Deep Research Agent

A multi-agent research assistant built with Deep Agents on EdgeOne Makers — a lead researcher delegates sub-questions to expert subagents with web search, then synthesizes a comprehensive answer.

Framework: Deep Agents · Category: Quick Start · Language: TypeScript

Overview

Deep Research Agent turns a single question into a multi-step research process. A lead researcher plans the approach, dispatches expert subagents to search the web in parallel, and synthesizes their findings into a final answer — all streamed to the frontend in real time.

• Multi-agent research — a coordinator delegates 2-3 sub-questions to expert researcher subagents running in parallel
• Web search — each subagent performs multiple web searches to gather information from multiple angles
• Real-time streaming — SSE delivers planning text, subagent progress, tool calls, and the final synthesis token-by-token
• Conversation history — previous research sessions are persisted via LangGraph checkpointer and restorable from the home history list
• Stop generation — abort a running research at any time; pending subagents are cancelled gracefully

Environment Variables

Variable	Required	Description
AI_GATEWAY_API_KEY	Yes	Model gateway API key. Use your Makers Models API Key, or any OpenAI-compatible provider key.
AI_GATEWAY_BASE_URL	Yes	Gateway base URL. For Makers Models, use https://ai-gateway.edgeone.link/v1.
WSA_API_KEY	No	Tencent Cloud Web Search API (WSA) key for the platform's built-in search tool.

This template follows the OpenAI-compatible standard — you can point these variables at Makers Models or any other compatible gateway / provider.

How to get AI_GATEWAY_API_KEY

1. Open the Makers Console.
2. Sign in and enable Makers.
3. Go to Makers → Models → API Key and create a key.
4. Copy it into AI_GATEWAY_API_KEY (set AI_GATEWAY_BASE_URL to https://ai-gateway.edgeone.link/v1).

Built-in models (@makers/deepseek-v4-flash, @makers/hy3-preview, @makers/minimax-m2.7) are free and rate-limited — great for prototyping. For production, bind your own provider key (BYOK) in the console.

How to get WSA_API_KEY

The platform provides a built-in search tool powered by Tencent Cloud Web Search API (WSA). To use it:

1. Enable the Web Search (WSA) service in the Tencent Cloud WSA Console.
2. Obtain your API Key and set it as WSA_API_KEY.
3. See the WSA API Documentation for details.

If you prefer not to use Tencent Cloud WSA, you can integrate a third-party search service (e.g. Exa, Tavily) by replacing the web_search tool implementation.

Local Development

Prerequisites: Node.js, npm

npm install
cp .env.example .env
edgeone makers dev

Open http://localhost:8080/agent-metrics for the local observability panel.

Project Structure

deepagents-research-nodejs/
├── agents/
│   ├── stream.ts          # /stream — main research endpoint (SSE streaming + history + delete)
│   ├── stop.ts            # /stop — abort an active research run
│   └── _logger.ts         # Shared logger factory
├── src/
│   ├── components/        # React UI components (ChatPage, SubAgentCard, etc.)
│   ├── hooks/
│   │   ├── useAgentStream.ts  # SSE streaming hook + state management
│   │   └── useLanguage.tsx    # i18n context provider
│   ├── i18n/              # Internationalization (en/zh)
│   └── lib/types.ts       # Shared TypeScript types
├── edgeone.json           # Agent runtime configuration
└── package.json

Files prefixed with _ are private modules — not exposed as public routes by EdgeOne.

How It Works

The agent runs as a session-mode runtime: requests sharing the same conversation_id are routed to the same instance with persistent state.

Workflow

1. User asks a question — the frontend sends a POST to /stream with makers-conversation-id header.
2. Planning — the lead researcher outputs a brief research plan (1-2 sentences) and dispatches 2-3 sub-questions in parallel via the task tool.
3. Researching — each expert subagent performs multiple web searches, gathers data, and writes a summary. Progress streams as subagent_pending → tool_call → tool → ai → subagent_complete events.
4. Synthesizing — once all subagents finish, the lead researcher synthesizes a comprehensive final answer.
5. Complete — the stream ends with [DONE]; the conversation state is checkpointed for future restoration.

Key Mechanisms

• Deep Agents + LangGraph: the createDeepAgent function builds a LangGraph graph with subagent orchestration, middleware (retry), checkpointer, and store.
• Dual stream mode: uses stream() with streamMode: ['updates', 'messages'] and subgraphs: true to capture both lifecycle events and token-level text, mapped to typed SSE events (ai, tool_call, tool, subagent_pending/step/complete, error).
• Platform tools: web_search is provided by the EdgeOne Makers runtime via context.tools.toLangChainTools().
• Checkpointer: conversation state (messages, tool results) is persisted via context.store.langgraphCheckpointer, enabling history restoration.

Routes

Route	Method	Description
/stream	POST	Main research (SSE streaming), also handles action: "history" and action: "delete"
/stop	POST	Abort an active research run

The conversation_id is passed via the makers-conversation-id request header.

Resources

• Makers Agents Documentation
• Quick Start: Agent Development
• Makers Models

License

MIT