Agentic Research Workflow

An automated research workflow powered by LangChain, Ollama, and LangGraph.
This project takes a research topic, generates research points, summarizes them, writes a full article, and finally critiques the result using a multi-node workflow.

---

Features

• Planner Node: Estimates research complexity and token usage.
• Research Node: Generates research points based on topic and complexity.
• Summarizer Node: Summarizes research results for writing.
• Writer Node: Produces a comprehensive article from the summary.
• Critic Node: Evaluates the article and determines next steps.

---

Tech Stack

• Python 3.11
• FastAPI for API endpoints
• LangChain + LangChain-Ollama for LLM interactions
• LangGraph for workflow orchestration
• Pydantic for state management
• Uvicorn as the ASGI server

---

Project Workflow

flowchart TD
    Start --> Planner["Planner Node"]
    Planner --> Researcher["Research Node"]
    Researcher --> Summarizer["Summarizer Node"]
    Summarizer --> Writer["Writer Node"]
    Writer --> Critic["Critic Node"]
    Critic -->|Pass| End["End"]
    Critic -->|Fail: research| Researcher
    Critic -->|Fail: summarize| Summarizer
    Critic -->|Fail: write| Writer

Loading

Installation

1. Clone the repository:

git clone <repository-url>
cd agentic-research-workflow

2. Install the dependencies

poetry install

3. Run the FastAPI server:

poetry run uvicorn app.main:app --reload

or Use launch.json