knowledge-graph-context
Project context from ChromaDB knowledge graph. Triggers on "how does this project work", "architecture", "where is X", "why was Y chosen", "project structure", "component relationships", "design decisions".
When & Why to Use This Skill
This Claude skill leverages a ChromaDB-powered knowledge graph to provide deep, structural project context. It enables developers to instantly understand complex system architectures, component relationships, and the underlying rationale behind design decisions, significantly reducing the cognitive load required to navigate and modify large codebases.
Use Cases
- Architectural Onboarding: Rapidly grasp the high-level system structure and design patterns of a new repository by querying its architecture and component relationships.
- Impact Analysis: Identify dependencies and interaction patterns between different modules before implementing major code changes or refactors.
- Design Rationale Retrieval: Access historical design decisions and the 'why' behind specific technical choices to ensure consistency with the project's original intent.
- Context-Aware Navigation: Quickly locate where specific functionalities are implemented and understand how they integrate into the broader project workflow.
| name | Knowledge Graph Context |
|---|---|
| description | | |
| version | 2.0.0 |
Knowledge Graph Context Skill
Access ChromaDB knowledge graph via coconut-knowledge-mcp.
When to Use
- Understand project architecture before changes
- Find where functionality is implemented
- Understand design decisions and rationale
- Learn component relationships
- Discover integration patterns
- Get project context
Get Collection Name
CRITICAL: Collection = {author}_{repo} from git remote.
git remote get-url origin 2>/dev/null | sed -E 's|.*[:/]([^/]+)/([^/.]+)(\.git)?$|\1_\2|' | tr '[:upper:]' '[:lower:]' | tr '-' '_'
Quick Reference
Search Knowledge
chroma_query_documents(
collection_name="<collection>",
query_texts=["your question"],
n_results=5
)
Filter by Type
chroma_query_documents(
collection_name="<collection>",
query_texts=["your question"],
n_results=5,
where={"type": "architecture"}
)
Get Document Count
chroma_get_collection_count(collection_name="<collection>")
Get Specific Documents
chroma_get_documents(
collection_name="<collection>",
ids=["component-auth-service"]
)
Node Types
| Type | Description |
|---|---|
| architecture | System structure, design |
| component | Modules, services |
| workflow | Processes, data flows |
| integration | APIs, databases |
| convention | Patterns, naming |
| decision | Choices with rationale |
Search Strategies
Architecture: where={"type": "architecture"}
Components: where={"type": "component"}
Decisions: where={"type": "decision"}
Workflow
- Get collection name from git remote
- Check collection exists:
chroma_list_collections - Search for context:
chroma_query_documents - If stale: suggest
/coconut-knowledge:update
Add New Knowledge
Use /coconut-knowledge:add or:
chroma_add_documents(
collection_name="<collection>",
documents=["content"],
ids=["type-name"],
metadatas=[{"name": "...", "type": "...", "tags": "..."}]
)
Troubleshooting
- No results: Check collection exists, try broader terms
- Stale info: Run
/coconut-knowledge:update - Not initialized: Run
/coconut-knowledge:init