recall

name	recall
description	Search Awareness memory for past implementations, decisions, or relevant context.
user-invocable	true
disable-model-invocation	false

Search Awareness Memory for relevant context.

Query: $ARGUMENTS

How to call Awareness tools

Try MCP tools first (awareness_recall). If MCP tools are NOT available, use Bash to call the local daemon HTTP API directly:

# F-053: single-parameter — daemon picks scope/mode/detail/weights.
curl -s -X POST http://localhost:37800/mcp -H "Content-Type: application/json" -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"awareness_recall","arguments":{"query":"why did we pick pgvector?","limit":10}}}'

The response is JSON-RPC: result.content[0].text contains the tool output as JSON string.

Steps

1. REWRITE the user query into a complete natural-language question with context. Example: "auth bug" → "authentication bug in login flow, JWT token handling, session management"

2. Call awareness_recall with ONE parameter:

   • query: the rewritten natural-language question
   • (optional) limit: default 6, max 30
   • (optional) token_budget: 5K (default, card-heavy) / 30K (mixed) / 60K+ (raw-heavy)

   Daemon auto-routes across memories + knowledge cards + workspace graph and picks the right detail level for your token budget. You do NOT need to choose scope, recall_mode, detail, ids, or weights.

3. Present results clearly:

   • Existing implementations that can be reused (include file paths)
   • Architectural decisions already made
   • Related past work and warnings

Rules:

• Pass ONE query string — daemon handles the rest.
• If results are empty, say so clearly — do not hallucinate.
• Do not dump raw JSON — summarize in plain language.