[ROADMAP][1.5] Implement deep agent chat in LangGraph

[MODSetter]

Feature Description

Create a new deep/long-running agent implementation using LangGraph, designed for complex research tasks that require multi-step reasoning, tool usage, and extended context. This is the core agent that will power SurfSense's advanced research capabilities.

Target Deployment

• SurfSense Cloud (hosted version)
• Self-hosted version

Problem Statement

Current agent implementation lacks:

• Support for long-running, multi-step research tasks
• Sophisticated state management for complex workflows
• Human-in-the-loop capabilities for validation checkpoints
• Optimized tool orchestration for research use cases

Reference: LangChain Deep Agents Blog