Documentation Index
Fetch the complete documentation index at: https://docs.helicone.ai/llms.txt
Use this file to discover all available pages before exploring further.
Web Search Overview
Helicone Gateway supports web search for Anthropic models, allowing Claude to search the internet and provide up-to-date information with citations. This feature is enabled by appending :online to the model name, following the same pattern as OpenRouter.
How it Works
When you append :online to an Anthropic model name (e.g., claude-3-5-sonnet-20241022:online), Helicone automatically:
- Enables the web search tool for the request
- Routes the request to Anthropic with the appropriate web search configuration
- Returns the response with citations formatted as annotations
Quick Start
import openai
client = openai.OpenAI(
api_key="YOUR_ANTHROPIC_API_KEY",
base_url="https://gateway.helicone.ai/v1",
default_headers={
"Helicone-Auth": "Bearer YOUR_HELICONE_API_KEY",
"Helicone-Target-Url": "https://api.anthropic.com",
}
)
response = client.chat.completions.create(
model="claude-3-5-sonnet-20241022:online", # Note the :online suffix
messages=[
{"role": "user", "content": "What are the latest developments in AI?"}
]
)
# Access citations if available
if response.choices[0].message.annotations:
for annotation in response.choices[0].message.annotations:
print(f"Source: {annotation['url_citation']['title']}")
print(f"URL: {annotation['url_citation']['url']}")
import OpenAI from 'openai';
const openai = new OpenAI({
apiKey: 'YOUR_ANTHROPIC_API_KEY',
baseURL: 'https://gateway.helicone.ai/v1',
defaultHeaders: {
'Helicone-Auth': 'Bearer YOUR_HELICONE_API_KEY',
'Helicone-Target-Url': 'https://api.anthropic.com',
}
});
const response = await openai.chat.completions.create({
model: 'claude-3-5-sonnet-20241022:online', // Note the :online suffix
messages: [
{ role: 'user', content: 'What are the latest developments in AI?' }
]
});
// Access citations if available
if (response.choices[0].message.annotations) {
response.choices[0].message.annotations.forEach(annotation => {
console.log(`Source: ${annotation.url_citation.title}`);
console.log(`URL: ${annotation.url_citation.url}`);
});
}
curl https://gateway.helicone.ai/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_ANTHROPIC_API_KEY" \
-H "Helicone-Auth: Bearer YOUR_HELICONE_API_KEY" \
-H "Helicone-Target-Url: https://api.anthropic.com" \
-d '{
"model": "claude-3-5-sonnet-20241022:online",
"messages": [
{
"role": "user",
"content": "What are the latest developments in AI?"
}
]
}'
Advanced Configuration
You can customize the web search behavior by including a plugins parameter in your request body:
{
"model": "claude-3-5-sonnet-20241022:online",
"messages": [...],
"plugins": [
{
"id": "web",
"max_uses": 5,
"allowed_domains": ["wikipedia.org", "arxiv.org"],
"blocked_domains": ["example.com"],
"user_location": {
"type": "approximate",
"country": "US"
}
}
]
}
Plugin Options
| Parameter | Type | Description |
|---|
max_uses | integer | Maximum number of web searches allowed per request (default: unlimited) |
allowed_domains | array | Restrict searches to specific domains |
blocked_domains | array | Exclude specific domains from search results |
user_location | object | Provide approximate user location for localized results |
{
"choices": [
{
"message": {
"role": "assistant",
"content": "Based on recent developments, AI has made significant progress in...",
"annotations": [
{
"type": "url_citation",
"url_citation": {
"url": "https://example.com/article",
"title": "Recent AI Breakthroughs",
"content": "The source text that was cited...",
"start_index": 0,
"end_index": 67
}
}
]
}
}
]
}
Understanding Annotations
url: The source URL of the cited informationtitle: The title of the source pagecontent: The relevant excerpt from the sourcestart_index: Character position where the citation begins in the responseend_index: Character position where the citation ends in the response
Pricing
Web search requests are billed at standard Anthropic rates plus any additional costs for web search usage. The usage statistics include:
{
"usage": {
"input_tokens": 1000,
"output_tokens": 500,
"server_tool_use": {
"web_search_requests": 1
}
}
}
