> ## Documentation Index
> Fetch the complete documentation index at: https://simili.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Custom Workflows

> Create custom pipeline workflows

# Custom Workflows

Create custom workflows by selecting which pipeline steps to run.

## Pipeline steps

Simili Bot has 13 modular steps you can mix and match:

1. **gatekeeper** - Check if repo is enabled
2. **command\_handler** - Process bot commands
3. **vectordb\_prep** - Create collections
4. **similarity\_search** - Find related issues
5. **transfer\_check** - Rule-based routing
6. **llm\_router** - AI routing
7. **duplicate\_detector** - Identify duplicates
8. **quality\_checker** - Assess quality
9. **triage** - Suggest labels
10. **response\_builder** - Build comment
11. **action\_executor** - Post to GitHub
12. **indexer** - Add to vector DB
13. **pending\_action\_scheduler** - Schedule actions

## Preset workflows

Use built-in presets:

```bash theme={null}
simili process --workflow issue-triage    # Full pipeline
simili process --workflow similarity-only # Search only
simili process --workflow index-only      # Indexing only
```

Or in GitHub Actions:

```yaml theme={null}
- uses: similigh/simili-bot@v0.2.0
  with:
    workflow: "issue-triage"
```

## Workflow presets

### issue-triage (Default)

Complete issue analysis pipeline:

* Find similar issues
* Check for duplicates
* Assess quality
* Suggest labels
* Route if configured

### similarity-only

Semantic search only:

* Find related issues
* No AI analysis
* Fast and lightweight

### index-only

Indexing to vector database:

* Add issues to Qdrant
* No analysis or comments

## Custom workflows

Define custom workflows in configuration (future feature):

```yaml theme={null}
# Planned: Custom workflow support
workflows:
  fast-only:
    steps:
      - gatekeeper
      - vectordb_prep
      - similarity_search
      - response_builder
      - action_executor

  full-analysis:
    steps:
      - gatekeeper
      - command_handler
      - vectordb_prep
      - similarity_search
      - transfer_check
      - llm_router
      - duplicate_detector
      - quality_checker
      - triage
      - response_builder
      - action_executor
      - indexer
```

## Tips for efficiency

**Reduce API Calls:**

* Use `similarity-only` when you don't need AI analysis
* Disable unnecessary steps

**Improve Speed:**

* Skip steps you don't need
* Use dry-run for testing

**Optimize Costs:**

* Every Gemini API call has cost
* Use lightweight workflows for high-volume repos

## Next steps

<CardGroup cols={1}>
  <Card title="Architecture overview" href="/overview/architecture">
    Understand pipeline architecture
  </Card>
</CardGroup>