Load PR object
Classify changed files by quality impact before reasoning.
Context TriageQuality Engineering Demo
Automated Code Review Agent
A QE review agent that triages PR files, attaches team standards, routes review depth, and emits gated comments.
Structure Blueprint
Clickable pattern simulator
Interactive Browser Demo
This simulator does not replace the CLI run, but it makes each step, state, log, gate and final output visible in the browser.
Live pipeline
State inspector
Execution log
Current output
Step-by-step Design
Attach standards
Pin team coding rules and project conventions.
Hierarchical RetentionRoute depth
Fast path for lint-like issues, deeper path for risk-bearing changes.
Complexity RoutingGenerate comments
Emit concise review comments with file, severity and rationale.
Prompt ChainingApply gate
Block only when quality or release risk crosses the threshold.
Approval GateExpose trace
Show file route, rule source, score and gate reason.
Observability HarnessScreenshots
Pattern Mapping
| Pattern | Design role |
|---|---|
| Context Triage | Large PRs need file priority before review judgment. |
| Hierarchical Retention | Team standards should be stable context, not repeated prompt text. |
| Complexity Routing | Different review questions deserve different model depth. |
| Approval Gate | The system comments freely but blocks cautiously. |
Design Insights
Attention is the first quality control
Most bad code review agents fail before reasoning, because the wrong files enter the context.
Memory turns style into policy
A team style guide should behave like pinned memory, otherwise every PR starts from zero.
The gate defines autonomy
A review agent can be useful without owning merge authority.
Run Path
cd workshops/workshop1-code-review
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python src/main.py
open visual/index.htmlworkshops/workshop1-code-review/README.mdimages/training/visual/workshop1-code-review-blueprint.png