How to Configure the memory-tencentdb Plugin: Complete Environment and File Settings

The memory-tencentdb plugin is configured through environment variables, YAML/JSON gateway configuration files, and OpenClaw plugin manifests, with settings resolving in a specific priority order from explicit file paths to environment overrides.

The memory-tencentdb plugin serves as the core memory backend for the TencentDB Agent, persisting conversation history and contextual data for AI-driven database operations. According to the TencentCloud/TencentDB-Agent-Memory source code, the plugin supports multiple configuration layers that allow you to customize everything from network binding to LLM integration. Understanding these configuration settings ensures secure deployment across development, staging, and production environments.

Core Configuration Methods

The plugin accepts configuration through three primary channels: environment variables for runtime overrides, structured configuration files for persistent settings, and plugin registry entries for OpenClaw and Hermes integration.

Environment Variables

The MemoryCore gateway reads several environment variables to control runtime behavior. These variables override file-based configurations and are essential for containerized deployments.

• TDAI_GATEWAY_HOST: Bind address for the HTTP gateway. Default is 127.0.0.1. Set to 0.0.0.0 for external exposure.
• TDAI_GATEWAY_PORT: TCP port for the gateway listener. Default is 8420.
• TDAI_GATEWAY_API_KEY: Bearer token required when binding to non-loopback addresses. No default value; must be set for secure remote access.
• TDAI_CORS_ORIGINS: Comma-separated list of allowed origins for cross-origin requests. Empty by default.
• TDAI_DATA_DIR: Root directory for persisted memory files including SQLite databases and binary blobs. Defaults to ~/.memory-tencentdb/memory-tdai.
• TDAI_LLM_API_KEY: API key for the OpenAI-compatible LLM service used for embeddings and generation. Empty by default.
• TDAI_LLM_BASE_URL: Base URL for the LLM endpoint. Default is https://api.openai.com/v1.
• TDAI_LLM_MODEL: Model identifier for memory-related LLM calls. Default is gpt-4o.
• TDAI_GATEWAY_CONFIG: Explicit path to a YAML or JSON configuration file. If unset, the gateway auto-discovers tdai-gateway.yaml or tdai-gateway.json in the current or data directories.

Gateway Configuration Files

In MemoryCore/tdai-gateway.yaml, you define persistent gateway settings including TLS certificates, logging levels, and storage backends. The gateway resolves configuration files in the following priority order as implemented in MemoryCore/src/gateway/server.ts:

1. File specified by TDAI_GATEWAY_CONFIG environment variable
2. tdai-gateway.yaml or tdai-gateway.json in the current working directory
3. Same configuration files in the data directory (TDAI_DATA_DIR)
4. Environment variables (which override any file values)

Plugin Registration

The plugin registers in OpenClaw through the manifest defined in MemoryCore/openclaw-plugin/openclaw.plugin.json. The canonical slot name is memory-tencentdb-client for v3 implementations, while memory-tencentdb remains supported for backward compatibility.

For Hermes integration, configure the provider in ~/.hermes/config.yaml under the memory key, referencing the same gateway endpoint and API key used by MemoryCore.

Installation and Setup Examples

Configure the gateway and install the plugin using the provided installation script and environment exports.

# Set up the gateway environment

export TDAI_GATEWAY_HOST="0.0.0.0"
export TDAI_GATEWAY_API_KEY="my-strong-token"
export TDAI_DATA_DIR="${HOME}/.memory-tencentdb/memory-tdai"
export TDAI_LLM_API_KEY="sk-xxxxxxxxxxxx"
export TDAI_LLM_MODEL="gpt-4o-mini"

# Start the MemoryCore gateway

node --import tsx MemoryCore/src/gateway/server.ts

Install the plugin into OpenClaw using the automated script:

bash MemoryCore/scripts/install-openclaw-plugin.sh

The install-openclaw-plugin.sh script performs three operations: it downloads the npm package @tencentdb-agent-memory/memory-tencentdb, builds the TypeScript sources, and registers the plugin in ~/.openclaw/openclaw.json under the plugins.slots.memory key.

Verify the installation via CLI:

openclaw plugins list | grep memory-tencentdb

Configure Hermes to use the memory provider:

# ~/.hermes/config.yaml

memory:
  provider: memory-tencentdb
  endpoint: "http://127.0.0.1:8420"
  api_key: "my-strong-token"

Key Configuration Files

These source files define the complete configuration surface for the memory-tencentdb plugin:

File Path	Purpose
MemoryCore/README.md	Documents all runtime environment variables and quick-start procedures
MemoryCore/tdai-gateway.yaml	Default gateway configuration template with bind addresses and storage paths
MemoryCore/openclaw-plugin/openclaw.plugin.json	Declares plugin ID (memory-tencentdb) and capability slots
MemoryCore/scripts/install-openclaw-plugin.sh	Automates npm installation and OpenClaw registry updates
MemoryCore/hermes-plugin/memory/memory_tencentdb/README.md	Documents Hermes provider aliases and default data locations
MemoryCore/SKILL.md	Defines OpenClaw skill triggers for automatic plugin configuration

Summary

• The memory-tencentdb plugin uses environment variables for runtime overrides and YAML/JSON files for persistent configuration.
• Configuration resolution follows a strict priority order: explicit config file → discovered config files → environment variables.
• All persistent data defaults to ~/.memory-tencentdb/memory-tdai but is configurable via TDAI_DATA_DIR.
• The install-openclaw-plugin.sh script automates registration in ~/.openclaw/openclaw.json using the slot name memory-tencentdb-client.
• Hermes integration requires specifying the memory_tencentdb provider with matching endpoint and API key credentials.

Frequently Asked Questions

What is the default port for the memory-tencentdb gateway?

The default port is 8420, controlled by the TDAI_GATEWAY_PORT environment variable. You can override this in tdai-gateway.yaml or via the environment variable before starting the gateway in MemoryCore/src/gateway/server.ts.

How do I secure the gateway when binding to external interfaces?

Set TDAI_GATEWAY_HOST to 0.0.0.0 for external access, but you must configure TDAI_GATEWAY_API_KEY with a strong bearer token. The gateway rejects unauthenticated requests when bound to non-loopback addresses, as enforced by the authentication middleware in the MemoryCore source.

Where does the plugin store conversation history and embeddings?

All persisted data including SQLite databases, binary blobs, and index files reside in the directory specified by TDAI_DATA_DIR, defaulting to ~/.memory-tencentdb/memory-tdai. This path is created automatically during the first run if it does not exist.

Can I use a local LLM instead of OpenAI for embeddings?

Yes. Override TDAI_LLM_BASE_URL to point to your local OpenAI-compatible endpoint (such as Ollama or vLLM) and set TDAI_LLM_MODEL to your local model identifier. The plugin uses standard OpenAI SDK calls, so any API-compatible service works without code changes.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}