GoClaw Docs

English

Tiếng Việt

English

Tiếng Việt

Appearance

Memory and Knowledge Graph ​

Memory Route: /memoryKnowledge Graph Route: /knowledge-graphSidebar Group: Data Access: Operator+

Overview ​

GoClaw provides two long-term knowledge storage mechanisms for agents:

  • Memory — stores structured text documents, supports semantic search via vector embeddings
  • Knowledge Graph — models relationships between entities, organized by agent and session

Memory ​

User Interface ​

Route: /memory

The page displays a memory documents table with columns: path, agent, scope (personal/global), hash, updated date. An embedding status badge is shown at the top.

Document Management ​

Create a new document

  1. Click the "Create Memory Document" button
  2. Fill in the dialog:
    • Agent — select the agent that owns the document
    • Scope — Global / Existing scope / Custom
    • Path — document identifier (e.g., user/preferences)
    • Content — text content
    • Auto-index — enable to embed vectors immediately after creation
  3. Click Create

View and edit

  • Click a document row to open the view dialog
  • Content tab: edit content, view path and metadata
  • Chunks tab: view indexed chunks with line ranges and embedding status

Delete a document

  • Click the delete button on a row, confirm in the dialog

Indexing ​

  • Re-index a single document — re-embed vectors for a specific document
  • Index All — bulk re-embed all documents (use when switching embedding providers)
  1. Click "Search Memory"
  2. Enter a search query (required)
  3. Optionally filter by user ID
  4. Results display: path, line range, similarity score bar, excerpt

Filters ​

  • Filter by agent — display documents for a specific agent or view all
  • Filter by user scope — filter by the user who owns the document

Knowledge Graph ​

User Interface ​

Route: /knowledge-graph

A knowledge graph entity viewer organized by agent and session. Includes an agent selector, scope selector (from existing sessions), and embedding status badge.

Usage ​

  1. Select an agent — choose the agent whose knowledge graph you want to view
  2. Filter by scope — filter by specific user or group
  3. View entities — browse the list of KG entities identified by the agent

Concepts ​

  • Entity — an object the agent recognizes (person, organization, concept, etc.)
  • Scope — the user or group that owns the entity
  • Graph — the network of relationships between entities

Note: Knowledge Graph is only available in the Standard edition. Desktop (Lite) does not support this feature.

Workflow Example ​

Saving user information to memory:

The agent automatically saves when it detects important information. Administrators can also create entries manually:

Path: users/nguyen-van-a/preferences
Content:
- Preferred language: Vietnamese
- Timezone: Asia/Ho_Chi_Minh
- Reply style: concise, with lists

Searching before responding:

The agent uses semantic search to retrieve relevant documents before generating a response, ensuring consistency with known user information.

Notes ​

  • Vector embeddings require an embedding provider configuration (e.g., OpenAI text-embedding-3-small)
  • If embeddings are not configured, memory still works but without semantic search capability
  • Documents with "Global" scope are shared across all users of that agent
  • Documents with "Personal" scope belong to a specific user only
  • Knowledge Graph depends on the LLM's entity recognition capability — quality varies by provider

See Also ​