mirror of
https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools.git
synced 2026-06-18 15:29:36 +00:00
123 lines
4.2 KiB
Python
123 lines
4.2 KiB
Python
"""
|
|
Anthropic Claude client.
|
|
عميل Claude.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import Any
|
|
|
|
import httpx
|
|
from tenacity import retry, retry_if_exception_type, stop_after_attempt, wait_exponential
|
|
|
|
from core.llm.base import LLMClient, LLMResponse, Message
|
|
|
|
|
|
class AnthropicClient(LLMClient):
|
|
"""Anthropic Claude API client."""
|
|
|
|
provider_name = "anthropic"
|
|
API_URL = "https://api.anthropic.com/v1/messages"
|
|
API_VERSION = "2023-06-01"
|
|
|
|
def __init__(
|
|
self,
|
|
api_key: str,
|
|
model: str = "claude-sonnet-4-5-20250929",
|
|
base_url: str | None = None,
|
|
timeout: int = 60,
|
|
) -> None:
|
|
super().__init__(api_key=api_key, model=model, base_url=base_url, timeout=timeout)
|
|
|
|
# Min tokens to trigger prompt caching (Anthropic requires >=1024 for Sonnet).
|
|
CACHE_MIN_TOKENS: int = 1024
|
|
# Rough heuristic: 1 token ≈ 4 chars (Arabic slightly higher, still safe).
|
|
CACHE_MIN_CHARS: int = 4 * CACHE_MIN_TOKENS
|
|
|
|
@retry(
|
|
stop=stop_after_attempt(3),
|
|
wait=wait_exponential(multiplier=1, min=2, max=10),
|
|
retry=retry_if_exception_type((httpx.TimeoutException, httpx.HTTPStatusError)),
|
|
reraise=True,
|
|
)
|
|
async def chat(
|
|
self,
|
|
messages: list[Message],
|
|
*,
|
|
max_tokens: int = 4096,
|
|
temperature: float = 0.7,
|
|
system: str | None = None,
|
|
cache_system: bool = True,
|
|
**kwargs: Any,
|
|
) -> LLMResponse:
|
|
"""Send chat completion to Anthropic API with optional prompt caching.
|
|
|
|
When ``cache_system=True`` (default) and the system prompt is long enough,
|
|
the system field is sent as a cache-enabled content block:
|
|
|
|
system = [{"type": "text", "text": PROMPT,
|
|
"cache_control": {"type": "ephemeral"}}]
|
|
|
|
Anthropic keeps cached prompts for ~5 minutes; subsequent calls with the
|
|
same prefix are billed at $0.30/mtok instead of $3/mtok (90% savings).
|
|
"""
|
|
# Separate system from messages (Anthropic API convention)
|
|
clean_messages: list[dict[str, str]] = []
|
|
extracted_system: str | None = system
|
|
|
|
for msg in messages:
|
|
if msg.role == "system" and extracted_system is None:
|
|
extracted_system = msg.content
|
|
else:
|
|
clean_messages.append(msg.to_dict())
|
|
|
|
payload: dict[str, Any] = {
|
|
"model": self.model,
|
|
"max_tokens": max_tokens,
|
|
"temperature": temperature,
|
|
"messages": clean_messages,
|
|
}
|
|
if extracted_system:
|
|
# Prompt-cache the system prompt when long enough. The system field
|
|
# accepts either a plain string or an array of content blocks.
|
|
if cache_system and len(extracted_system) >= self.CACHE_MIN_CHARS:
|
|
payload["system"] = [
|
|
{
|
|
"type": "text",
|
|
"text": extracted_system,
|
|
"cache_control": {"type": "ephemeral"},
|
|
}
|
|
]
|
|
else:
|
|
payload["system"] = extracted_system
|
|
|
|
headers = {
|
|
"x-api-key": self.api_key,
|
|
"anthropic-version": self.API_VERSION,
|
|
"content-type": "application/json",
|
|
}
|
|
|
|
async with httpx.AsyncClient(timeout=self.timeout) as client:
|
|
response = await client.post(self.API_URL, json=payload, headers=headers)
|
|
response.raise_for_status()
|
|
data = response.json()
|
|
|
|
# Extract text from content blocks
|
|
content_blocks = data.get("content", [])
|
|
text = "".join(
|
|
block.get("text", "") for block in content_blocks if block.get("type") == "text"
|
|
)
|
|
|
|
usage = data.get("usage", {})
|
|
return LLMResponse(
|
|
content=text,
|
|
provider=self.provider_name,
|
|
model=data.get("model", self.model),
|
|
input_tokens=usage.get("input_tokens", 0),
|
|
output_tokens=usage.get("output_tokens", 0),
|
|
cached_tokens=usage.get("cache_read_input_tokens", 0)
|
|
+ usage.get("cache_creation_input_tokens", 0),
|
|
finish_reason=data.get("stop_reason"),
|
|
raw=data,
|
|
)
|