File size: 4,354 Bytes
95a5cb0 2e3dc8a 95a5cb0 2e3dc8a 95a5cb0 d69ddeb 95a5cb0 d69ddeb 95a5cb0 d69ddeb 95a5cb0 2e3dc8a 95a5cb0 bcabbeb 95a5cb0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 | import asyncio
from helper import MODEL_NAME, AUTO_ANSWERS, ROLE_ASSISTANT
from typing import List
from huggingface_hub import InferenceClient
import os
class RAGModel:
""" Class for model related functions, such as loading the model, API/model interaction and such."""
def __init__(self, api_key: str, model_name: str = MODEL_NAME):
self.api_key = api_key
self.model_name = model_name
self.client = InferenceClient(provider="nebius", token=api_key)
# Backup (maybe)
#try:
# self.tokenizer = AutoTokenizer.from_pretrained(model_name)
#except Exception as e:
# print(f"Lokaler Tokenizer nicht verfügbar: {e}")
# self.tokenizer = None
def get_auto_answer(self, reason: str):
return f"Apologies: {reason}"
def generate_response(self,
query: str,
context: List[str],
temperature: float = 0.7,
max_tokens: int = 512,
top_p: float = 0.9,
top_k: int = None) -> str:
"""Generate an answer via Hugging Face API with configurable parameters
Args:
query: User input string
context: Context retrieved from context DB
temperature: Sampling temperature (0.0 to 2.0)
max_tokens: Maximum tokens to generate
top_p: Nucleus sampling parameter
top_k: Top-k sampling parameter (not used in HF API)
"""
if len(context) >= 10:
context_text = "\n".join(context[:10]) # limit to 10 most relevant chunks
else:
context_text = "\n".join(context)
UNIVERSITY_ASSISTANT_SYSTEM_PROMPT = """
You are a university assistant that helps ONLY with university-related topics using the available database.
## APPROPRIATE RESPONSES
You can help with:
- "What courses is [student] taking?"
- "Who teaches [course]?"
- "Which students are in [professor]'s class?"
- "What is [professor]'s email?"
- "What department is [faculty] in?"
## STRICT BOUNDARIES
Never discuss:
- Grades, academic performance, or GPA
- Financial information, tuition, or payments
- Sensitive student data beyond basic directory info
- Any non-university topics (medical, legal, financial advice)
## RESPONSE STYLE
- Be helpful and professional
- Redirect inappropriate requests: "I can only help with university academic topics"
- For sensitive data: "I don't have access to that information. Please contact [relevant office]"
- Only share information appropriate for academic purposes
## KNOWLEDGE USAGE
Use the provided Context below to answer user requests.
"""
prompt = f"{UNIVERSITY_ASSISTANT_SYSTEM_PROMPT} \n\nContext: {context_text}\nQuestion: {query}\nAnswer: Based on the given context,"
try:
response = self.client.chat.completions.create(
model=self.model_name,
messages=[
{
"role": ROLE_ASSISTANT,
"content": [
{
"type": "text",
"text": prompt
},
]
}
],
max_tokens=max_tokens, # Use configurable parameter
temperature=temperature, # Use configurable parameter
top_p=top_p, # Use configurable parameter
)
if response:
return response.choices[0].message.content
else:
print(f"NO RESPONSE")
return self.get_auto_answer(AUTO_ANSWERS.COULD_NOT_GENERATE.value)
except asyncio.TimeoutError:
print("API Timeout")
return self.get_auto_answer(AUTO_ANSWERS.REQUEST_TIMED_OUT.value)
except Exception as e:
print(f"API Fehler: {e}")
return self.get_auto_answer(AUTO_ANSWERS.UNEXPECTED_ERROR.value)
|