mi/werkrevisor
1
0
Fork 0
Code Issues Pull Requests 2 Actions Packages Projects Releases Wiki Activity

Concept: classifier Jev combined with the LLM review #1

Manually merged
mi merged 0 commits from mihoe/jev-classifier-concept into main 2026-10-05 14:24:40 +02:00
Conversation 0 Commits - Files Changed -
mi commented 2026-10-05 07:58:01 +02:00
Owner
Copy Link
Copy Source

Konzept: der Classifier Jev (TypeSafe) und der LLM-Review in vier Stufen kombiniert.
Viele billige Jev-Aufrufe je Datei, je Abschnitt des PR-Dokuments und je Gherkin-Szenario; danach ein LLM-Review, der den Jev-Ergebnissen folgt.
Noch keine Umsetzung; Voraussetzung ist ein Jev-Zugang.

Dokument: docs/konzept-jev-classifier.md.

🤖 Generated with Claude Code

Konzept: der Classifier Jev (TypeSafe) und der LLM-Review in vier Stufen kombiniert. Viele billige Jev-Aufrufe je Datei, je Abschnitt des PR-Dokuments und je Gherkin-Szenario; danach ein LLM-Review, der den Jev-Ergebnissen folgt. Noch keine Umsetzung; Voraussetzung ist ein Jev-Zugang. Dokument: `docs/konzept-jev-classifier.md`. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
mi added 1 commit 2026-10-05 07:58:01 +02:00
docs: concept for combining the classifier Jev with the LLM review aac6fe8fe5 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mi added 1 commit 2026-10-05 08:25:12 +02:00
docs: Jev rates the risk of every file change by its kind; mechanical changes skip the LLM budget e11d6206b7 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mi added 3 commits 2026-10-05 11:19:20 +02:00
Merge branch 'main' into mihoe/jev-classifier-concept b752759dbe
werkrevisor 0.5.0: Jev judges the sections and Gherkin scenarios of the PR documents 04461a274a 
[engine] mode chooses the review: "llm" (default, unchanged), "classifier" (Jev alone, no LLM)
or "combined" (Jev, then the LLM with Jev's flagged units as hints in the user prompt).
werkrevisor splits each changed PR document into ##/### sections and #### Scenario units,
asks Jev once per unit with the operator's questions, and maps every answer to a problem value:
each question asks for a problem, a score runs from "no problem" to "worst problem".
A clear spec problem is RED, a doubtful one or an oversized unit YELLOW, a Jev error ERROR.
The hints carry file, line, scenario id, question and value, but no PR text.
tools/jev-eval measures the questions offline on local PR documents.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
docs: classifier stage 1 is implemented; TYPESAFE_API_KEY and [engine] in the operations guides
werkator/server build succeeded after 00:00
Details
werkator build succeeded after 00:06
Details
werkator/werkrevisor-statistics build succeeded after 00:00
Details
5b53d48bca 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mi marked the pull request as ready for review 2026-10-05 14:22:58 +02:00
mi manually merged commit 5b53d48bca into main 2026-10-05 14:24:40 +02:00
Sign in to join this conversation.
Reviewers
No Reviewers
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
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: mi/werkrevisor#1