Skip to main content

​
Overview

CheckThat integrates with Anthropic’s Claude models, providing access to advanced AI capabilities with strong reasoning, analysis, and conversation abilities. Claude models excel at nuanced understanding, detailed explanations, and maintaining context across long conversations.

​
Available Models

The following Claude models are available through CheckThat:
​
claude-sonnet-4-20250514
string
Claude Sonnet 4 - Balanced performance with excellent reasoning capabilities and speed
​
claude-opus-4-1-20250805
string
Claude Opus 4.1 - Most capable Claude model with superior performance on complex tasks

​
Configuration

​
API Key Setup

​
api_key
string
required
Your Anthropic API key. Get your key from Anthropic Console.
​
model
string
required
The model identifier from the available models list above.

​
Request Parameters

​
system
string
System prompt that sets the assistant’s behavior and context.
​
messages
array
required
Array of message objects. Must alternate between user and assistant roles.
[
  {"role": "user", "content": "Hello!"},
  {"role": "assistant", "content": "Hi! How can I help?"},
  {"role": "user", "content": "Tell me about AI."}
]
​
max_tokens
integer
default:"8192"
Maximum number of tokens to generate. Claude models default to 8192 tokens.
​
temperature
number
default:"1.0"
Controls randomness in responses. Range: 0.0 to 1.0.
​
stream
boolean
default:"false"
Enable streaming responses for real-time output.

​
Usage Examples

​
Basic Chat Completion

import requests

url = "https://api.checkthat.ai/v1/chat/completions"
headers = {
    "Authorization": "Bearer YOUR_CHECKTHAT_API_KEY",
    "Content-Type": "application/json"
}

payload = {
    "model": "claude-sonnet-4-20250514",
    "provider": "anthropic",
    "anthropic_api_key": "YOUR_ANTHROPIC_API_KEY",
    "messages": [
        {"role": "user", "content": "Explain the concept of neural networks."}
    ],
    "system_prompt": "You are an expert AI educator who explains complex topics clearly."
}

response = requests.post(url, json=payload, headers=headers)
print(response.json())

​
Streaming Response

import requests

url = "https://api.checkthat.ai/v1/chat/completions"
headers = {
    "Authorization": "Bearer YOUR_CHECKTHAT_API_KEY",
    "Content-Type": "application/json"
}

payload = {
    "model": "claude-opus-4-1-20250805",
    "provider": "anthropic",
    "anthropic_api_key": "YOUR_ANTHROPIC_API_KEY",
    "messages": [
        {"role": "user", "content": "Write a detailed analysis of renewable energy trends."}
    ],
    "stream": True
}

with requests.post(url, json=payload, headers=headers, stream=True) as response:
    for line in response.iter_lines():
        if line:
            print(line.decode('utf-8'))

​
Multi-turn Conversation

import requests

url = "https://api.checkthat.ai/v1/chat/completions"
headers = {
    "Authorization": "Bearer YOUR_CHECKTHAT_API_KEY",
    "Content-Type": "application/json"
}

payload = {
    "model": "claude-sonnet-4-20250514",
    "provider": "anthropic",
    "anthropic_api_key": "YOUR_ANTHROPIC_API_KEY",
    "messages": [
        {"role": "user", "content": "What are the main types of machine learning?"},
        {"role": "assistant", "content": "The three main types are: supervised learning, unsupervised learning, and reinforcement learning."},
        {"role": "user", "content": "Can you explain supervised learning in detail?"}
    ]
}

response = requests.post(url, json=payload, headers=headers)
print(response.json())

​
Structured Output (Limited Support)

import requests

url = "https://api.checkthat.ai/v1/chat/completions"
headers = {
    "Authorization": "Bearer YOUR_CHECKTHAT_API_KEY",
    "Content-Type": "application/json"
}

# Note: Anthropic structured outputs currently use instructor library
# and support Pydantic models only (not JSON schema)
payload = {
    "model": "claude-sonnet-4-20250514",
    "provider": "anthropic",
    "anthropic_api_key": "YOUR_ANTHROPIC_API_KEY",
    "messages": [
        {"role": "user", "content": "Analyze this claim: 'Electric vehicles reduce carbon emissions.'"}
    ],
    "system_prompt": "You are a fact-checker. Analyze claims and provide structured responses."
}

response = requests.post(url, json=payload, headers=headers)
print(response.json())

​
Features and Capabilities

​
Long Context Window

Claude models support extensive context windows, making them ideal for:
  • Long document analysis
  • Extended conversations
  • Multi-turn dialogues with rich history

​
Conversation History Management

CheckThat automatically formats conversation history for Anthropic’s API (anthropic.py:38-42):
  • Alternates user/assistant messages correctly
  • Extracts system instructions properly
  • Maintains conversation context across turns

​
Streaming Support

Real-time streaming with Anthropic’s native streaming API (anthropic.py:69-74): 
with client.messages.stream(
    max_tokens=8192,
    model=model,
    system=system_instruction,
    messages=messages,
) as stream:
    for text in stream.text_stream:
        yield text

​
OpenAI-Compatible Response Format

CheckThat transforms Anthropic responses to OpenAI-compatible format (anthropic.py:131-278), including:
  • Standard message structure
  • Usage statistics (prompt_tokens, completion_tokens)
  • Finish reason mapping
  • Stop sequence handling

​
Structured Outputs

​
Current Limitations

Anthropic structured outputs in CheckThat have limited support:
  • Pydantic models only: Currently supports Pydantic model formats via instructor library
  • No JSON schema: Dict-based JSON schema formats not yet supported
  • Error handling: Returns 400 error for unsupported formats
Supported:
  • claude-sonnet-4-20250514 (with Pydantic models)
Implementation (anthropic.py:81-129): 
if isinstance(response_format, dict):
    raise HTTPException(
        status_code=400, 
        detail="Anthropic provider currently only supports Pydantic model response formats"
    )

​
Implementation Details

CheckThat’s Anthropic integration (anthropic.py:21-279) provides:
  • Native Anthropic client: Uses official anthropic Python SDK
  • Instructor integration: Structured outputs via instructor library
  • Response transformation: Converts to OpenAI-compatible format
  • Conversation formatting: Automatic message formatting for Anthropic API

​
Response Format Transformation

The _format_to_openai_response method ensures compatibility:
  • Extracts content from Anthropic’s content blocks
  • Maps finish reasons (end_turn → stop, max_tokens → length)
  • Preserves usage metadata (input_tokens → prompt_tokens)
  • Adds Anthropic-specific extensions for debugging

​
Rate Limits and Pricing

Rate limits and pricing are determined by your Anthropic API tier. CheckThat does not impose additional limits. Refer to Anthropic’s pricing page for current rates:
  • Claude Sonnet 4: Balanced pricing for production use
  • Claude Opus 4.1: Premium pricing for highest capability

​
Error Handling

try:
    response = requests.post(url, json=payload, headers=headers)
    response.raise_for_status()
    result = response.json()
except requests.exceptions.HTTPError as e:
    if e.response.status_code == 400:
        print(f"Bad request: {e.response.json()}")
    elif e.response.status_code == 429:
        print("Rate limit exceeded")
    else:
        print(f"API Error: {e}")
except Exception as e:
    print(f"Request failed: {e}")
Common error codes:
  • 400: Invalid request or unsupported structured output format
  • 401: Invalid API key
  • 429: Rate limit exceeded
  • 500: Anthropic service error

​
Best Practices

  1. Use system prompts effectively: Set clear context and behavior expectations
  2. Alternate messages properly: Ensure user/assistant message alternation
  3. Leverage long context: Claude excels at analyzing long documents
  4. Stream for UX: Enable streaming for better user experience
  5. Handle max_tokens: Default is 8192; adjust based on needs
  6. Monitor usage: Track token usage for cost optimization
  7. Structured outputs: Use Pydantic models when structured data is needed

Build docs developers (and LLMs) love

Get started for freeTalk to us