Back to Integrations
Integration/Ollama

Ollama Integration

Run local LLMs with Ollama and execute code in Cognitora sandboxes.

Category
LLM Providers
Difficulty
Intermediate
Tags
3 features
Setup Time
5-15 min
Local Models
Privacy
Open Source

Quick Start Guide

Installation

Terminal
pip install ollama cognitora

Implementation

Python
python
1import ollama
2from cognitora import Cognitora
3import re
4
5cognitora = Cognitora(api_key="your-cognitora-api-key")
6
7def extract_code_block(llm_response):
8    pattern = re.compile(r'```python\n(.*?)\n```', re.DOTALL)
9    match = pattern.search(llm_response)
10    if match:
11        return match.group(1)
12    return ""
13
14async def run_ollama_code_interpreter(user_query: str):
15    session = cognitora.sessions.create(
16        image="docker.io/library/python:3.11-slim",
17        timeout=300,
18        persistent=True
19    )
20
21    system_prompt = """You are a helpful coding assistant that can execute python code. 
22You are given tasks to complete and you run Python code to solve them.
23ALWAYS FORMAT YOUR RESPONSE IN MARKDOWN
24ALWAYS RESPOND ONLY WITH CODE IN CODE BLOCK LIKE THIS:
25```python
26{code}
27```"""
28
29    response = ollama.chat(
30        model="llama3.2",
31        messages=[
32            {"role": "system", "content": system_prompt},
33            {"role": "user", "content": user_query}
34        ]
35    )
36
37    content = response['message']['content']
38    code = extract_code_block(content)
39
40    if code:
41        execution = cognitora.compute.execute(
42            session_id=session.id,
43            command=["python", "-c", code]
44        )
45        
46        return {
47            "code": code,
48            "result": execution.stdout,
49            "error": execution.stderr,
50            "exit_code": execution.exit_code
51        }
52    
53    return {"error": "No code block found in response"}

Ready to integrate Ollama?

Get started with Ollama and Cognitora in minutes. Secure, scalable, and ready for anything.