All templates By NappAI team Deep Research Agent
Este flujo implementa un Agente de Investigación Autónomo que sigue un riguroso protocolo de cuatro fases (Planificar, Actuar, Procesar, Sintetizar) para responder a consultas complejas.
graph TD
%%{init: {'theme': 'mc','layout': 'elk'}}%%
ChatInput-im722[<div><img src="/_astro/messages-square.BaSDmT6g.svg" style="height: 20px !important;width: 20px !important"/></div>Chat Input]
style ChatInput-im722 stroke:#a170ff
ChatOutput-7h4uy[<div><img src="/_astro/messages-square.BaSDmT6g.svg" style="height: 20px !important;width: 20px !important"/></div>Chat Output]
style ChatOutput-7h4uy stroke:#a170ff
OpenAIModel-xyjkl[<div><img src="/_astro/openAI.BhmuxEs3.svg" style="height: 20px !important;width: 20px !important"/></div>OpenAI]
style OpenAIModel-xyjkl stroke:#a170ff
DeepResearchAgent-af1xa[Deep Research Agent]
style DeepResearchAgent-af1xa stroke:#a170ff
OpenAIWebSearchToolComponent-mezqb[<div><img src="/_astro/openAI.BhmuxEs3.svg" style="height: 20px !important;width: 20px !important"/></div>OpenAI Built-in tools]
style OpenAIWebSearchToolComponent-mezqb stroke:#a170ff
ChatInput-im722 -.- DeepResearchAgent-af1xa
linkStyle 0 stroke:#a170ff
DeepResearchAgent-af1xa -.- ChatOutput-7h4uy
linkStyle 1 stroke:#a170ff
OpenAIModel-xyjkl -.- DeepResearchAgent-af1xa
linkStyle 2 stroke:#a170ff
OpenAIWebSearchToolComponent-mezqb -.- DeepResearchAgent-af1xa
linkStyle 3 stroke:#a170ff
Deep Research Agent
🧩 Overview
This workflow implements an autonomous Deep Research Agent designed to handle complex information-gathering and analysis tasks. It automates a structured, multi-phase research process that begins with strategic planning, incorporates human validation, and culminates in the execution of comprehensive web searches and data synthesis. The agent is built to deliver coherent, analytical, and well-structured final reports based on user queries.
⚙️ Main Features
- Automates a four-phase research protocol: Plan, Act, Process, and Synthesize.
- Implements a Human-in-the-Loop (HITL) validation step, requiring user approval of the research plan before execution.
- Leverages a language model for strategic planning and decision-making.
- Utilizes integrated web search tools to gather information from the internet.
- Delivers final, structured responses directly to a chat interface.
🔄 Workflow Steps
| Component Name | Role in the Workflow | Key Inputs | Key Outputs |
|---|---|---|---|
| Chat Input | Captures the initial user query or command to initiate the research process. | User prompt text | User message for the agent |
| OpenAI Model | Serves as the core reasoning engine for the agent, handling strategic planning and decision-making. | Model configuration parameters | Language model instance |
| OpenAI Built-in Tools | Provides the agent with the capability to perform web searches to gather external information. | Tool configuration | Web search tool |
| Deep Research Agent | The central autonomous agent that orchestrates the entire research process, including planning, validation, and execution. | User query, Language model, Research tools | Final research response |
| Chat Output | Presents the agent's final, synthesized research report to the user. | Agent's response message | Displayed message in the chat interface |
🧠 Notes
- The workflow requires a configured OpenAI API key to power the language model and web search tools.
- The Human-in-the-Loop feature is enabled by default, pausing execution to present the research plan for user approval.
- The agent's memory and state management are configured to support iterative, stateful conversations.
- The final output language can be auto-detected from the input or specified by the user.