NLP-RAG / models /qwen_3_5_9b.py
Qar-Raz's picture
Sync backend Docker context from GitHub main
c64aaec verified
import os
from openai import OpenAI
class Qwen_3_5_9B:
def __init__(self, token=None):
openrouter_token = os.getenv("OPENROUTER_API_KEY")
if not openrouter_token:
print("Warning: OPENROUTER_API_KEY environment variable is not set")
# Fallback to the token passed in if available
openrouter_token = token
self.client = OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=openrouter_token,
)
self.model_id = "qwen/qwen3.5-9b"
def generate_stream(self, prompt, max_tokens=100000, temperature=0.1):
for message in self.client.chat.completions.create(
model=self.model_id,
messages=[{"role": "user", "content": prompt}],
max_tokens=max_tokens,
temperature=temperature,
stream=True,
):
if message.choices and len(message.choices) > 0:
content = message.choices[0].delta.content
if content:
yield content
def generate(self, prompt, max_tokens=100000, temperature=0.1):
return "".join(self.generate_stream(prompt, max_tokens=max_tokens, temperature=temperature))