Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.lemondata.cc/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

LemonData supports three native API formats with a single API key. Choose the format that best fits your use case - no configuration changes needed.

OpenAI Format

/v1/chat/completions Standard format, widest compatibility

Anthropic Format

/v1/messages Extended thinking, native Claude features

Gemini Format

/v1beta/models/:model:generateContent Google ecosystem integration

​
Why Multi-Format?

BenefitDescription
No SDK switchingUse any model with your preferred SDK
Native featuresAccess format-specific capabilities
Easy migrationSwitch from official APIs with just a base URL change
Single billingOne account, one API key, all formats

​
Format Comparison

FeatureOpenAIAnthropicGemini
Endpoint/v1/chat/completions/v1/messages/v1beta/models/:model:generateContent
Auth HeaderAuthorization: Bearerx-api-keyAuthorization: Bearer
System PromptIn messages arraySeparate system fieldIn systemInstruction
Extended Thinking❌✅❌
Streaming✅ SSE✅ SSE✅ SSE
Tool Calling✅✅✅
Vision✅✅✅

​
OpenAI Format

The most widely compatible format and the default starting point for most new integrations. Works across the broadest set of LemonData models. 
from openai import OpenAI

client = OpenAI(
    api_key="sk-your-lemondata-key",
    base_url="https://api.lemondata.cc/v1"
)

# Works with ANY model
response = client.chat.completions.create(
    model="claude-sonnet-4-6",  # Claude via OpenAI format
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"}
    ]
)
Best for:
  • General use
  • Existing OpenAI SDK integrations
  • Maximum compatibility

​
Anthropic Format

Native Anthropic Messages API. Required for Claude-specific features like extended thinking. 
from anthropic import Anthropic

client = Anthropic(
    api_key="sk-your-lemondata-key",
    base_url="https://api.lemondata.cc"  # No /v1 suffix!
)

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    system="You are a helpful assistant.",  # Separate system field
    messages=[
        {"role": "user", "content": "Hello!"}
    ]
)

​
Extended Thinking (Claude Opus 4.6)

Only available in Anthropic format: 
message = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 10000
    },
    messages=[{"role": "user", "content": "Solve this complex problem..."}]
)

# Access thinking process
for block in message.content:
    if block.type == "thinking":
        print(f"Thinking: {block.thinking}")
    elif block.type == "text":
        print(f"Answer: {block.text}")
Best for:
  • Claude-specific features
  • Extended thinking mode
  • Native Anthropic SDK users

​
Gemini Format

Native Google Gemini API format for Google ecosystem integration. 
curl "https://api.lemondata.cc/v1beta/models/gemini-2.5-flash:generateContent" \
  -H "Authorization: Bearer sk-your-lemondata-key" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "parts": [{"text": "Hello!"}]
    }],
    "systemInstruction": {
      "parts": [{"text": "You are a helpful assistant."}]
    }
  }'

​
Streaming

curl "https://api.lemondata.cc/v1beta/models/gemini-2.5-flash:streamGenerateContent?alt=sse" \
  -H "Authorization: Bearer sk-your-lemondata-key" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"parts": [{"text": "Write a story"}]}]
  }'
Best for:
  • Google Cloud integrations
  • Existing Gemini SDK code
  • Native Gemini features
Gemini Files and Cache: /upload/v1beta/files, /v1beta/files, /v1beta/files:register, and /v1beta/cachedContents are available on the native Gemini route. Files use Gemini File API-compatible upstream channels; explicit cache resources can also route through Vertex AI channels. Resources created through LemonData are bound to the same upstream channel/key for later generateContent calls.

​
Tool Compatibility Boundary

Function tools can be converted between formats when the target route supports them. Provider-native tools must stay on their native route:
  • OpenAI Responses hosted and native tools such as tool_search, web_search, file_search, code_interpreter, MCP, shell/apply_patch, and computer-use tools require /v1/responses.
  • Anthropic server/native tools such as web_search_*, web_fetch_*, code_execution_*, tool_search_*, bash, computer-use, and text-editor tools require /v1/messages.
  • Gemini built-in tools such as googleSearch, codeExecution, urlContext, computerUse, and similar tools fields require /v1beta.
If LemonData cannot route a request with native tools to a native-capable upstream path, it returns an explicit unsupported-field error instead of dropping the tool or pretending it is a Chat Completions function. User-defined function tools remain the portable path.

​
Choosing the Right Format

​
Migration Guides

​
From OpenAI Official API

# Before (OpenAI)
client = OpenAI(api_key="sk-openai-key")

# After (LemonData)
client = OpenAI(
    api_key="sk-lemondata-key",
    base_url="https://api.lemondata.cc/v1"  # Add this line
)
# That's it! Same code works

​
From Anthropic Official API

# Before (Anthropic)
client = Anthropic(api_key="sk-ant-key")

# After (LemonData)
client = Anthropic(
    api_key="sk-lemondata-key",
    base_url="https://api.lemondata.cc"  # Add this line (no /v1!)
)

​
From Google AI Studio

# Before (Google)
import google.generativeai as genai
genai.configure(api_key="google-api-key")

# After (LemonData) - Use REST API
import requests

response = requests.post(
    "https://api.lemondata.cc/v1beta/models/gemini-2.5-flash:generateContent",
    headers={"Authorization": "Bearer sk-lemondata-key"},
    json={"contents": [{"parts": [{"text": "Hello"}]}]}
)

​
Cross-Model Compatibility

The magic of LemonData: use any SDK with any model. The gateway automatically handles format conversion.

​
Any SDK → Any Model

# Anthropic SDK with GPT-4o (auto-converts to OpenAI format)
from anthropic import Anthropic

client = Anthropic(
    api_key="sk-lemondata-key",
    base_url="https://api.lemondata.cc"
)

response = client.messages.create(
    model="gpt-4o",  # ✅ Works! Auto-converted
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}]
)

# Same SDK, different models - no code changes
response = client.messages.create(model="gemini-2.5-flash", ...)  # ✅ Works!
response = client.messages.create(model="deepseek-r1", ...)       # ✅ Works!

​
OpenAI SDK → All Models

from openai import OpenAI

client = OpenAI(base_url="https://api.lemondata.cc/v1", api_key="sk-...")

# All these work with the same SDK:
response = client.chat.completions.create(model="gpt-4o", ...)
response = client.chat.completions.create(model="claude-sonnet-4-6", ...)
response = client.chat.completions.create(model="gemini-2.5-flash", ...)

​
Industry Comparison

PlatformOpenAI FormatAnthropic FormatGemini FormatResponses API
LemonData✅ All models✅ All models✅ All models✅ All models
OpenRouter✅ All models❌❌❌
Together AI✅ All models❌❌❌
Fireworks✅ All models❌❌❌
While cross-format works for most features, format-specific features (like Anthropic extended thinking) require the native format.