Search hybrid BM25 and vector indexes for relevant past sessions and notes.
npx skills add https://github.com/CarbeneAI/Forge --skill semanticmemory
Please help me install this Agent Skill. Skill: SemanticMemory Source: https://github.com/CarbeneAI/Forge/tree/main/.claude/skills/SemanticMemory Command: npx skills add https://github.com/CarbeneAI/Forge --skill semanticmemory
Help users find, recall, and reuse past AI sessions, learnings, research, and personal notes so context and decisions are not lost between conversations or projects.
Run a semantic search for past work by invoking the SemanticSearch tool with your query, for example: bun ~/.claude/skills/SemanticMemory/tools/SemanticSearch.ts "How did we configure Traefik SSL?"
High-intent search queries and answers about installing and using this skill.
To search past AI sessions, you can invoke a semantic search tool that uses hybrid BM25 and vector similarity to retrieve relevant past work, learnings, and research from your history. This surfaces context from previous conversations without losing decisions between projects.
Hybrid BM25 and vector indexing combines full-text search with embedding similarity to retrieve relevant past sessions and notes. This approach ensures both exact keyword matches and semantic meaning are used to surface the right context at the right time.
Yes, local SQLite is supported for semantic memory using the vec0 and FTS5 extensions. This allows you to run hybrid semantic search and store embeddings directly in a local database without requiring external infrastructure.
Discovering cross-project connections is best achieved using temporal knowledge graph tunnels. These tunnels identify topic bridges across your sessions, learnings, and research notes to surface relationships between different workflows.
You can use either Gemini or OpenAI as your embedding provider to build a temporal knowledge graph, with fallback support included. These embeddings power the vector similarity needed to map time-bounded facts and entity relations.
You minimize token usage by using a layered context loading approach that progresses from Identity to Essential, On-Demand, and Deep context. This tiered loading ensures only the most relevant information is surfaced during conversations and automation.
statesetSet up the StateSet iCommerce engine with CLI installation, DB initialization, and demo data seeding.
jannolan200211-ship-itConstruct and evolve a typed knowledge graph with enforced constraints.
winsorllcProvides semantic search across documents with Tunrsity and Tunrsity-like features.
lanejSearch and audit personal knowledge bases with hybrid vector and BM25 indexing.
