Overview
CheckThat integrates with Google’s Gemini models, providing access to multimodal AI capabilities with strong performance on reasoning, analysis, and generation tasks. Gemini models are designed for versatility and efficiency.
Available Models
The following Gemini models are available through CheckThat:
Gemini 2.5 Pro - Most capable model with superior performance on complex reasoning tasks
Gemini 2.5 Flash - Optimized for speed and efficiency while maintaining high quality
Configuration
API Key Setup
The model identifier from the available models list above.
Request Parameters
System instruction that sets the model’s behavior and context.
Array of content parts. Can include text and other modalities.[
{"role": "user", "parts": [{"text": "Hello!"}]},
{"role": "model", "parts": [{"text": "Hi there!"}]}
]
Controls randomness in responses. Range: 0.0 to 2.0.
Expected MIME type of response (e.g., application/json for structured outputs).
JSON schema for structured output generation.
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": "gemini-2.5-pro",
"provider": "gemini",
"gemini_api_key": "YOUR_GEMINI_API_KEY",
"messages": [
{"role": "user", "content": "Explain machine learning in simple terms."}
],
"system_prompt": "You are a helpful AI assistant."
}
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": "gemini-2.5-flash",
"provider": "gemini",
"gemini_api_key": "YOUR_GEMINI_API_KEY",
"messages": [
{"role": "user", "content": "Write a creative story about space exploration."}
],
"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'))
Structured Output with JSON Schema
import requests
url = "https://api.checkthat.ai/v1/chat/completions"
headers = {
"Authorization": "Bearer YOUR_CHECKTHAT_API_KEY",
"Content-Type": "application/json"
}
schema = {
"type": "object",
"properties": {
"topic": {"type": "string"},
"key_points": {
"type": "array",
"items": {"type": "string"}
},
"complexity": {
"type": "string",
"enum": ["simple", "moderate", "advanced"]
}
},
"required": ["topic", "key_points"]
}
payload = {
"model": "gemini-2.5-pro",
"provider": "gemini",
"gemini_api_key": "YOUR_GEMINI_API_KEY",
"messages": [
{"role": "user", "content": "Summarize the key concepts of quantum computing."}
],
"response_format": {
"type": "json_schema",
"json_schema": {
"name": "topic_summary",
"schema": schema
}
}
}
response = requests.post(url, json=payload, headers=headers)
result = response.json()
print(result)
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": "gemini-2.5-pro",
"provider": "gemini",
"gemini_api_key": "YOUR_GEMINI_API_KEY",
"messages": [
{"role": "user", "content": "What is photosynthesis?"},
{"role": "assistant", "content": "Photosynthesis is the process by which plants convert light energy into chemical energy."},
{"role": "user", "content": "What are the main steps involved?"}
]
}
response = requests.post(url, json=payload, headers=headers)
print(response.json())
Features and Capabilities
Structured Output Support
Gemini models have excellent support for structured outputs using JSON schemas (gemini.py:81-146):
Full JSON Schema Support:
response = client.models.generate_content(
model=model,
contents=contents,
config=types.GenerateContentConfig(
system_instruction=system_instruction,
response_mime_type='application/json',
response_schema=schema
)
)
- gemini-2.5-pro
- gemini-2.5-flash
- gemini-1.5-pro (legacy)
- gemini-1.5-flash (legacy)
Conversation History Management
CheckThat automatically formats conversation history for Gemini’s API (gemini.py:38-43):
- Converts messages to Gemini’s
Part format
- Maintains system instructions separately
- Handles multi-turn conversations seamlessly
Streaming Support
Real-time streaming with native Gemini streaming API (gemini.py:57-79):
stream = client.models.generate_content_stream(
model=model,
contents=contents,
config=types.GenerateContentConfig(
system_instruction=system_instruction
)
)
for chunk in stream:
if hasattr(chunk, 'text') and chunk.text:
yield chunk.text
gemini.py:148-244):
- Extracts content from candidates
- Maps finish reasons (STOP, MAX_TOKENS, SAFETY, etc.)
- Includes usage metadata (prompt_token_count, candidates_token_count)
- Preserves safety ratings for content filtering
Implementation Details
CheckThat’s Gemini integration (gemini.py:19-244) provides:
- Native Google GenAI client: Uses official
google.genai SDK
- Full structured output support: JSON schema with response validation
- Safety ratings: Includes content safety information in responses
- Response transformation: Converts to OpenAI-compatible format
Structured Response Object
For JSON schema responses, CheckThat returns a StructuredResponse object:
class StructuredResponse:
def __init__(self, content: str, parsed: Any):
self.content = content # Raw JSON string
self.parsed = parsed # Parsed Python object
Safety Ratings
Gemini responses include safety ratings for content filtering:
"gemini_safety_ratings": [
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
}
]
Rate Limits and Pricing
Rate limits and pricing are determined by your Google AI API tier. CheckThat does not impose additional limits.
Refer to Google AI pricing for current rates:
- Gemini 2.5 Pro: Premium pricing for complex tasks
- Gemini 2.5 Flash: Optimized pricing for high-volume use
Free Tier
Google AI offers a generous free tier:
- 15 requests per minute
- 1 million tokens per minute
- 1,500 requests per day
Error Handling
try:
response = requests.post(url, json=payload, headers=headers)
response.raise_for_status()
result = response.json()
# Check for safety filters
if 'gemini_safety_ratings' in result:
for rating in result['gemini_safety_ratings']:
if rating['probability'] not in ['NEGLIGIBLE', 'LOW']:
print(f"Safety warning: {rating['category']}")
except requests.exceptions.HTTPError as e:
print(f"API Error: {e}")
print(f"Response: {e.response.text}")
except ValueError as e:
print(f"Invalid API key or configuration: {e}")
- 400: Invalid request or API key required
- 429: Rate limit exceeded
- 500: Gemini service error
Best Practices
- Choose the right model: Use Flash for speed, Pro for complex reasoning
- Leverage JSON schema: Take advantage of native structured output support
- Monitor safety ratings: Check content safety in responses
- Use streaming: Enable streaming for better user experience
- Handle API key errors: Gemini requires API key; validate before requests
- Optimize for free tier: Stay within rate limits for cost-effective use
- System instructions: Use clear system instructions for consistent behavior
