recall
npx machina-cli add skill parcadei/Continuous-Claude-v3/recall --openclawRecall - Semantic Memory Retrieval
Query the memory system for relevant learnings from past sessions.
Usage
/recall <query>
Examples
/recall hook development patterns
/recall wizard installation
/recall TypeScript errors
What It Does
- Runs semantic search against stored learnings (PostgreSQL + BGE embeddings)
- Returns top 5 results with full content
- Shows learning type, confidence, and session context
Execution
When this skill is invoked, run:
cd $CLAUDE_OPC_DIR && PYTHONPATH=. uv run python scripts/core/recall_learnings.py --query "<ARGS>" --k 5
Where <ARGS> is the query provided by the user.
Output Format
Present results as:
## Memory Recall: "<query>"
### 1. [TYPE] (confidence: high, id: abc123)
<full content>
### 2. [TYPE] (confidence: medium, id: def456)
<full content>
Options
The user can specify options after the query:
--k N- Return N results (default: 5)--vector-only- Use pure vector search (higher precision)--text-only- Use text search only (faster)
Example: /recall hook patterns --k 10 --vector-only
Source
git clone https://github.com/parcadei/Continuous-Claude-v3/blob/main/.claude/skills/recall/SKILL.mdView on GitHub Overview
Recall taps the memory system to fetch relevant learnings from prior sessions. It performs a semantic search over stored learnings (PostgreSQL + BGE embeddings) and returns the top 5 results with full content, learning type, confidence, and session context.
How This Skill Works
When invoked, Recall runs a semantic search against stored learnings using embeddings, retrieving the top 5 results. Results are presented with learning type, confidence, and session context, formatted for quick review. The process is executed via recall_learnings.py with the provided query and --k parameter, enabling options like --vector-only or --text-only.
When to Use It
- You need context from past sessions to inform a current problem or decision
- You want to review recurring patterns or solutions (e.g., hook development patterns)
- You need installation steps or troubleshooting notes surfaced from prior work (e.g., wizard installation)
- You want to surface TypeScript errors and corresponding learnings from history
- You want to compare current approach against documented learnings to avoid past mistakes
Quick Start
- Step 1: Issue a query, e.g., /recall <query> [--k N] [--vector-only|--text-only]
- Step 2: The system runs recall_learnings.py and returns the top results with content and context
- Step 3: Review the results and apply relevant learnings to your current task
Best Practices
- Craft specific queries to narrow results (e.g., include a topic plus a constraint)
- Adjust the number of results with --k N to balance completeness and noise
- Use --vector-only for higher precision when available
- Always verify confidence and review session context before acting
- Treat recall results as reference material; corroborate with primary sources when possible
Example Use Cases
- Recall hook development patterns
- Recall wizard installation
- Recall TypeScript errors
- Recall API usage patterns from prior sessions
- Recall performance tuning learnings
