Deep Research Agent

Two pieces of deep research agent developed [independent pieces for easy POC]

• Scoping Phase → Understands user intent and generates a research brief
• Researching Phase → Gathers, filters, and synthesizes information and generates consolidated summary for the reearch brief

Designed for high-quality, citation-backed insights while optimizing for token efficiency and source reliability.

---

🧠 Architecture Overview

🔹 1. Scoping Phase (Research Brief Generation)

• Clarifies user intent (if needed)
• Generates a structured research brief
• Ensures alignment before deep research begins

Flow

---

🔹 2. Researching Phase (Deep Research Execution)

• Executes iterative search + reasoning loop
• Filters low-quality sources
• Produces:
  • ✅ Consolidated summary
  • ✅ Report-style findings with citations

Flow

---

⚙️ Models & Tools

Component	Tool / Model	Purpose
🌐 Web Search	Tavily	Retrieve search results
🧠 Reasoning	llama-3.3-70b-versatile	Generate research brief
✂️ Summarization	llama-3.1-8b-instant	Chunking + summarization
🎯 Orchestration	openai/gpt-oss-120b	Agent control & decision making

---

🔍 Key Features

✅ High-Quality Source Filtering

• Skips sources without raw_content
• Avoids low-signal aggregator/listing pages
• Focuses on insightful, content-rich sources

---

✅ Token Optimization Strategy

• Chunk-based summarization
• Avoids full-page ingestion by smart chunk selection strategy

---

✅ Rate Limit Handling

• Introduces 15-second delay between searches
• Prevents token-per-minute overflow issues

---

---

⚙️ Configuration

🔹 1. Create Environment

Create a separate environment and install dependencies:

python -m venv langchain_env
source langchain_env/bin/activate   # Mac/Linux
langchain_env\Scripts\activate      # Windows

pip install -r requirements.txt

---

🔹 2. Setup Environment Variables

Create a .env file in the project root:

GROQ_API_KEY=your_key
TAVILY_API_KEY=your_key
LANGSMITH_TRACING=True
LANGSMITH_API_KEY=your_key
LANGSMITH_PROJECT=deep_research_agent

---

🔹 3. Setup Jupyter Kernel (Important)

Run these commands after activating the environment:

pip install ipykernel
python -m ipykernel install --user --name=langchain_env --display-name "Python (langchain_env)"

Then select Python (langchain_env) as the kernel inside Jupyter Notebook.

---

🔹 4. LangGraph Deployment

• Ensure langgraph.json is present in the project
• Start local deployment:

langgraph dev