17 · Ollama API, Modelfiles & Integrations

The local API

While Ollama runs, it serves an HTTP API at http://localhost:11434. Everything — n8n, Open WebUI, Hermes, your scripts — talks to it.

Endpoint	Use
`POST /api/chat`	Chat with messages (roles: system, user, assistant, tool)
`POST /api/generate`	Single prompt → completion
`POST /api/embed`	Create embeddings for RAG
`GET /api/tags`	List local models
`/v1/chat/completions` etc.	OpenAI-compatible endpoints

terminalbash

curl http://localhost:11434/api/chat -d '{
  "model": "qwen3:8b",
  "messages": [
    {"role": "system", "content": "You write short, upbeat marketing copy."},
    {"role": "user", "content": "3 taglines for a surf school in Siargao"}
  ],
  "stream": false
}'

chat.pypython

# pip install ollama
from ollama import chat

response = chat(
    model="qwen3:8b",
    messages=[{"role": "user", "content": "Summarize in 3 bullets: ..."}],
    options={"temperature": 0.2},
)
print(response.message.content)

# streaming
for part in chat(model="qwen3:8b", messages=[{"role": "user", "content": "Tell a short story"}], stream=True):
    print(part.message.content, end="", flush=True)

chat.mjsjavascript

// npm i ollama
import ollama from "ollama";

const response = await ollama.chat({
  model: "qwen3:8b",
  messages: [{ role: "user", content: "Write a 2-line description for handmade abaca bags" }],
});
console.log(response.message.content);

openai_compat.pypython

# pip install openai — any OpenAI-SDK tool can point at Ollama
from openai import OpenAI

client = OpenAI(base_url="http://localhost:11434/v1", api_key="ollama")  # key is required but ignored
r = client.chat.completions.create(
    model="qwen3:8b",
    messages=[{"role": "user", "content": "Hello from my laptop!"}],
)
print(r.choices[0].message.content)

This is why Ollama plugs into so much software: anywhere that accepts an “OpenAI-compatible base URL”, enter http://localhost:11434/v1.

Structured outputs: guaranteed JSON

Pass a JSON schema in the format field and Ollama constrains the model to produce matching JSON — perfect for extraction pipelines.

extract_receipt.pypython

from ollama import chat
from pydantic import BaseModel

class Receipt(BaseModel):
    vendor: str
    date: str
    total_php: float
    vat_php: float | None
    category: str

text = """ISLAND SUPPLY CO. Iloilo City  2026-09-21  Inv IS-88213
Printer ink x2, Bond paper x5   VAT 312.00   TOTAL 2,912.00"""

r = chat(
    model="qwen3:8b",
    messages=[{"role": "user", "content": f"Extract the receipt data. Category is one of: Office supplies, Software, Meals, Travel, Other.\n\n{text}"}],
    format=Receipt.model_json_schema(),
    options={"temperature": 0},
)
receipt = Receipt.model_validate_json(r.message.content)
print(receipt)

Modelfiles: package your own assistant

A Modelfile bundles a base model with a system prompt and parameters under a new name — like a custom GPT that runs locally.

ModelfileModelfile

FROM qwen3:8b

PARAMETER temperature 0.3
PARAMETER num_ctx 16384

SYSTEM """
You are the Barefoot Freelancer assistant for a Philippine-based web designer.
You write clear, friendly client emails, proposals and project updates.
- Default to concise answers with bullets where helpful.
- Prices are in PHP unless stated. Never invent prices or dates — ask.
- Client-facing text: warm and professional English.
"""

terminalbash

ollama create barefoot-assistant -f Modelfile
ollama run barefoot-assistant "Draft a project kickoff email for Kapé Iloilo"

Instruction	Purpose
`FROM`	Base model (from the library, or a local GGUF file path)
`PARAMETER`	Defaults like `temperature`, `num_ctx` (context), `stop`
`SYSTEM`	The built-in system prompt
`MESSAGE`	Seed example conversation turns (few-shot)
`ADAPTER`	Apply a fine-tuned LoRA adapter

Where to plug in your local models

⚙️

n8n

Ollama Chat Model and Embeddings nodes — local AI inside workflows (Lesson 14).

☤

Hermes Agent

Custom endpoint http://localhost:11434/v1 (Lesson 12).

💬

Open WebUI

A full ChatGPT-style interface for your team (next lesson).

🧑‍💻

Aider / VS Code

Local coding assistants via the OpenAI-compatible API.

17Ollama API, Modelfiles & Integrations

What you'll learn

The local API

Structured outputs: guaranteed JSON

Modelfiles: package your own assistant

Where to plug in your local models

n8n

Hermes Agent

Open WebUI

Aider / VS Code

Key takeaways

Knowledge check

Official links & further reading