BMAD-METHOD/src/bmm/workflows/4-implementation/genai-knowledge-sync/knowledge-index-template.md

87 lines
2.1 KiB
Markdown

---
project_name: ''
user_name: ''
date: ''
total_chunks: 0
sources_indexed: 0
tag_vocabulary_size: 0
retrieval_tested: false
status: 'draft'
---
# Knowledge Index for {{project_name}}
_RAG-optimized knowledge base for AI agent retrieval. Each chunk is self-contained and tagged for semantic search._
---
## Index Summary
- **Total Chunks:** {{total_count}}
- **Critical:** {{critical_count}} | **High:** {{high_count}} | **Standard:** {{standard_count}} | **Reference:** {{ref_count}}
- **Sources Indexed:** {{source_count}}
- **Last Synced:** {{date}}
---
## Critical Knowledge
<!-- Critical-priority chunks go here. These are retrieved for every implementation task. -->
---
## Architecture Knowledge
<!-- Architecture decisions, system design patterns, and technology choices. -->
---
## Requirements Knowledge
<!-- Business rules, acceptance criteria, and constraints. -->
---
## Implementation Knowledge
<!-- Coding patterns, conventions, and implementation rules. -->
---
## Domain Knowledge
<!-- Business domain concepts, terminology, and definitions. -->
---
## Operations Knowledge
<!-- Deployment, monitoring, and workflow rules. -->
---
## Quality Knowledge
<!-- Testing patterns, review standards, and anti-patterns. -->
---
## Retrieval Configuration
### Query Mapping
| Query Pattern | Target Categories | Priority Filter | Expected Chunks |
|---|---|---|---|
| "how to implement \*" | implementation, architecture | critical, high | 3-5 |
| "testing requirements for \*" | quality, implementation | critical, high | 2-4 |
| "business rules for \*" | requirements, domain | all | 2-3 |
| "architecture decision for \*" | architecture | all | 1-3 |
| "deployment process for \*" | operations | all | 1-2 |
### Embedding Recommendations
- **Model:** Use an embedding model that handles technical content well
- **Chunk Overlap:** 50-100 characters overlap between adjacent chunks from the same source
- **Metadata Filters:** Always filter by category and priority for focused retrieval
- **Top-K:** Retrieve 3-5 chunks per query for optimal context balance