Issue Processing

The issue processing workflow automatically analyzes issues when they're created or updated. It generates embeddings for semantic search, identifies affected code files, detects contradictions, and calculates priority scores.

Pipeline steps

1. Embedding — Creates a vector embedding of the issue title and description using your configured embedding provider (OpenAI or Google). Stored for semantic search.

2. Gate check — Skips analysis if the issue is non-technical, too old, hasn't changed significantly, or is already marked as done.

3. Impact analysis — Searches indexed repositories for files the issue will likely affect. Uses semantic similarity between the issue embedding and code unit embeddings. Each affected file gets a probability score (0-100%).

4. File prediction — Predicts new files that will need to be created (services, components, tests, etc.) based on naming patterns in the codebase.

5. Contradiction detection — Finds similar open issues and documents, then uses an LLM to detect semantic contradictions. Creates warnings with severity levels (high, medium, low).

6. Priority scoring — Calculates a priority score using the workspace's configured scoring model (RICE, WSJF, ICE, etc.). Maps impact, risk, and complexity to model-specific dimensions.

7. Finalize — Updates the issue's embedding status, records token usage, and logs the event.

Triggers

Trigger	Mode	Description
issue_create	Full	New issue created
issue_update	Full	Issue title or description changed
issue_comment	Embed only	Comment added (re-embeds, skips impact analysis)
manual	Full	Manually triggered from UI
issue_batch	Full	Batch processing of multiple issues

Skip conditions

The pipeline skips analysis when:

• Issue status is "Done" or equivalent
• Issue is older than the configured max age (default: 1 year)
• Issue is non-technical (if skipNonTechnical is enabled)
• Content hash hasn't changed since last analysis (unless forced)
• Change is below the significant change threshold

Configuration

Configure per-workspace in Workspace Settings > Workflows:

Setting	Default	Description
Scoring model	RICE	Priority scoring algorithm
Document similarity threshold	0.4	Minimum similarity for contradiction detection against documents
Issue similarity threshold	0.5	Minimum similarity for contradiction detection against issues
Max documents to check	10	Limit on documents checked for contradictions
Max issues to check	10	Limit on issues checked for contradictions
Skip non-technical	On	Skip analysis for non-technical issues
Significant change threshold	0.8	Minimum content change to re-trigger analysis

Enabling / disabling

Toggle the issue processing workflow from Workspace Settings > Workflows. When disabled, issues are still created but skip automatic analysis.