main.py

import os
import glob
import shutil
import subprocess
from typing import Any, Dict, List

from fastapi import FastAPI, HTTPException, UploadFile, WebSocket
from fastapi.staticfiles import StaticFiles

from pydantic import BaseModel

# import torch
from langchain.chains import RetrievalQA
from langchain.embeddings import HuggingFaceInstructEmbeddings
from langchain.prompts import PromptTemplate
from langchain.memory import ConversationBufferMemory
from langchain.callbacks.base import BaseCallbackHandler
from langchain.schema import LLMResult

# from langchain.embeddings import HuggingFaceEmbeddings
from load_models import load_model

# from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from langchain.vectorstores import Chroma

from constants import CHROMA_SETTINGS, EMBEDDING_MODEL_NAME, PERSIST_DIRECTORY, MODEL_ID, MODEL_BASENAME, PATH_NAME_SOURCE_DIRECTORY

class Predict(BaseModel):
    prompt: str

class Delete(BaseModel):
    filename: str

class MyCustomHandler(BaseCallbackHandler):
    def on_llm_new_token(self, token: str, **kwargs) -> None:
        print(f" token: {token}")

    async def on_llm_start(
        self, serialized: Dict[str, Any], prompts: List[str], **kwargs: Any
    ) -> None:
        class_name = serialized["name"]
        print("start")

    async def on_llm_end(self, response: LLMResult, **kwargs: Any) -> None:
        print("finish")

# if torch.backends.mps.is_available():
#     DEVICE_TYPE = "mps"
# elif torch.cuda.is_available():
#     DEVICE_TYPE = "cuda"
# else:
#     DEVICE_TYPE = "cpu"

DEVICE_TYPE = "cuda"
SHOW_SOURCES = True

EMBEDDINGS = HuggingFaceInstructEmbeddings(model_name=EMBEDDING_MODEL_NAME, model_kwargs={"device": DEVICE_TYPE})

# load the vectorstore
DB = Chroma(
    persist_directory=PERSIST_DIRECTORY,
    embedding_function=EMBEDDINGS,
    client_settings=CHROMA_SETTINGS,
)

RETRIEVER = DB.as_retriever()

LLM = load_model(device_type=DEVICE_TYPE, model_id=MODEL_ID, model_basename=MODEL_BASENAME, stream=True, callbacks=[MyCustomHandler()])

template = """you are a helpful, respectful and honest assistant. When answering questions, you should only use the documents provided.
You should only answer the topics that appear in these documents.
Always answer in the most helpful and reliable way possible, if you don't know the answer to a question, just say you don't know, don't try to make up an answer,
don't share false information. you should use no more than 15 sentences and all your answers should be as concise as possible.
Always say "Thank you for asking!" at the end of your answer.
Context: {history} \n {context}
Question: {question}
"""

memory = ConversationBufferMemory(input_key="question", memory_key="history")

QA_CHAIN_PROMPT = PromptTemplate(input_variables=["history", "context", "question"], template=template)

QA = RetrievalQA.from_chain_type(
    llm=LLM,
    chain_type="stuff",
    retriever=RETRIEVER,
    return_source_documents=SHOW_SOURCES,
    chain_type_kwargs={
        "prompt": QA_CHAIN_PROMPT,
        "memory": memory,
    },
)

app = FastAPI(title="homepage-app")
api_app = FastAPI(title="api app")

app.mount("/api", api_app, name="api")
app.mount("/", StaticFiles(directory="static",html = True), name="static")

@api_app.get("/training")
def run_ingest_route():
    global DB
    global RETRIEVER
    global QA

    try:
        if os.path.exists(PERSIST_DIRECTORY):
            try:
                shutil.rmtree(PERSIST_DIRECTORY)
            except OSError as e:
                raise HTTPException(status_code=500, detail=f"Error: {e.filename} - {e.strerror}.")
        else:
            raise HTTPException(status_code=500, detail="The directory does not exist")

        run_langest_commands = ["python", "ingest.py"]

        if DEVICE_TYPE == "cpu":
            run_langest_commands.append("--device_type")
            run_langest_commands.append(DEVICE_TYPE)

        result = subprocess.run(run_langest_commands, capture_output=True)

        if result.returncode != 0:
            raise HTTPException(status_code=400, detail="Script execution failed: {}")

        # load the vectorstore
        DB = Chroma(
            persist_directory=PERSIST_DIRECTORY,
            embedding_function=EMBEDDINGS,
            client_settings=CHROMA_SETTINGS,
        )

        RETRIEVER = DB.as_retriever()

        QA = RetrievalQA.from_chain_type(
            llm=LLM,
            chain_type="stuff",
            retriever=RETRIEVER,
            return_source_documents=SHOW_SOURCES,
            chain_type_kwargs={
                "prompt": QA_CHAIN_PROMPT,
                "memory": memory
            },
        )


        return {"response": "The training was successfully completed"}
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Error occurred: {str(e)}")

@api_app.get("/api/files")
def get_files():
    upload_dir = os.path.join(os.getcwd(), PATH_NAME_SOURCE_DIRECTORY)
    files = glob.glob(os.path.join(upload_dir, '*'))

    return {"directory": upload_dir, "files": files}

@api_app.delete("/api/delete_document")
def delete_source_route(data: Delete):
    filename = data.filename
    path_source_documents = os.path.join(os.getcwd(), PATH_NAME_SOURCE_DIRECTORY)
    file_to_delete = f"{path_source_documents}/{filename}"

    if os.path.exists(file_to_delete):
        try:
            os.remove(file_to_delete)
            print(f"{file_to_delete} has been deleted.")

            return {"message": f"{file_to_delete} has been deleted."}
        except OSError as e:
            raise HTTPException(status_code=400, detail=print(f"error: {e}."))
    else:
         raise HTTPException(status_code=400, detail=print(f"The file {file_to_delete} does not exist."))

@api_app.post('/predict')
async def predict(data: Predict):
    global QA
    user_prompt = data.prompt
    if user_prompt:
        res = QA(user_prompt)

        answer, docs = res["result"], res["source_documents"]

        prompt_response_dict = {
            "Prompt": user_prompt,
            "Answer": answer,
        }

        prompt_response_dict["Sources"] = []
        for document in docs:
            prompt_response_dict["Sources"].append(
                (os.path.basename(str(document.metadata["source"])), str(document.page_content))
            )

        return {"response": prompt_response_dict}
    else:
        raise HTTPException(status_code=400, detail="Prompt Incorrect")

@api_app.post("/save_document/")
async def create_upload_file(file: UploadFile):
    # Get the file size (in bytes)
    file.file.seek(0, 2)
    file_size = file.file.tell()

    # move the cursor back to the beginning
    await file.seek(0)

    if file_size > 10 * 1024 * 1024:
        # more than 10 MB
        raise HTTPException(status_code=400, detail="File too large")

    content_type = file.content_type

    if content_type not in [
        "text/plain",
        "text/markdown",
        "text/x-markdown",
        "text/csv",
        "application/msword",
        "application/pdf",
        "application/vnd.ms-excel",
        "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
        "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
        "text/x-python",
        "application/x-python-code"]:
        raise HTTPException(status_code=400, detail="Invalid file type")

    upload_dir = os.path.join(os.getcwd(), PATH_NAME_SOURCE_DIRECTORY)
    if not os.path.exists(upload_dir):
        os.makedirs(upload_dir)

    dest = os.path.join(upload_dir, file.filename)

    with open(dest, "wb") as buffer:
        shutil.copyfileobj(file.file, buffer)

    return {"filename": file.filename}

@api_app.websocket("/ws")
async def websocket_endpoint(websocket: WebSocket):
    await websocket.accept()
    while True:
        data = await websocket.receive_text()

        res = QA(data)

        qa_chain_response = res.stream(
            {"query": data},
        )

        print(f"{qa_chain_response} stream")

        await websocket.send_text(f"Message text was: {data}")