---
title: Report Writer Agent
emoji: 🏆
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 5.28.0
app_file: app.py
pinned: false
license: mit
short_description: Report Writing Agent by Nader Afshar
---

# Research-Driven Report Writing Agent

# 🔎 Research Agent

The **Research Agent** is an autonomous, multi-step research system built using [LangGraph](https://github.com/langchain-ai/langgraph), [Cohere](https://cohere.com), and [Tavily](https://www.tavily.com/). It mimics the behavior of a human research assistant — planning, researching, drafting, critiquing, and revising — to produce high-quality **research reports** on any topic the user provides.

---

## 🧠 What Sets It Apart?

Unlike typical AI content generators, this agent performs **real-time research** to ensure that reports are grounded in current, verifiable information.

### ✨ Highlights:
- Uses **Cohere’s `command-a-03-2025` LLM** for language generation, outlining, and critique.
- Uses **Tavily** to perform real-time, internet-based research (not just LLM memory).
- Produces **structured, refined research reports**, not informal essays or chatbot replies.
- Streams its **research and reasoning process** live so users can follow along.

---

## 🧩 How It Works

1. **User provides a topic or research question** through the UI.
2. The agent:
   - **Plans** the structure of the research.
   - **Conducts multiple web searches** via Tavily.
   - **Drafts a full report** using retrieved content and context.
   - **Critiques** and **revises** the report internally.
3. The output includes:
   - A live, streaming **log of the research process**.
   - A clean, **final research report** as a separate result.

---

## ⏳ Why It Takes Time

The Research Agent behaves more like a human researcher than a chatbot:

- It may take **~30 seconds before the first visible output**, while:
  - Querying the web.
  - Organizing thoughts.
  - Planning the structure.
- It takes **1–2 minutes or more** to complete a full research and writing cycle with critique + revision.

🧠 **This is intentional**: speed is traded for depth, accuracy, and reasoning.

---

## 🖥 Interface

Built with **Gradio**, the app includes:

- A **textbox** for entering a topic or question.
- A **streaming log** of research, planning, and revision steps.
- A **final output panel** for the completed research report.

---

## 🛠️ Technologies Used

| Component      | Purpose                                  |
|----------------|------------------------------------------|
| LangGraph      | Agent state machine + step logic         |
| Cohere LLM     | `command-a-03-2025` model for text tasks |
| Tavily Search  | Real-time research via web search        |
| LangChain Core | Messaging and prompt orchestration       |
| Gradio         | Interactive web interface                |

---

## 🚀 Getting Started

1. Install dependencies:

   ```bash
   pip install -r requirements.txt