Spaces:
Sleeping
Sleeping
Download llm_handler.py from abstikatech/AbstikaAI: direct link, hf CLI and curl.
- Browser
- Download file 2.25 kB
-
https://huggingface.co/spaces/abstikatech/AbstikaAI/resolve/main/llm_handler.py
- Command line
-
hf download hf://spaces/abstikatech/AbstikaAI/llm_handler.py
-
curl -L -o llm_handler.py https://huggingface.co/spaces/abstikatech/AbstikaAI/resolve/main/llm_handler.py
2.25 kB
| import httpx | |
| import os | |
| from pathlib import Path | |
| USER_SESSIONS = {} | |
| GLOBAL_CONTEXT_PATH = Path(__file__).parent / "global_context.txt" | |
| with open(GLOBAL_CONTEXT_PATH, "r", encoding="utf-8") as f: | |
| GLOBAL_PROMPT = f.read() | |
| HF_TOKEN = os.getenv("HF_TOKEN") | |
| HF_MODEL_ID = os.getenv("HF_MODEL_ID") | |
| def build_prompt(user_id: str, user_message: str) -> str: | |
| history = USER_SESSIONS.setdefault(user_id, []) | |
| history.append(f"User: {user_message}") | |
| history_text = "\n".join(history) | |
| full_prompt = f"""{GLOBAL_PROMPT} | |
| Conversation: | |
| {history_text} | |
| Assistant:""" | |
| return full_prompt | |
| async def call_hf_inference(prompt: str) -> str: | |
| if not HF_TOKEN or not HF_MODEL_ID: | |
| return "Hugging Face token or model ID not set A." | |
| url = f"https://api-inference.huggingface.co/models/{HF_MODEL_ID}" | |
| headers = { | |
| "Authorization": f"Bearer {HF_TOKEN}", | |
| "Content-Type": "application/json" | |
| } | |
| payload = {"inputs": prompt} | |
| async with httpx.AsyncClient() as client: | |
| try: | |
| response = await client.post(url, headers=headers, json=payload, timeout=60) | |
| response.raise_for_status() | |
| except httpx.HTTPStatusError as exc: | |
| print(f"HTTP error: {exc.response.status_code}") | |
| print(f"Response: {exc.response.text}") | |
| return "HTTP error" | |
| except httpx.RequestError as exc: | |
| print(f"Request error: {exc}") | |
| return "Request error" | |
| except Exception as exc: | |
| print(f"Other error: {exc}") | |
| return "Unknown error" | |
| data = response.json() | |
| # The response format may vary by model; adjust as needed | |
| if isinstance(data, list) and "generated_text" in data[0]: | |
| return data[0]["generated_text"].strip() | |
| elif isinstance(data, dict) and "error" in data: | |
| return f"Error: {data['error']}" | |
| else: | |
| return str(data) | |
| async def get_chat_response(user_id: str, user_message: str) -> str: | |
| # prompt = build_prompt(user_id, user_message) | |
| # If you want to use the prompt with history, uncomment the above line and use 'prompt' below | |
| answer = await call_hf_inference(user_message) | |
| # USER_SESSIONS[user_id].append(f"Assistant: {answer}") | |
| return answer | |