TDaí Memory Suite for Openclaw

1. Installation

Install the complete package directly using the package manager for Openclaw Skills:

clawhub install tdai-memory-suite

Or deploy individual components depending on your pipeline requirements:

# Core memory pipeline (required)
clawhub install tdai-core

# Ontology knowledge graph (optional)
clawhub install tdai-ontology

# Atlas visualization (optional)
clawhub install tdai-atlas

2. Configuration

To activate the suite, update your system-wide openclaw.json configuration to configure local plugins and set the embedding provider to local search:

{
  "plugins": {
    "allow": ["mx", "memory-tencentdb"],
    "load": {
      "paths": ["<path-to-tdai-core>"]
    },
    "entries": {
      "memory-tencentdb": {
        "enabled": true,
        "config": {
          "pipeline": {
            "everyNConversations": 3,
            "enableWarmup": true,
            "l1IdleTimeoutSeconds": 10
          },
          "store": {
            "backend": "sqlite"
          }
        }
      }
    }
  },
  "memorySearch": {
    "provider": "local",
    "model": "nomic-embed-text-v1.5.Q4_K_M.gguf"
  }
}

3. Ontology Auto-Sync (Optional Cron Job)

Add an automated daily cron-like action inside your workflow configs to keep your entity maps updated:

{
  "name": "Ontology Daily Sync",
  "schedule": { "kind": "cron", "expr": "0 22 * * *", "tz": "Asia/Shanghai" },
  "payload": {
    "kind": "agentTurn",
    "message": "Run ontology sync: python scripts/ontology_sync.py --days 7"
  },
  "sessionTarget": "isolated"
}

Memory Pipeline Outputs

Ontology Knowledge Graph Schema

Relationships and concepts are declared under memory/ontology/ inside a JSONL format backed by schemas:

• People & Orgs: Track identities via Person and Organization types.
• Work Entities: Organize workflows via Project, Task, and Goal structures.
• Spatio-Temporal Data: Detail coordinates using Event and Location contexts.
• Information Systems: Log records using Document, Note, and Message taxonomies.
• Meta Attributes: Classify system context using Technology, Skill, Action, and Policy labels.