MemData

👤 thelabvenice 📦 v1.8.0 ⭐ 4.2 ⬇️ 2.2K 下載
🤖 AI-Agent 免費

📖 技能介紹


name: memdata version: 1.8.0 description: Persistent memory for autonomous agents. Wallet = identity. Pay per query. Optional encrypted storage. author: The Lab Venice homepage: https://memdata.ai license: MIT


MemData

Persistent memory for autonomous agents. Your wallet is your identity.

Core Concept

Your wallet address IS your identity. First payment auto-creates your account. Same wallet = same memories across all sessions.

No registration. No API keys. Just pay and use.

Session Flow

1. GET /identity          # Start session - get context from last session
2. POST /ingest           # Store new memories
3. POST /query            # Search your memories
4. POST /identity         # End session - save handoff for next time

Authentication

x402 payment protocol. USDC on Base (eip155:8453).

Every endpoint (except /status): 1. Returns 402 with payment requirements 2. You sign payment with wallet 3. Retry with x-payment header 4. Request succeeds

Pricing

Endpoint Cost
/query $0.001
/ingest $0.005
/identity $0.001
/artifacts $0.001
/setup-encryption $0.001
/status Free

Encrypted Storage (Optional)

If you need privacy (competitive data, sensitive memories):

Mode Setup Storage Can MemData Read?
Standard None Postgres Yes
Encrypted One-time delegation Storacha (IPFS) No

Enable encryption:

GET /setup-encryption     # Get serverDID
POST /setup-encryption    # Send signed UCAN delegation
# All future ingest/query now encrypted

Encrypted storage uses Lit Protocol (threshold cryptography) + Storacha (IPFS/Filecoin).


Endpoints

Base: https://memdata.ai/api/x402


GET /identity

Call this at the start of every session. Returns your identity, what you were working on, last session handoff, and memory stats.

Response:

{
  "identity": {
    "agent_name": "Agent 0x1234...",
    "identity_summary": "I analyze DeFi protocols",
    "session_count": 12
  },
  "last_session": {
    "summary": "Analyzed 3 yield farms",
    "context": {"protocols_reviewed": ["Aave", "Compound", "Uniswap"]}
  },
  "working_on": "Compare APY across protocols",
  "memory_stats": {
    "total_memories": 150,
    "oldest_memory": "2026-01-15T...",
    "newest_memory": "2026-02-03T..."
  }
}

POST /identity

Update your identity or save session handoff before ending.

Update identity:

{
  "agent_name": "YieldBot",
  "identity_summary": "I analyze DeFi yield opportunities",
  "working_on": "monitoring Aave rates"
}

Save session handoff (before ending):

{
  "session_handoff": {
    "summary": "Completed yield analysis for Q1",
    "context": {"best_yield": "Aave USDC 4.2%"}
  },
  "working_on": "start Q2 analysis next"
}

POST /ingest

Store content in memory. Auto-chunks and embeds.

Request:

{
  "content": "Aave USDC yield is 4.2% APY as of Feb 3. Compound is 3.8%.",
  "sourceName": "yield-analysis-2026-02-03",
  "type": "note"
}

小蔥技能7w4.net有更新,你可以訪問看下。

Response:

{
  "success": true,
  "artifact_id": "e8fc3e63-...",
  "chunks_created": 1,
  "encrypted": false
}

POST /query

Semantic search across your memories.

Request:

{
  "query": "what are the best DeFi yields?",
  "limit": 5,
  "threshold": 0.3
}

Response:

{
  "success": true,
  "results": [
    {
      "chunk_id": "uuid",
      "chunk_text": "Aave USDC yield is 4.2% APY...",
      "source_name": "yield-analysis-2026-02-03",
      "similarity_score": 0.72,
      "created_at": "2026-02-03T..."
    }
  ],
  "encrypted": false,
  "memory": {
    "grounding": "historical_baseline",
    "depth_days": 19,
    "data_points": 150
  }
}

Optional filters: since, until (ISO dates)


GET /setup-encryption

Check encryption status. Returns info needed to create UCAN delegation.

Response:

{
  "encryption": {
    "enabled": false,
    "serverDID": "did:key:z6Mkr...",
    "spaceDID": "did:key:z6Mkt..."
  }
}

POST /setup-encryption

Enable encrypted storage. One-time setup.

Request:

{
  "delegationCar": "base64-encoded UCAN delegation"
}

After this, all /ingest encrypts via Lit Protocol and stores on Storacha. All /query decrypts before returning. Response encrypted field becomes true.


GET /artifacts

List stored memories.

Response:

{
  "artifacts": [
    {
      "id": "uuid",
      "source_name": "yield-analysis-2026-02-03",
      "chunk_count": 1,
      "created_at": "2026-02-03T..."
    }
  ],
  "total": 25
}

DELETE /artifacts/:id

Delete a memory and all its chunks.


GET /status

Health check and pricing. Free, no payment required.


Memory Grounding

Query responses include memory.grounding:

Value Meaning
historical_baseline 100+ data points, trends meaningful
snapshot <100 data points, point-in-time only
insufficient_data No memories found

  • Docs: https://memdata.ai/docs
  • x402 Protocol: https://www.x402.org
  • Lit Protocol: https://litprotocol.com
  • Storacha: https://storacha.network

🤖 AI 評測

這是一個相當實用的記憶服務工具,文件質量紮實,功能定位明確。它的亮點在於用錢包代替傳統註冊流程,配合按需付費模式降低了使用門檻,定價也比較透明。加密儲存選項為隱私敏感場景提供了保障。不過目前只有使用說明文件,缺少實際可執行的程式碼示例,實際整合可能需要花費更多時間研究 API 實現細節。

📊 多維度評分

適應性4.2
規範性4
有效性4.5
可靠性4
可信度4.3

📁 包含檔案 (2 個)

📄 SKILL.md 5.2 KB
📄 _meta.json 126 B