🤖 Gemma AI API — OpenAI-Compatible Cloudflare Worker

A free, OpenAI-compatible AI API powered by Google Gemma 4 26B running on Cloudflare Workers AI at the edge. Drop-in replacement for OpenAI's /v1/chat/completions endpoint — no credit card, no GPU needed.

Built by Nimesh Piyumal

---

🚀 One-Click Deploy

Click the button above to instantly deploy your own instance to Cloudflare Workers — no local setup needed!

---

✨ Features

• 🆓 Free to run — powered by Cloudflare Workers AI free tier (10,000 neurons/day)
• ⚡ Edge inference — runs on Cloudflare's 300+ global locations for ultra-low latency
• 🔌 OpenAI-compatible — works with any OpenAI SDK by just changing the base_url
• 🌊 Streaming support — real-time token streaming via Server-Sent Events (SSE)
• 🌍 CORS enabled — ready for browser-based frontends
• 🧠 Powered by Gemma 4 26B — Google's latest open model with reasoning capabilities

---

🚀 Quick Start

Deploy Your Own

1. Clone this repo

   git clone https://github.com/nimesh-piyumal/Gemma-API.git
   cd Gemma-API

2. Install Wrangler CLI

   npm install -g wrangler
   wrangler login

3. Deploy to Cloudflare Workers

   wrangler deploy

4. Done! Your API is live at https://your-worker.your-subdomain.workers.dev

---

📡 API Endpoints

GET /

Returns API info.

{
  "success": true,
  "creator": "Nimesh Piyumal",
  "website": "https://ceylonnet.com"
}

---

POST /v1/chat/completions

OpenAI-compatible chat endpoint. Supports both streaming and non-streaming.

Request

{
  "model": "gemma-4-26b-a4b-it",
  "messages": [
    { "role": "user", "content": "Explain quantum computing in one sentence." }
  ],
  "stream": false
}

Response

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1779494702,
  "model": "gemma-4-26b-a4b-it",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Quantum computing uses quantum mechanics to process information in multiple states simultaneously..."
      },
      "finish_reason": "stop"
    }
  ]
}

---

GET /debug

Returns the raw Cloudflare AI response object. Useful for debugging model output format.

---

🐍 Python Usage

import requests
import json

BASE_URL = "https://your-worker.workers.dev"

# Non-streaming
response = requests.post(f"{BASE_URL}/v1/chat/completions", json={
    "model": "gemma-4-26b-a4b-it",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": False
})
print(response.json()["choices"][0]["message"]["content"])

# Streaming
with requests.post(f"{BASE_URL}/v1/chat/completions", json={
    "model": "gemma-4-26b-a4b-it",
    "messages": [{"role": "user", "content": "Count from 1 to 5."}],
    "stream": True
}, stream=True) as r:
    for line in r.iter_lines(decode_unicode=True):
        if line.startswith("data: ") and line != "data: [DONE]":
            data = json.loads(line[6:])
            delta = data["choices"][0].get("delta", {})
            if "content" in delta:
                print(delta["content"], end="", flush=True)

---

🔌 OpenAI SDK Usage

from openai import OpenAI

client = OpenAI(
    base_url="https://your-worker.workers.dev/v1",
    api_key="not-needed"  # Required by SDK but ignored by the worker
)

response = client.chat.completions.create(
    model="gemma-4-26b-a4b-it",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

// Node.js
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://your-worker.workers.dev/v1",
  apiKey: "not-needed",
});

const response = await client.chat.completions.create({
  model: "gemma-4-26b-a4b-it",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0].message.content);

---

🧠 Model Info

Property	Value
Model	Google Gemma 4 26B Instruct
Model ID	@cf/google/gemma-4-26b-a4b-it
Provider	Cloudflare Workers AI
Context Window	Large
Capabilities	Text generation, reasoning, instruction following
Vision	✅ Supported
Function Calling	✅ Supported

---

💸 Pricing

Plan	Neurons/day	Cost
Free	10,000	$0
Paid	Unlimited	$0.011 / 1,000 neurons

A typical ~500 token response costs roughly 400–600 neurons.

---

⚙️ Configuration

To swap the model, edit worker.js:

const aiModel = '@cf/google/gemma-4-26b-a4b-it'; // Change this

Other free models you can use:

• @cf/meta/llama-3.1-8b-instruct — Fast, general purpose
• @cf/meta/llama-3.3-70b-instruct-fp8-fast — High quality
• @cf/mistralai/mistral-7b-instruct-v0.2 — 32k context
• @cf/zai-org/glm-4.7-flash — Multilingual, tool calling

---

📄 License

MIT — free to use, modify, and deploy.

---

Made with ❤️ by Nimesh Piyumal