Groq Deep Research Agent Example

This guide demonstrates how to use the groq-deep-research-agent-51 graph for automated research tasks.

Basic Usage

from lmsystems import SyncLmsystemsClient, APIError
import os

# Initialize client with graph name
client = SyncLmsystemsClient(
    graph_name="groq-deep-research-agent-51",
    api_key=os.environ.get("LMSYSTEMS_API_KEY")
)

def basic_research(topic: str):
    try:
        # Create research thread
        thread = client.threads.create()

        # Stream research results
        for chunk in client.stream_run(
            thread=thread,
            input={
                "research_topic": topic
            },
            config={
                "configurable": {
                    "llm": "llama-3.1-8b-instant",
                    "groq_api_key": os.environ.get("GROQ_API_KEY"),
                    "tavily_api_key": os.environ.get("TAVILY_API_KEY"),
                    "max_web_research_loops": 3  # Optional
                }
            },
            stream_mode=["messages", "updates"]
        ):
            if "message" in chunk:
                print(f"New message: {chunk['message']['content']}")
            elif "update" in chunk:
                print(f"Research update: {chunk['update']}")

    except APIError as e:
        print(f"Research failed: {str(e)}")

if __name__ == "__main__":
    basic_research("AI safety alignment recent developments")

Input Parameters Guide

Parameter	Required	Type	Description
research_topic	Yes	string	Primary research subject
search_query	No	string	Custom search query (default: generated from topic)
running_summary	No	string	Initial summary for iterative research
research_loop_count	No	integer	Start from specific iteration count
web_research_results	No	array	Pre-existing research materials

Configuration Options

config = {
    "configurable": {
        "llm": "",  # Required Groq model
        "groq_api_key": "your-groq-key",  # Required
        "tavily_api_key": "your-tavily-key",  # Required
        "max_web_research_loops": 5  # Optional (default: 3)
    }
}

Error Handling

from lmsystems import (
    AuthenticationError,
    InputError,
    GraphExecutionError
)

try:
    # Research execution code
except InputError as e:
    print(f"Missing required parameter: {e}")
except AuthenticationError as e:
    print(f"Check your API keys: {e}")
except GraphExecutionError as e:
    print(f"Research failed at step {e.node_id}: {e}")

Best Practices

1. API Keys

   • Get Groq API key from Groq Console
   • Get Tavily key from Tavily
   • Never hardcode keys - use environment variables

2. Stream Handling

   async for chunk in client.stream_run(...):
       if "message" in chunk:
           # Handle assistant messages
       elif "value" in chunk:
           # Track state changes like loop_count
       elif "update" in chunk:
           # Monitor research progress

3. Model Selection

   • Supported Groq models: I believe all models are supported (including deepseek-r1), although i've only tested llama and deepseek models.

4. Result Processing

   final_state = client.get_thread_state(thread)
   print(f"Final summary: {final_state['running_summary']}")
   print(f"Sources used: {len(final_state['sources_gathered'])}")