Doc Generator
A sophisticated n8n workflow that transforms minimal input into publication-ready technical documentation through a multi-stage LLM feedback loop.Overview
The Doc Generator emulates the quality standards and technical finesse of senior engineers who have worked across complex systems at Tesla, Airbus, GCHQ, and NASA.Key Capabilities
- Minimal Input: Transform brief requests into comprehensive docs
- Iterative Refinement: Draft-review cycles until quality threshold
- Multi-Output: GitHub PR, Slack notification, Airtable logging
- Consistent Quality: 85%+ quality score requirement
Pipeline Stages
| Stage | Purpose | Model | Temperature |
|---|---|---|---|
| Conceptual Design | Document architecture planning | Claude Sonnet | 0.3 |
| Draft Generator | Content creation | Claude Sonnet | 0.5 |
| Quality Review | Scoring and feedback | Claude Sonnet | 0.2 |
| Finalization | Frontmatter and formatting | Claude Sonnet | 0.1 |
Quality Dimensions
Each draft is scored across five dimensions:| Dimension | Weight | Description |
|---|---|---|
| Technical Accuracy | 25% | Correctness of information |
| Clarity & Structure | 25% | Readability and organization |
| Completeness | 20% | Coverage of topic |
| Audience Fit | 15% | Appropriate for readers |
| Actionability | 15% | Practical usefulness |
Output Destinations
- GitHub: Automatic PR creation with the document
- Slack: Notification with summary and link
- Airtable: Execution log for tracking
Performance
- Quality Threshold: 85%
- Max Iterations: 3 (configurable)
- Typical Duration: 2-5 minutes
- Token Efficiency: Optimized prompt design