Store decision traces and query past precedents via semantic search.
npx skills add https://github.com/ingpoc/SKILLS --skill context-graph
Please help me install this Agent Skill. Skill: context-graph Source: https://github.com/ingpoc/SKILLS/tree/main/context-graph Command: npx skills add https://github.com/ingpoc/SKILLS --skill context-graph
Context-Graph provides a living record of decisions and their outcomes, enabling teams to recall past choices, justify actions, and learn from history.
Store a decision trace with context and immediately query for similar precedents to validate the decision.
High-intent search queries and answers about installing and using this skill.
You can store decision traces with timestamps and metadata like category and session state. This builds a searchable memory for precedent retrieval, enabling semantic search across past decisions to validate current choices.
Semantic search across past precedents uses Voyage AI vector embeddings to find similar historical decisions by meaning rather than exact keywords. This enables teams to recall past choices and justify actions based on contextual similarity.
Yes, Voyage AI and ChromaDB are required dependencies. Voyage AI generates vector embeddings for semantic search, while ChromaDB serves as the vector store for querying stored decision traces and past precedents.
Yes, pattern extraction surfaces recurring decision patterns to support reflexion-style improvements. This enables learning loops across multiple projects by analyzing historical decision traces stored with metadata.
The best way to audit cross-project decisions is storing traces with rich metadata like feature_id and state in a vector store. ChromaDB then enables semantic retrieval of past precedents for retrospective debugging and decision auditing.
Semantic search finds precedents by meaning rather than exact keywords, retrieving contextually similar past decisions even when terminology differs. This prevents missing relevant historical outcomes during debugging or learning loops.
giuseppe-trisciuoglioConfigure LangChain4J vector stores for embedding storage and retrieval in Java applications.
shynlee04Search the web and extract content via Tavily and Exa using the inference.sh CLI.
omega-memoryStore persistent memory across coding sessions with semantic search.
lifangdaSearch archived Claude Code conversations semantically and by exact text with filters.
