Fixing Ollama Connection Refused (HTTP 11434) in Python and Docker

When calling a local Ollama instance from Python scripts, AI agents, or Docker containers, you may encounter:

httpx.ConnectError: [Errno 111] Connection refused
# or
urllib.error.URLError: <urlopen error [Errno 61] Connection refused>
# or
Failed to connect to localhost port 11434: Connection refused

This error happens for two primary reasons:
1. Host Binding: By default, Ollama binds exclusively to loopback 127.0.0.1, rejecting inbound requests originating outside localhost (such as from Docker containers or local network peers).
2. Daemon Status: The Ollama background service is not running or crashed.

Here is the exact step-by-step fix for Linux, macOS, and Docker environments.

Step 1: Ensure Ollama Service is Active

Before troubleshooting network bindings, check if the service is alive:

# On Linux (systemd)
systemctl status ollama

# If stopped, restart it:
sudo systemctl restart ollama

# On macOS
# Ensure the Ollama app icon is running in your menu bar, or start via terminal:
ollama serve

Test with curl http://127.0.0.1:11434/. If running, it will respond with:

Ollama is running

Step 2: Fix Host Binding (0.0.0.0) for Network & Docker Access

If calling Ollama from a Docker container or another machine, Ollama must listen on all network interfaces (0.0.0.0), not just 127.0.0.1.

On Linux (systemd service)

Edit the service configuration:

sudo systemctl edit ollama.service

Add the following lines under [Service]:

[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"
Environment="OLLAMA_ORIGINS=*"

Save, reload, and restart:

sudo systemctl daemon-reload
sudo systemctl restart ollama

On macOS

Quit the Ollama application, then launch it with the environment variable set:

launchctl setenv OLLAMA_HOST "0.0.0.0:11434"
# Or run directly in terminal:
OLLAMA_HOST=0.0.0.0:11434 ollama serve

Step 3: Accessing Ollama from Docker Containers

If your Python app runs inside Docker and Ollama runs on your host machine, localhost:11434 will point to the container itself, causing a connection failure.

Use host.docker.internal instead of localhost:

import os
from openai import OpenAI

# Automatically use host.docker.internal when inside Docker
base_url = os.getenv("OLLAMA_BASE_URL", "http://host.docker.internal:11434/v1")

client = OpenAI(
    base_url=base_url,
    api_key="ollama"
)

response = client.chat.completions.create(
    model="deepseek-r1:14b",
    messages=[{"role": "user", "content": "Hello!"}]
)

print(response.choices[0].message.content)

In docker-compose.yml, ensure the container can resolve the host:

services:
  agent_app:
    build: .
    extra_hosts:
      - "host.docker.internal:host-gateway"
    environment:
      - OLLAMA_BASE_URL=http://host.docker.internal:11434/v1

Next Steps

To implement autonomous tool calling with local models, follow our step-by-step tutorial on Local LLM Tool Calling with Ollama and Python.

To check memory requirements before downloading large models, use our Local LLM & DeepSeek-R1 VRAM Hardware Sizer.

Related posts:

Fixing Common LangChain Memory Deprecation and Migration to LangGraph Deploying DeepSeek-R1 and Llama 3.3 Locally with vLLM and Docker Quantizing and Running DeepSeek-R1 with Ollama, llama.cpp, and GGUF in Python Human-in-the-Loop AI Agents: Implementing Approval Breakpoints in LangGraph