app.py

import logging

from fastapi import FastAPI
from llama_index.core.memory import ChatMemoryBuffer
from llama_index.llms.llama_cpp import LlamaCPP
from transformers import AutoTokenizer
from llama_index.core import set_global_tokenizer
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.core import SimpleDirectoryReader
from llama_index.core import VectorStoreIndex


logging.basicConfig(
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    level=logging.INFO
)
logger = logging.getLogger(__name__)

logger.info("Запускаемся... 🥳🥳🥳")

app = FastAPI()

model_url = "https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF/resolve/main/qwen2.5-7b-instruct-q3_k_m.gguf"

SYSTEM_PROMPT = ''

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B-Instruct")

set_global_tokenizer(
    AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B-Instruct").encode
)

# embed_model = HuggingFaceEmbedding(model_name="sentence-transformers/all-mpnet-base-v2")
#
# documents = SimpleDirectoryReader("./data/").load_data()

def messages_to_prompt(messages):
    messages = [{"role": m.role.value, "content": m.content} for m in messages]
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    return prompt


def completion_to_prompt(completion):
    messages = [
        {"role": "system", "content": SYSTEM_PROMPT or "Ответ должен быть точным, кратким и с юмором."},
        {"role": "user", "content": completion},
    ]
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    return prompt


llm = LlamaCPP(
    # You can pass in the URL to a GGML model to download it automatically
    model_url=model_url,
    # optionally, you can set the path to a pre-downloaded model instead of model_url
    model_path=None,
    temperature=0.1,
    max_new_tokens=256,
    # llama2 has a context window of 4096 tokens, but we set it lower to allow for some wiggle room
    context_window=2046,
    # kwargs to pass to __call__()
    generate_kwargs={},
    # kwargs to pass to __init__()
    # set to at least 1 to use GPU
    model_kwargs={"n_gpu_layers": -1, "num_return_sequences": 1, "no_repeat_ngram_size": 2, "n_threads": 2},
    # transform inputs into Llama2 format
    messages_to_prompt=messages_to_prompt,
    completion_to_prompt=completion_to_prompt,
    verbose=True,
)

memory = ChatMemoryBuffer.from_defaults(token_limit=3900)
index = VectorStoreIndex.from_documents({})
chat_engine = index.as_chat_engine(
    chat_mode="context",
    memory=memory,
    llm=llm,
    context_propt=(
        "Вы - чат-бот, способный нормально взаимодействовать.\n"
        "Используйте предыдущую историю чата или приведенный выше контекст, чтобы взаимодействовать с пользователем и помогать ему."
    )
)

def generate_response(completion_response):
    try:
        response_text = completion_response.text
        return response_text.strip() if response_text else "Пустой ответ"
    except Exception as e:
        logger.error(f"Ошибка обработки ответа: {str(e)}")
        return "Ошибка генерации"


@app.get("/")
def greet_json():
    return {"Hello": "World!"}

@app.put("/system-prompt")
async def set_system_prompt(text: str):
    logger.info('post/system-prompt')
    # global SYSTEM_PROMPT
    # SYSTEM_PROMPT = text

@app.post("/predict")
async def predict(text: str):
    # Генерация ответа с помощью модели
    logger.info('post/predict')
    logger.info('ЗАПРОС:')
    logger.info(text)
    # response = llm.complete(text)
    response = chat_engine.chat(text)
    logger.info('ОТВЕТ:')
    logger.info(response)
    # text_response = generate_response(response)
    # return {"response": text_response}
    return {"response": response}