smara-memory（斯马拉记忆）

Smara Memory Skill

Gives your agent persistent memory across conversations. Store facts about users, search by meaning, and retrieve full context — powered by Smara's Ebbinghaus decay scoring.

When to use

• When the agent learns something about a user that should persist (preferences, facts, context)
• When the agent needs to recall what it knows about a user
• When the agent should check if it already knows something before asking again
• After meaningful conversations to extract and store key facts

Setup

1. Get a free API key at https://smara.io
2. Set SMARA_API_KEY in your environment

Actions

Store a memory

curl -X POST https://api.smara.io/v1/memories \
  -H "Authorization: Bearer $SMARA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "user_id_here",
    "fact": "User prefers dark mode and uses vim keybindings",
    "importance": 0.7
  }'

Search memories

curl "https://api.smara.io/v1/memories/search?user_id=user_id_here&query=editor+preferences&limit=5" \
  -H "Authorization: Bearer $SMARA_API_KEY"

Get full user context

curl "https://api.smara.io/v1/users/user_id_here/context" \
  -H "Authorization: Bearer $SMARA_API_KEY"

Delete a memory

curl -X DELETE "https://api.smara.io/v1/memories/MEMORY_ID" \
  -H "Authorization: Bearer $SMARA_API_KEY"

Instructions for the agent

1. After conversations: Extract key facts (preferences, decisions, context) and store them as memories with relevant tags
2. Before responding: Search for relevant memories to personalize responses
3. Contradiction handling: Smara automatically handles contradictions — if a user changes a preference, just store the new one and the old one is soft-deleted
4. Duplicate handling: Smara skips duplicates automatically — safe to store the same fact multiple times
5. Decay scoring: Memories naturally lose weight over time. Recent, frequently-accessed memories rank higher. This is automatic.

Example workflow

User: "I switched to Neovim last week"

Agent thinks:
1. Search memories for "editor preferences" → finds "Uses vim keybindings"
2. Store new memory: "Switched to Neovim (from vim)" with tags ["preferences", "editor"]
3. Smara auto-detects contradiction with old vim memory → soft-deletes it
4. Respond acknowledging the switch

API Reference

Full docs: https://api.smara.io/docs/