72 lines
2.7 KiB
Python
72 lines
2.7 KiB
Python
"""DeepSeek API client - OpenAI-compatible, zero extra dependencies."""
|
|
|
|
import json
|
|
import logging
|
|
import time
|
|
import requests
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class DeepSeekClient:
|
|
"""Client for DeepSeek API (OpenAI-compatible format)."""
|
|
|
|
BASE_URL = "https://api.deepseek.com/v1/chat/completions"
|
|
|
|
def __init__(self, api_key, model="deepseek-chat", temperature=0.7, max_retries=3):
|
|
self.api_key = api_key
|
|
self.model = model
|
|
self.temperature = temperature
|
|
self.max_retries = max_retries
|
|
self.session = requests.Session()
|
|
self.session.headers.update({
|
|
"Authorization": f"Bearer {self.api_key}",
|
|
"Content-Type": "application/json"
|
|
})
|
|
|
|
def ask(self, system_prompt, user_prompt, temperature=None):
|
|
"""Send a chat completion request. Returns response text."""
|
|
messages = [
|
|
{"role": "system", "content": system_prompt},
|
|
{"role": "user", "content": user_prompt}
|
|
]
|
|
payload = {
|
|
"model": self.model,
|
|
"messages": messages,
|
|
"temperature": temperature or self.temperature
|
|
}
|
|
|
|
last_error = None
|
|
for attempt in range(self.max_retries):
|
|
try:
|
|
resp = self.session.post(self.BASE_URL, json=payload, timeout=60)
|
|
resp.raise_for_status()
|
|
data = resp.json()
|
|
result = data["choices"][0]["message"]["content"]
|
|
logger.info(
|
|
"DeepSeek API OK | model=%s | input=%d chars | output=%d chars",
|
|
self.model, len(user_prompt), len(result)
|
|
)
|
|
return result
|
|
|
|
except requests.exceptions.Timeout:
|
|
last_error = "Timeout"
|
|
logger.warning("Attempt %d/%d timed out", attempt + 1, self.max_retries)
|
|
except requests.exceptions.HTTPError as e:
|
|
last_error = str(e)
|
|
logger.warning("Attempt %d/%d HTTP error: %s", attempt + 1, self.max_retries, e)
|
|
if "insufficient_quota" in str(e).lower() or "invalid_api_key" in str(e).lower():
|
|
raise # don't retry these
|
|
except (json.JSONDecodeError, KeyError) as e:
|
|
last_error = f"Parse error: {e}"
|
|
logger.warning("Attempt %d/%d parse error: %s", attempt + 1, self.max_retries, e)
|
|
|
|
if attempt < self.max_retries - 1:
|
|
sleep_time = 2 ** attempt
|
|
logger.info("Retrying in %ds...", sleep_time)
|
|
time.sleep(sleep_time)
|
|
|
|
raise RuntimeError(
|
|
f"DeepSeek API failed after {self.max_retries} retries. Last error: {last_error}"
|
|
)
|