YouTip

1. Start the Ollama Service

\\n

Ollama provides an HTTP-based API that allows developers to interact with models programmatically.

\\n

This article will detail the usage of the Ollama API, including request formats, response formats, and example code.

\\n

---

\\n

2. API Endpoints

\\n

Ollama provides the following main API endpoints:

\\n

Generate Text

\\n

\\n
• Endpoint: POST /api/generate
\\n
• Function: Send a prompt to the model and retrieve the generated text.
\\n
• Request Format:
\\n

{ "model": "", // model name "prompt": "", // input prompt "stream": false, // enable streaming response（default false) "options": { // Optional parameter "temperature": 0.7, // Temperature parameter "max_tokens": 100 // Maximum token count }}

\\n
• Response Format:
\\n

{ "response": "", // generated text "done": true // done}

\\n

\\n

Chat

\\n

\\n
• Endpoint: POST /api/chat
\\n
• Function: Supports multi-turn conversations; the model retains context.
\\n
• Request Format:
\\n

{ "model": "", // model name "messages": [ // message list { "role": "user", // user role "content": "" // User input } ], "stream": false, // enable streaming response "options": { // Optional parameter "temperature": 0.7, "max_tokens": 100 }}

\\n
• Response Format:
\\n

{ "message": { "role": "assistant", // assistant role "content": "" // generated text }, "done": true}

\\n

\\n

List Local Models

\\n

\\n
• Endpoint: GET /api/tags
\\n
• Function: List locally downloaded models.
\\n
• Response Format:
\\n

{ "models": [ { "name": "", // model name "size": "", // model size "modified_at": "" // modified time } ]}

\\n

\\n

Pull Model

\\n

\\n
• Endpoint: POST /api/pull
\\n
• Function: Pull a model from the model library.
\\n
• Request Format:
\\n

{ "name": "" // model name}

\\n
• Response Format:
\\n

{ "status": "downloading", // download status "digest": "" // model summary}

\\n

\\n

---

\\n

3. Usage Examples

\\n

Generate Text

\\n

Send a request using curl:

\\n

Example

\\n

curl http://localhost:11434/api/generate -d'{\\n\\n  "model": "deepseek-coder",\\n\\n  "prompt": "Hello, can you help me write a code snippet?",\\n\\n  "stream": false\\n\\n  }'

\\n

Multi-turn Conversation

\\n

Send a request using curl:

\\n

Example

\\n

curl http://localhost:11434/api/chat -d'{\\n\\n  "model": "deepseek-coder",\\n\\n  "messages": [\\n\\n   {\\n\\n    "role": "user",\\n\\n    "content": "Hello, can you help me write a Python code snippet?"\\n\\n   }\\n\\n  ],\\n\\n  "stream": false\\n\\n  }'

\\n

List Local Models

\\n

Send a request using curl:

\\n

curl http://localhost:11434/api/tags

\\n

Pull Model

\\n

Send a request using curl:

\\n

Example

\\n

curl http://localhost:11434/api/pull -d'{\\n\\n  "name": "deepseek-coder"\\n\\n  }'

\\n

---

\\n

4. Streaming Responses

\\n

Ollama supports streaming responses, which are suitable for scenarios requiring real-time text generation.

\\n

Enable Streaming Response

\\n

Set "stream": true in the request, and the API will return the generated text line by line.

\\n

Example

\\n

curl http://localhost:11434/api/generate -d'{\\n\\n  "model": "deepseek-coder",\\n\\n  "prompt": "Hello, can you help me write a code snippet?",\\n\\n  "stream": true\\n\\n  }'

\\n

Response Format

\\n

Each line returns a JSON object:

\\n

Example

\\n

{\\n\\n"response":"",// partially generated text\\n\\n"done":false// done\\n\\n}

\\n

---

\\n

5. Programming Language Examples

\\n

Python: Interact with the Ollama API using the requests library:

\\n

Example

\\n

import requests\\n\\n# generate text\\n\\n response = requests.post(\\n\\n"http://localhost:11434/api/generate",\\n\\n json={\\n\\n"model": "deepseek-coder",\\n\\n"prompt": "Hello, can you help me write a code snippet?",\\n\\n"stream": False\\n\\n}\\n\\n)\\n\\nprint(response.json())

\\n

Multi-turn conversation:

\\n

Example

\\n

response = requests.post(\\n\\n"http://localhost:11434/api/chat",\\n\\n json={\\n\\n"model": "deepseek-coder",\\n\\n"messages": [\\n\\n{\\n\\n "role": "user",\\n\\n "content": "Hello, can you help me write a Python code snippet?"\\n\\n}\\n\\n],\\n\\n"stream": False\\n\\n}\\n\\n)\\n\\nprint(response.json())

\\n

JavaScript: Interact with Ollama using the fetch API:

\\n

Example

\\n

// generate text\\n\\n fetch("http://localhost:11434/api/generate",{\\n\\n  method:"POST",\\n\\n  headers:{"Content-Type":"application/json"},\\n\\n  body: JSON.stringify({\\n\\n   model:"deepseek-coder",\\n\\n   prompt:"Hello, can you help me write a code snippet?",\\n\\n   stream:false\\n\\n})\\n\\n})\\n\\n .then(response => response.json())\\n\\n .then(data => console.log(data));

\\n

Multi-turn conversation:

\\n

Example

\\n

fetch("http://localhost:11434/api/chat",{\\n\\n  method:"POST",\\n\\n  headers:{"Content-Type":"application/json"},\\n\\n  body: JSON.stringify({\\n\\n   model:"deepseek-coder",\\n\\n   messages:[\\n\\n{\\n\\n  role:"user",\\n\\n  content:"Hello, can you help me write a Python code snippet?"\\n\\n}\\n\\n],\\n\\n stream:false\\n\\n})\\n\\n})\\n\\n .then(response => response.json())\\n\\n .then(data => console.log(data));