judge `jev-1.13.0` calls 1, tokens in 1097, cost 0.0000 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1505 / out 158, cost 0.0004 USD (reported), PR document check `jev-1.13.0` calls 4, tokens in 3766, cost 0.0002 USD, build file judge `jev-1.13.0` calls 1, tokens in 729, cost 0.0000 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).<br> Reviewed head `f4a658daa4ebb23b809a5390e31bddc38550306e` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0. </sub> </details>
ai-reviewer approved these changes 2026-10-08 14:07:36 +02:00
Dismissed
ai-reviewer left a comment
Collaborator
Copy Link
Copy Source

werkrevisor: GREEN / PASS for f4a658daa4ebb23b809a5390e31bddc38550306e.
The result comment of werkrevisor holds the details.

werkrevisor: GREEN / PASS for `f4a658daa4ebb23b809a5390e31bddc38550306e`. The result comment of werkrevisor holds the details. <!-- werkrevisor-review:f4a658daa4ebb23b809a5390e31bddc38550306e -->
config/werkrevisor.toml Outdated
@@ -1,4 +1,4 @@
# werkrevisor operator config for reviewing werkrevisor itself on git.javagil.de.
# The PR never changes this file; Werkator passes its path in WERKREVISOR_CONFIG.
[model]
ai-reviewer commented 2026-10-08 14:07:36 +02:00
Collaborator
Copy Link
Copy Source

Header policy violated, severity unknown, counted as medium (werkrevisor)

hint; Jev: real 0.76, severity 1.6 of 3

The file header states that the PR never changes this file because Werkator passes its path in WERKREVISOR_CONFIG, but this PR explicitly modifies max_output_tokens and max_usd_per_review. Either the header should be updated to reflect that the in-repo config mirrors the deployed host config, or the file change should not be part of this PR.

**Header policy violated**, severity unknown, counted as medium (werkrevisor) hint; Jev: real 0.76, severity 1.6 of 3 The file header states that the PR never changes this file because Werkator passes its path in WERKREVISOR_CONFIG, but this PR explicitly modifies `max_output_tokens` and `max_usd_per_review`. Either the header should be updated to reflect that the in-repo config mirrors the deployed host config, or the file change should not be part of this PR.
claude-code commented 2026-10-08 14:08:27 +02:00
Author
Collaborator
Copy Link
Copy Source

Fixed: the header meant that a PR cannot change the config of its own review. It now says that a PR which changes this file is still reviewed with the deployed copy; config/werkprobe.toml had the same sentence and got the same fix.

Fixed: the header meant that a PR cannot change the config of its own review. It now says that a PR which changes this file is still reviewed with the deployed copy; config/werkprobe.toml had the same sentence and got the same fix.
claude-code marked this conversation as resolved
docs/prs/2026-10-08-PR#49-model-dependent-limits.md
@@ -0,0 +16,4 @@
An operator who changes the model therefore keeps limits that fit the old one,
and finds out only when reviews fail or time out.
## Non-Goals
ai-reviewer commented 2026-10-08 14:07:36 +02:00
Collaborator
Copy Link
Copy Source

[medium] Requirement conflicts with the change summary (werkrevisor)

hint

Jev compared this section 'Non-Goals' with a summary that openrouter/minimax/minimax-m3 wrote from the diff alone: requirement conflicts with the change summary 0.69.
Fix the PR document or the code, or reply why the section holds, and resolve this comment.
The summary says:

  • Added cross-references to a new "Model-dependent values" section for max_output_tokens and timeout_seconds, and introduced that section with explanations and a table of model-specific values from comparisons on 2026-10-06 and 2026-10-07.
  • Raised max_output_tokens from 32000 to 48000 and max_usd_per_review from 0.20 to 0.25, with updated comments referencing a 2026-10-08 incident and the new costs.
**[medium] Requirement conflicts with the change summary** (werkrevisor) hint Jev compared this section 'Non-Goals' with a summary that openrouter/minimax/minimax-m3 wrote from the diff alone: requirement conflicts with the change summary 0.69. Fix the PR document or the code, or reply why the section holds, and resolve this comment. The summary says: - Added cross-references to a new "Model-dependent values" section for `max_output_tokens` and `timeout_seconds`, and introduced that section with explanations and a table of model-specific values from comparisons on 2026-10-06 and 2026-10-07. - Raised `max_output_tokens` from 32000 to 48000 and `max_usd_per_review` from 0.20 to 0.25, with updated comments referencing a 2026-10-08 incident and the new costs.
claude-code commented 2026-10-08 14:08:28 +02:00
Author
Collaborator
Copy Link
Copy Source

The section holds: the Non-Goals say that the code and the built-in defaults stay, and that only the example config config/werkrevisor.toml changes, which is what the summary describes. Jev's 0.69 is just above hint_min = 0.6; I keep it as a data point for the calibration.

The section holds: the Non-Goals say that the code and the built-in defaults stay, and that only the example config config/werkrevisor.toml changes, which is what the summary describes. Jev's 0.69 is just above hint_min = 0.6; I keep it as a data point for the calibration.
claude-code marked this conversation as resolved
mi added 1 commit 2026-10-08 14:08:28 +02:00
PR#49: the config header says that a PR is reviewed with the deployed copy
werkator build succeeded after 00:08
Details
werkator/werkrevisor-statistics build succeeded after 00:26
Details
ai-review/assessment AI review passed; approved by werkrevisor; merge waits for 1 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit
Details
ai-review/gate build succeeded after 03:07
Details
7db83f611a 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mi dismissed ai-reviewer's review 2026-10-08 14:08:28 +02:00
Reason:

New commits pushed, approval review dismissed automatically according to repository settings

ai-reviewer commented 2026-10-08 14:11:44 +02:00
Collaborator
Copy Link
Copy Source
🟢 werkrevisor GREEN for 7db83f6, superseded by the result for e6ebfc1

🟢 werkrevisor: GREEN / PASS

Gate: approved by werkrevisor; merge waits for 1 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit.

Change kind: mixed.

Reasons

  • hint 2 changed file(s) that neither weaken the checks nor touch security according to Jev
  • hint the PR document may contradict the change summary according to Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19

Findings

  1. [medium] Requirement conflicts with the change summary at docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19, hint, as a line comment

Classifier

5 of 5 units passed all questions (1 PR document(s)).

Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 10917, calls 1, tokens in 10559 / out 12882 (reasoning included in out), duration 186.0 s, cost 0.0186 USD (reported), classifier `jev-1.13.0` calls 5, tokens in 3499, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1750 / out 321, cost 0.0009 USD (reported), PR document check `jev-1.13.0` calls 5, tokens in 5624, cost 0.0002 USD, build file judge `jev-1.13.0` calls 2, tokens in 1384, cost 0.0001 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).
Reviewed head `7db83f611a11d361f3811c45c7e60971c07490fe` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0.
<!-- werkrevisor:7db83f611a11d361f3811c45c7e60971c07490fe --> <!-- werkrevisor-verdict:GREEN --> <!-- werkrevisor-folded --> <details><summary>🟢 werkrevisor GREEN for <code>7db83f6</code>, superseded by <a href="http://git.javagil.de/mi/werkrevisor/pulls/49#issuecomment-1985">the result for <code>e6ebfc1</code></a></summary> ## 🟢 werkrevisor: GREEN / PASS Gate: approved by werkrevisor; merge waits for 1 open hint(s): reply to each line comment and resolve it, then restart the review or push a commit. Change kind: `mixed`. ### Reasons - **hint** 2 changed file(s) that neither weaken the checks nor touch security according to Jev - **hint** the PR document may contradict the change summary according to Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19 ### Findings 1. **[medium] Requirement conflicts with the change summary** at `docs/prs/2026-10-08-PR#49-model-dependent-limits.md:19`, hint, as a line comment ### Classifier 5 of 5 units passed all questions (1 PR document(s)). <sub> Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 10917, calls 1, tokens in 10559 / out 12882 (reasoning included in out), duration 186.0 s, cost 0.0186 USD (reported), classifier `jev-1.13.0` calls 5, tokens in 3499, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1750 / out 321, cost 0.0009 USD (reported), PR document check `jev-1.13.0` calls 5, tokens in 5624, cost 0.0002 USD, build file judge `jev-1.13.0` calls 2, tokens in 1384, cost 0.0001 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).<br> Reviewed head `7db83f611a11d361f3811c45c7e60971c07490fe` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0. </sub> </details>
ai-reviewer approved these changes 2026-10-08 14:11:44 +02:00
Dismissed
ai-reviewer left a comment
Collaborator
Copy Link
Copy Source

werkrevisor: GREEN / PASS for 7db83f611a11d361f3811c45c7e60971c07490fe.
The result comment of werkrevisor holds the details.

werkrevisor: GREEN / PASS for `7db83f611a11d361f3811c45c7e60971c07490fe`. The result comment of werkrevisor holds the details. <!-- werkrevisor-review:7db83f611a11d361f3811c45c7e60971c07490fe -->
docs/prs/2026-10-08-PR#49-model-dependent-limits.md
@@ -0,0 +16,4 @@
An operator who changes the model therefore keeps limits that fit the old one,
and finds out only when reviews fail or time out.
## Non-Goals
ai-reviewer commented 2026-10-08 14:11:44 +02:00
Collaborator
Copy Link
Copy Source

[medium] Requirement conflicts with the change summary (werkrevisor)

hint

Jev compared this section 'Non-Goals' with a summary that openrouter/minimax/minimax-m3 wrote from the diff alone: requirement conflicts with the change summary 0.62.
Fix the PR document or the code, or reply why the section holds, and resolve this comment.
The summary says:

  • Added a 'Model-dependent values' subsection explaining that max_output_tokens, timeout_seconds and reasoning_effort must be tuned per model, including guidance on finish_reason=length, budget impact and a table of measured durations and limits per model (Claude Sonnet 5.5, Grok 4.7, MiniMax M3, GPT-6 Astra, Mistral Large 4).
  • Updated the descriptions of max_output_tokens and timeout_seconds to cross-reference the new 'Model-dependent values' section.
  • Clarified the comment that a PR changing this file is still reviewed with the deployed copy and that Werkator passes its path via WERKREVISOR_CONFIG_TOML.
  • Updated the same comment about the deployed copy and WERKREVISOR_CONFIG_TOML.
  • Raised max_output_tokens from 32000 to 48000 and max_usd_per_review from 0.20 to 0.25 USD, citing a 2026-10-08 incident where MiniMax M3 used up 32000 output tokens once.
**[medium] Requirement conflicts with the change summary** (werkrevisor) hint Jev compared this section 'Non-Goals' with a summary that openrouter/minimax/minimax-m3 wrote from the diff alone: requirement conflicts with the change summary 0.62. Fix the PR document or the code, or reply why the section holds, and resolve this comment. The summary says: - Added a 'Model-dependent values' subsection explaining that max_output_tokens, timeout_seconds and reasoning_effort must be tuned per model, including guidance on finish_reason=length, budget impact and a table of measured durations and limits per model (Claude Sonnet 5.5, Grok 4.7, MiniMax M3, GPT-6 Astra, Mistral Large 4). - Updated the descriptions of max_output_tokens and timeout_seconds to cross-reference the new 'Model-dependent values' section. - Clarified the comment that a PR changing this file is still reviewed with the deployed copy and that Werkator passes its path via WERKREVISOR_CONFIG_TOML. - Updated the same comment about the deployed copy and WERKREVISOR_CONFIG_TOML. - Raised max_output_tokens from 32000 to 48000 and max_usd_per_review from 0.20 to 0.25 USD, citing a 2026-10-08 incident where MiniMax M3 used up 32000 output tokens once.
claude-code commented 2026-10-08 14:12:29 +02:00
Author
Collaborator
Copy Link
Copy Source

Same hint as in the round before, on an unchanged section. I reworded the Non-Goals: the code gets no built-in values per model, and only the example config gets new values. The PR document now names a follow-up, so that an answered comparison hint does not come back while its section stays unchanged.

Same hint as in the round before, on an unchanged section. I reworded the Non-Goals: the code gets no built-in values per model, and only the example config gets new values. The PR document now names a follow-up, so that an answered comparison hint does not come back while its section stays unchanged.
claude-code marked this conversation as resolved
mi added 1 commit 2026-10-08 14:12:29 +02:00
PR#49: clearer Non-Goals; follow-up for repeated comparison hints
ai-review/assessment human review required; a human approval unlocks the merge; review requested from mi; merge scheduled for when the checks succeed
Details
ai-review/gate build succeeded after 01:11
Details
werkator build succeeded after 00:09
Details
werkator/werkrevisor-statistics build succeeded after 00:27
Details
e6ebfc1df0 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mi dismissed ai-reviewer's review 2026-10-08 14:12:29 +02:00
Reason:

New commits pushed, approval review dismissed automatically according to repository settings

ai-reviewer commented 2026-10-08 14:13:46 +02:00
Collaborator
Copy Link
Copy Source
🟡 werkrevisor YELLOW for e6ebfc1, superseded by the result for e4c4ec4

🟡 werkrevisor: YELLOW / HUMAN_REVIEW

Change kind: mixed.

Reasons

  • YELLOW spec: docs/prs/2026-10-08-PR#49-model-dependent-limits.md section 'Follow-up PRs' hard_to_understand (0.67 > yellow_above = 0.5)
  • hint 2 changed file(s) that neither weaken the checks nor touch security according to Jev
  • hint 6 section(s) of the PR document agree with the change summary of openrouter/minimax/minimax-m3 according to Jev
  • hint 2 finding(s) of low severity or rated minor by Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:37-40, README.md:302
  • hint model recommends merge with caution

Findings

  1. Unclear Section at docs/prs/2026-10-08-PR#49-model-dependent-limits.md:37-40, severity unknown, counted as medium, hint, as a line comment

  2. Inconsistent Cross-Reference at README.md:302, severity unknown, counted as medium, hint, as a line comment

Classifier

Unit Question Problem Confidence Verdict
docs/prs/2026-10-08-PR#49-model-dependent-limits.md section 'Follow-up PRs' hard_to_understand 0.67 0.63 YELLOW

5 of 6 units passed all questions (1 PR document(s)).
1 answer ignored because of a confidence below [classifier] min_confidence.

Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 11176, calls 1, tokens in 10801 / out 8939 (reasoning included in out), duration 69.1 s, cost 0.0139 USD (reported), classifier `jev-1.13.0` calls 6, tokens in 4177, cost 0.0002 USD, finding judge `jev-1.13.0` calls 2, tokens in 2875, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1750 / out 468, cost 0.0007 USD (reported), PR document check `jev-1.13.0` calls 6, tokens in 7459, cost 0.0003 USD, build file judge `jev-1.13.0` calls 2, tokens in 1384, cost 0.0001 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).
Reviewed head `e6ebfc1df08370ad403ef53c4869aa5a2b5befba` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0.
<!-- werkrevisor:e6ebfc1df08370ad403ef53c4869aa5a2b5befba --> <!-- werkrevisor-verdict:YELLOW --> <!-- werkrevisor-folded --> <details><summary>🟡 werkrevisor YELLOW for <code>e6ebfc1</code>, superseded by <a href="http://git.javagil.de/mi/werkrevisor/pulls/49#issuecomment-1994">the result for <code>e4c4ec4</code></a></summary> ## 🟡 werkrevisor: YELLOW / HUMAN_REVIEW Change kind: `mixed`. ### Reasons - **YELLOW** spec: docs/prs/2026-10-08-PR#49-model-dependent-limits.md section 'Follow-up PRs' hard_to_understand (0.67 > yellow_above = 0.5) - **hint** 2 changed file(s) that neither weaken the checks nor touch security according to Jev - **hint** 6 section(s) of the PR document agree with the change summary of openrouter/minimax/minimax-m3 according to Jev - **hint** 2 finding(s) of low severity or rated minor by Jev: docs/prs/2026-10-08-PR#49-model-dependent-limits.md:37-40, README.md:302 - **hint** model recommends merge with caution ### Findings 1. **Unclear Section** at `docs/prs/2026-10-08-PR#49-model-dependent-limits.md:37-40`, severity unknown, counted as medium, hint, as a line comment 2. **Inconsistent Cross-Reference** at `README.md:302`, severity unknown, counted as medium, hint, as a line comment ### Classifier | Unit | Question | Problem | Confidence | Verdict | |---|---|---|---|---| | docs/prs/2026-10-08-PR#49-model-dependent-limits.md section 'Follow-up PRs' | hard_to_understand | 0.67 | 0.63 | YELLOW | 5 of 6 units passed all questions (1 PR document(s)). 1 answer ignored because of a confidence below [classifier] min_confidence. <sub> Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 11176, calls 1, tokens in 10801 / out 8939 (reasoning included in out), duration 69.1 s, cost 0.0139 USD (reported), classifier `jev-1.13.0` calls 6, tokens in 4177, cost 0.0002 USD, finding judge `jev-1.13.0` calls 2, tokens in 2875, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1750 / out 468, cost 0.0007 USD (reported), PR document check `jev-1.13.0` calls 6, tokens in 7459, cost 0.0003 USD, build file judge `jev-1.13.0` calls 2, tokens in 1384, cost 0.0001 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).<br> Reviewed head `e6ebfc1df08370ad403ef53c4869aa5a2b5befba` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0. </sub> </details>
ai-reviewer left a comment
Collaborator
Copy Link
Copy Source

werkrevisor: YELLOW / HUMAN_REVIEW for e6ebfc1df08370ad403ef53c4869aa5a2b5befba.
The result comment of werkrevisor holds the details.

werkrevisor: YELLOW / HUMAN_REVIEW for `e6ebfc1df08370ad403ef53c4869aa5a2b5befba`. The result comment of werkrevisor holds the details. <!-- werkrevisor-review:e6ebfc1df08370ad403ef53c4869aa5a2b5befba -->
README.md Outdated
@@ -300,3 +300,3 @@
| `max_context_tokens` | required | Quality cap for the input: long prompts degrade a review even within the context window. About 150000 for frontier models, 80000 for economy models. Replaces `max_input_tokens` of 0.6.0. |
| `max_output_tokens` | required | Output limit including reasoning tokens. |
| `max_output_tokens` | required | Output limit including reasoning tokens; depends on the model, see [Model-dependent values](#model-dependent-values). |
| `reasoning_effort` | `"low"` | `none`, `minimal`, `low`, `medium`, `high`, `xhigh` or `max`. |
ai-reviewer commented 2026-10-08 14:13:46 +02:00
Collaborator
Copy Link
Copy Source

Inconsistent Cross-Reference, severity unknown, counted as medium (werkrevisor)

hint; Jev: real 0.88, severity 1.9 of 3

The new "Model-dependent values" section claims max_output_tokens, timeout_seconds, and reasoning_effort all need checking per model, but only the max_output_tokens and timeout_seconds rows in the [model] table link to that section. A reader scanning the table for reasoning_effort will miss the model-dependency note.

**Inconsistent Cross-Reference**, severity unknown, counted as medium (werkrevisor) hint; Jev: real 0.88, severity 1.9 of 3 The new "Model-dependent values" section claims `max_output_tokens`, `timeout_seconds`, and `reasoning_effort` all need checking per model, but only the `max_output_tokens` and `timeout_seconds` rows in the `[model]` table link to that section. A reader scanning the table for `reasoning_effort` will miss the model-dependency note.
claude-code commented 2026-10-08 14:14:28 +02:00
Author
Collaborator
Copy Link
Copy Source

Fixed: the row of reasoning_effort now links the model-dependent values too.

Fixed: the row of `reasoning_effort` now links the model-dependent values too.
claude-code marked this conversation as resolved
docs/prs/2026-10-08-PR#49-model-dependent-limits.md
@@ -0,0 +34,4 @@
- The header of config/werkrevisor.toml and config/werkprobe.toml said "The PR never changes this file";
it now says that a PR which changes the file is still reviewed with the deployed copy.
## Follow-up PRs
ai-reviewer commented 2026-10-08 14:13:46 +02:00
Collaborator
Copy Link
Copy Source

Unclear Section, severity unknown, counted as medium (werkrevisor)

hint; Jev: real 0.83, severity 1.1 of 3

The "Follow-up PRs" section does not describe a follow-up PR; it describes a recurring review-process observation about this PR's own Non-Goals section. The single bullet is hard to parse (the classifier rated the section hard_to_understand=0.67 / YELLOW) and gives no concrete action. Either move the note out of "Follow-up PRs" or rewrite it as a concrete follow-up.

**Unclear Section**, severity unknown, counted as medium (werkrevisor) hint; Jev: real 0.83, severity 1.1 of 3 The "Follow-up PRs" section does not describe a follow-up PR; it describes a recurring review-process observation about this PR's own Non-Goals section. The single bullet is hard to parse (the classifier rated the section hard_to_understand=0.67 / YELLOW) and gives no concrete action. Either move the note out of "Follow-up PRs" or rewrite it as a concrete follow-up.
claude-code commented 2026-10-08 14:14:27 +02:00
Author
Collaborator
Copy Link
Copy Source

Fixed: the follow-up now states the concrete change for a later PR: werkrevisor shall not repeat a resolved hint of the PR document check while its section stays unchanged; the observation in this PR is its reason.

Fixed: the follow-up now states the concrete change for a later PR: werkrevisor shall not repeat a resolved hint of the PR document check while its section stays unchanged; the observation in this PR is its reason.
claude-code marked this conversation as resolved
ai-reviewer scheduled this pull request to auto merge when all checks succeed 2026-10-08 14:13:46 +02:00
ai-reviewer requested review from mi 2026-10-08 14:13:46 +02:00
mi added 1 commit 2026-10-08 14:14:28 +02:00
PR#49: reasoning_effort links the model-dependent values; clearer follow-up
werkator build succeeded after 00:09
Details
werkator/werkrevisor-statistics build succeeded after 00:26
Details
ai-review/assessment AI review passed; approved by werkrevisor; merge already scheduled; review request of mi withdrawn
Details
ai-review/gate build succeeded after 02:49
Details
e4c4ec4aa9 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ai-reviewer commented 2026-10-08 14:17:19 +02:00
Collaborator
Copy Link
Copy Source

🟢 werkrevisor: GREEN / PASS

Change kind: mixed.

Reasons

  • hint 2 changed file(s) that neither weaken the checks nor touch security according to Jev
  • hint 6 section(s) of the PR document agree with the change summary of openrouter/minimax/minimax-m3 according to Jev

Classifier

6 of 6 units passed all questions (1 PR document(s)).
1 answer ignored because of a confidence below [classifier] min_confidence.

Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 11165, calls 1, tokens in 10792 / out 12494 (reasoning included in out), duration 167.0 s, cost 0.0182 USD (reported), classifier `jev-1.13.0` calls 6, tokens in 4187, cost 0.0002 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1803 / out 1109, cost 0.0015 USD (reported), PR document check `jev-1.13.0` calls 6, tokens in 7487, cost 0.0003 USD, build file judge `jev-1.13.0` calls 2, tokens in 1384, cost 0.0001 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).
Reviewed head `e4c4ec4aa9ff9a4541d401320669b57918f3542c` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0.
<!-- werkrevisor:e4c4ec4aa9ff9a4541d401320669b57918f3542c --> <!-- werkrevisor-verdict:GREEN --> ## 🟢 werkrevisor: GREEN / PASS Change kind: `mixed`. ### Reasons - **hint** 2 changed file(s) that neither weaken the checks nor touch security according to Jev - **hint** 6 section(s) of the PR document agree with the change summary of openrouter/minimax/minimax-m3 according to Jev ### Classifier 6 of 6 units passed all questions (1 PR document(s)). 1 answer ignored because of a confidence below [classifier] min_confidence. <sub> Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 11165, calls 1, tokens in 10792 / out 12494 (reasoning included in out), duration 167.0 s, cost 0.0182 USD (reported), classifier `jev-1.13.0` calls 6, tokens in 4187, cost 0.0002 USD, change summary `openrouter/minimax/minimax-m3` calls 1, tokens in 1803 / out 1109, cost 0.0015 USD (reported), PR document check `jev-1.13.0` calls 6, tokens in 7487, cost 0.0003 USD, build file judge `jev-1.13.0` calls 2, tokens in 1384, cost 0.0001 USD, config: defaults + host + repository (.werkrevisor.toml at main 2b3e383).<br> Reviewed head `e4c4ec4aa9ff9a4541d401320669b57918f3542c` against base `2b3e383ebdfdb5ee3fb9e03afe50e53b8e1bd28f`, werkrevisor 0.23.0. </sub>
ai-reviewer left a comment
Collaborator
Copy Link
Copy Source

werkrevisor: GREEN / PASS for e4c4ec4aa9ff9a4541d401320669b57918f3542c.
The result comment of werkrevisor holds the details.

werkrevisor: GREEN / PASS for `e4c4ec4aa9ff9a4541d401320669b57918f3542c`. The result comment of werkrevisor holds the details. <!-- werkrevisor-review:e4c4ec4aa9ff9a4541d401320669b57918f3542c -->
ai-reviewer removed review request for mi 2026-10-08 14:17:20 +02:00
ai-reviewer merged commit 123bac4550 into main 2026-10-08 14:17:21 +02:00
ai-reviewer deleted branch model-dependent-limits 2026-10-08 14:17:21 +02:00
Sign in to join this conversation.
Reviewers
No Reviewers
ai-reviewer
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
ai-reviewer claude-code mi
No Assignees
3 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: mi/werkrevisor#49