memories
Save and retrieve memories or embeddings via the repo helpers or API. Use when working with embedding config or memory storage.
When & Why to Use This Skill
This Claude skill provides a robust framework for managing agentic memory through the storage and retrieval of vector embeddings. It enables AI agents to maintain long-term context, store repository-specific facts, and perform semantic searches, significantly enhancing the agent's ability to recall information and maintain continuity across different sessions and complex tasks.
Use Cases
- Case 1: Repository Knowledge Enrichment - Save specific facts and architectural details about a codebase (e.g., 'Bumba repo facts') as embeddings to provide the agent with deep, persistent context during development and maintenance tasks.
- Case 2: Semantic Information Retrieval - Utilize the retrieve-memory function to find relevant project history, technical documentation, or previous workflow states based on the semantic meaning of a query rather than simple keyword matching.
- Case 3: Persistent Context for AI Agents - Enable AI agents to 'remember' user preferences, complex configuration details, or historical decision-making data by storing them in a dedicated memory storage system for future retrieval.
| name | memories |
|---|---|
| description | Save and retrieve memories or embeddings via the repo helpers or API. Use when working with embedding config or memory storage. |
Memories
Overview
Use the memories helpers to store and retrieve embeddings in a consistent way. Ensure embedding dimension matches the DB schema.
Save
bun run --filter memories save-memory --task-name bumba-enrich-repo --content "Enrichment facts for services/bumba" --summary "Bumba repo facts" --tags "bumba,enrich"
Retrieve
bun run --filter memories retrieve-memory --query "enrichFile workflow" --limit 5
Environment
OPENAI_API_BASE_URL/OPENAI_API_BASEOPENAI_EMBEDDING_MODELOPENAI_EMBEDDING_DIMENSIONOPENAI_EMBEDDING_TIMEOUT_MS
Resources
- Reference:
references/memories-runbook.md - Helper:
scripts/memories.sh - Example payload:
assets/memory-entry.json