Skip to content

DNALLM Documentation

Pydantic AI

zhangtaolab/DNALLM

Pydantic AI with Ollama and MCP Server¶

This tutorial demonstrates how to use Pydantic AI with Ollama models and the DNALLM MCP server for DNA sequence analysis. Pydantic AI provides structured output validation and type-safe tool calling.

Full Notebook¶

View Full Notebook

Prerequisites¶

# Install Python packages
uv pip install pydantic-ai nest-asyncio

# Start Ollama and pull the model
ollama pull qwen3.6:latest

Start the DNALLM MCP server in a separate terminal:

dnallm mcp-server --transport streamable-http

Configure Async Environment¶

import nest_asyncio

nest_asyncio.apply()

Set Up Ollama Model¶

from pydantic_ai.models.openai import OpenAIChatModel
from pydantic_ai.providers.ollama import OllamaProvider

ollama_model = OpenAIChatModel(
    model_name='qwen3.6:latest',
    provider=OllamaProvider(base_url='http://localhost:11434/v1'),
)

Connect MCP Server¶

from pydantic_ai.mcp import MCPServerStreamableHTTP

server = MCPServerStreamableHTTP('http://localhost:8000/mcp')

Create Agent¶

Create a Pydantic AI agent with MCP tool access and a system prompt that guides DNA analysis:

from pydantic_ai import Agent

# Create agent with MCP server tools and proper system prompt
agent_ollama = Agent(
    ollama_model, 
    toolsets=[server],
    system_prompt='''You are a DNA analysis assistant with access to specialized DNA analysis tools via MCP server.

When analyzing a DNA sequence, you should:
1. First call _list_loaded_models to see what models are available
2. Then call dna_multi_model_predict with the DNA sequence and appropriate model names
3. Interpret and explain the results in a comprehensive way, not just only list the results

Available tools should include:
- _list_loaded_models: Lists available DNA analysis models
- dna_multi_model_predict: Predicts DNA sequence properties using multiple models

Always use the tools to provide accurate analysis. Based on the returned results, make reasonable inferences with comprehensive biological functions of this sequence.'''
)

Analyze DNA Sequence¶

# Analyze DNA sequence using MCP server with proper async context
async def analyze_dna_sequence():
    async with agent_ollama:  # This ensures proper MCP server connection
        result = await agent_ollama.run(
            'What is the function of following DNA sequence? Please analyze it thoroughly using all available models: AGAAAAAACATGACAAGAAATCGATAATAATACAAAAGCTATGATGGTGTGCAATGTCCGTGTGCATGCGTGCACGCATTGCAACCGGCCCAAATCAAGGCCCATCGATCAGTGAATACTCATGGGCCGGCGGCCCACCACCGCTTCATCTCCTCCTCCGACGACGGGAGCACCCCCGCCGCATCGCCACCGACGAGGA
GGAGGCCATTGCCGGCGGCGCCCCCGGTGAGCCGCTGCACCACGTCCCTGA'
        )
        return result

# Run the analysis
result = await analyze_dna_sequence()

time.sleep(3)

Display Results¶

print("=== DNA Sequence Analysis Result ===")
print(result.output)
print("\n=== Usage Statistics ===")
print(result.usage())

List Available Models¶

You can also test individual tools:

# Alternative approach: Test individual tool calls to debug
async def test_individual_tools():
    async with agent_ollama:
        # First, let's try to list available tools
        print("=== Testing tool availability ===")

        # Try to get the agent to use the list models tool
        list_result = await agent_ollama.run(
            'Please list all the available DNA analysis models using the _list_loaded_models tool.'
        )
        print("List models result:")
        print(list_result.output)

        return list_result

# Run individual tool test
list_result = await test_individual_tools()