Persistent Memory Infrastructure for AI Agents

Project Overview

This project provides a persistent memory infrastructure for AI coding agents, allowing context and knowledge to remain available when switching between agents such as Claude Code and OpenCode.

Agents can change, but project context and knowledge remains persistent.

Architecture

                         Developer
                             │
                             ▼
                    ┌─────────────────┐
                    │   Agent Code    │
                    │                 │
                    │ Claude Code     │
                    │ OpenCode        │
                    └────────┬────────┘
                             │
                             ▼
                    ┌─────────────────┐
                    │      TDAI       │
                    │ TencentDB Agent │
                    │     Memory      │
                    └────────┬────────┘
                             │
                             ▼
                    ┌─────────────────┐
                    │    OmniRoute    │
                    │   LLM Provider  │
                    └─────────────────┘

The infrastructure is accessed remotely through Tailscale VPN.

Core Components

Component	Purpose
Claude Code / OpenCode	AI coding agents
TDAI (TencentDB Agent Memory)	Persistent memory and context management
OmniRoute	Centralized LLM provider and routing layer
Tailscale	Secure remote access

Cross-Agent Persistent Memory

TDAI provides a shared memory layer between different AI agents.

For example:

Developer
    │
    ▼
OpenCode
    │
    ▼
TDAI
    │
    ▼
Persistent Memory
    │
    ▼
Claude Code
    │
    ▼
TDAI
    │
    ▼
Previous Context

This allows the developer to switch between:

OpenCode → Claude Code → OpenCode

while maintaining shared project knowledge.

LLM Routing

TDAI uses OmniRoute as its LLM provider.

TDAI
 │
 ▼
OmniRoute
 │
 └── LLM Providers

OmniRoute provides a centralized interface for connecting TDAI with different LLM services.

Remote Access

The infrastructure can run on a separate server and be accessed through Tailscale VPN.

Developer PC
     │
     │ Tailscale VPN
     ▼
AI Infrastructure
     │
     ├── TDAI
     └── OmniRoute

Primary Goal

The system creates a persistent shared-memory layer for AI coding agents.

Core Principles

Persistent memory — context survives individual agent sessions.
Cross-agent continuity — different agents can access shared project knowledge.
LLM flexibility — OmniRoute provides the LLM interface for TDAI.
Remote infrastructure — TDAI and OmniRoute can operate independently from the developer's machine.

Agents and models can change, but the project's accumulated knowledge remains persistent.