app.py

import json
import logging
import os
import threading
from datetime import datetime

import requests
from dotenv import load_dotenv
from fastapi import FastAPI, Request
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse, JSONResponse
from fastapi.staticfiles import StaticFiles
from huggingface_hub import HfApi, create_repo, hf_hub_download
from openai import OpenAI
from pydantic import BaseModel, Field, field_validator
from pypdf import PdfReader
from slowapi import Limiter, _rate_limit_exceeded_handler
from slowapi.errors import RateLimitExceeded
from slowapi.util import get_remote_address
from starlette.middleware.base import BaseHTTPMiddleware


logger = logging.getLogger("assistant")


load_dotenv(override=True)

def _upload_to_hf(path_or_fileobj, path_in_repo, repo_id):
    """Upload en arrière-plan vers HF avec retry."""
    try:
        api = HfApi()
        api.upload_file(
            path_or_fileobj=path_or_fileobj,
            path_in_repo=path_in_repo,
            repo_id=repo_id,
            repo_type="dataset",
            token=os.getenv("HF_TOKEN"),
        )
    except Exception as e:
        print(f"[UPLOAD] Erreur upload HF ({path_in_repo}) : {e}", flush=True)


# Chemins dans le dataset HF (pas exposés dans le repo public me/)
KNOWLEDGE_HF_PATHS = (
    "me/cv.pdf",
    "me/summary.txt",
    "me/Travaux.txt",
)
KNOWLEDGE_LOCAL_PATHS = KNOWLEDGE_HF_PATHS


def _knowledge_config():
    """Retourne (token, dataset_id) ou (None, None)."""
    token = os.getenv("HF_TOKEN")
    dataset_id = os.getenv("KNOWLEDGE_DATASET_ID")
    if token and dataset_id:
        return token, dataset_id
    return None, None


def _hf_knowledge_file_path(path_in_repo: str) -> str | None:
    """Chemin cache HF (hors dossier me/ du Space)."""
    token, dataset_id = _knowledge_config()
    if not token or not dataset_id:
        return None
    try:
        return hf_hub_download(
            repo_id=dataset_id,
            repo_type="dataset",
            filename=path_in_repo,
            token=token,
        )
    except Exception:
        print(f"[KNOWLEDGE] {path_in_repo} absent sur HF", flush=True)
        return None


def _read_text_file(path: str) -> str:
    with open(path, "r", encoding="utf-8") as f:
        return f.read()


def _read_pdf_text(path: str) -> str:
    text = ""
    reader = PdfReader(path)
    for page in reader.pages:
        page_text = page.extract_text()
        if page_text:
            text += page_text
    return text


def remove_local_knowledge_copies():
    """Supprime les copies sensibles dans me/ (runtime du conteneur)."""
    cwd = os.getcwd()
    removed: list[str] = []
    absent: list[str] = []
    errors: list[str] = []

    for local_path in KNOWLEDGE_LOCAL_PATHS:
        full = os.path.join(cwd, local_path)
        if not os.path.isfile(local_path) and not os.path.isfile(full):
            absent.append(local_path)
            continue
        target = local_path if os.path.isfile(local_path) else full
        try:
            os.remove(target)
            removed.append(local_path)
        except OSError as exc:
            errors.append(f"{local_path} ({exc})")

    if removed:
        print(
            f"[KNOWLEDGE] Supprimé de me/ (runtime): {', '.join(removed)}",
            flush=True,
        )
    if absent:
        print(
            "[KNOWLEDGE] Déjà absents de me/ au runtime "
            f"(chargement HF uniquement): {', '.join(absent)}",
            flush=True,
        )
    if errors:
        print(
            f"[KNOWLEDGE] Échec suppression: {'; '.join(errors)}",
            flush=True,
        )
    if not removed and not absent and not errors:
        print("[KNOWLEDGE] Aucun fichier knowledge à traiter.", flush=True)


def init_knowledge_dataset():
    """Crée le dataset privé et bootstrap depuis me/ si besoin (one-shot)."""
    token, dataset_id = _knowledge_config()
    if not token or not dataset_id:
        print(
            "[KNOWLEDGE] HF_TOKEN ou KNOWLEDGE_DATASET_ID manquant, "
            "dataset non initialisé.",
            flush=True,
        )
        return
    try:
        create_repo(
            repo_id=dataset_id,
            repo_type="dataset",
            private=True,
            token=token,
            exist_ok=True,
        )
        print(f"[KNOWLEDGE] Dataset prêt : {dataset_id}", flush=True)
    except Exception as e:
        print(f"[KNOWLEDGE] Erreur init dataset : {e}", flush=True)
        return

    os.makedirs("me", exist_ok=True)
    uploaded = 0
    for path_in_repo in KNOWLEDGE_HF_PATHS:
        if _hf_knowledge_file_path(path_in_repo):
            continue
        if os.path.isfile(path_in_repo):
            _upload_to_hf(path_in_repo, path_in_repo, dataset_id)
            print(
                f"[KNOWLEDGE] {path_in_repo} envoyé vers HF (bootstrap)",
                flush=True,
            )
            uploaded += 1

    if uploaded:
        print(
            f"[KNOWLEDGE] {uploaded} fichier(s) bootstrap vers HF.",
            flush=True,
        )
    remove_local_knowledge_copies()


def init_logs_dataset():
    token = os.getenv("HF_TOKEN")
    dataset_id = os.getenv("LOGS_DATASET_ID")
    if not token or not dataset_id:
        print("[LOG] HF_TOKEN ou LOGS_DATASET_ID manquant, dataset non initialisé.", flush=True)
        return
    try:
        create_repo(repo_id=dataset_id, repo_type="dataset", private=True, token=token, exist_ok=True)
        print(f"[LOG] Dataset logs prêt : {dataset_id}", flush=True)
    except Exception as e:
        print(f"[LOG] Erreur init dataset : {e}", flush=True)
    # Récupérer le fichier existant depuis HF pour ne pas perdre les données après un rebuild
    try:
        path = hf_hub_download(repo_id=dataset_id, repo_type="dataset", filename="me/log.txt", token=token)
        with open(path, "r", encoding="utf-8") as src:
            content = src.read()
        with open("me/log.txt", "w", encoding="utf-8") as dst:
            dst.write(content)
        print(f"[LOG] Fichier log.txt récupéré ({len(content.splitlines())} lignes)", flush=True)
    except Exception:
        print("[LOG] Aucun fichier log.txt existant sur HF, démarrage à vide.", flush=True)


def init_questions_dataset():
    token = os.getenv("HF_TOKEN")
    questions_id = os.getenv("QUESTIONS_DATASET_ID")
    if not token or not questions_id:
        print("[QUESTIONS] HF_TOKEN ou QUESTIONS_DATASET_ID manquant, dataset non initialisé.", flush=True)
        return
    try:
        create_repo(repo_id=questions_id, repo_type="dataset", private=True, token=token, exist_ok=True)
        print(f"[QUESTIONS] Dataset questions prêt : {questions_id}", flush=True)
    except Exception as e:
        print(f"[QUESTIONS] Erreur init dataset : {e}", flush=True)
    # Récupérer le fichier existant depuis HF
    try:
        path = hf_hub_download(repo_id=questions_id, repo_type="dataset", filename="questions.txt", token=token)
        with open(path, "r", encoding="utf-8") as src:
            content = src.read()
        with open("me/questions.txt", "w", encoding="utf-8") as dst:
            dst.write(content)
        print(f"[QUESTIONS] Fichier questions.txt récupéré ({len(content.splitlines())} lignes)", flush=True)
    except Exception:
        print("[QUESTIONS] Aucun fichier questions.txt existant sur HF, démarrage à vide.", flush=True)


def push_question(question):
    """Enregistre chaque question posée par un utilisateur dans le dataset Questions."""
    now = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
    print(f"[QUESTIONS] {now} — {question}", flush=True)
    questions_path = "me/questions.txt"
    entry = f"[{now}] {question}\n"
    with open(questions_path, "a", encoding="utf-8") as f:
        f.write(entry)
    threading.Thread(
        target=_upload_to_hf,
        args=(questions_path, "questions.txt", os.getenv("QUESTIONS_DATASET_ID")),
        daemon=True,
    ).start()


def push(text):
    print(f"[LOG] {text}", flush=True)
    log_path = "me/log.txt"
    entry = f"[{datetime.now().strftime('%Y-%m-%d %H:%M:%S')}] {text}\n"
    with open(log_path, "a", encoding="utf-8") as f:
        f.write(entry)
    threading.Thread(
        target=_upload_to_hf,
        args=(log_path, log_path, os.getenv("LOGS_DATASET_ID")),
        daemon=True,
    ).start()


def record_user_details(email, name="Name not provided", notes="not provided"):
    push(f"Recording {name} with email {email} and notes {notes}")
    return {"recorded": "ok"}

def record_unknown_question(question):
    push(f"Recording {question}")
    return {"recorded": "ok"}

record_user_details_json = {
    "name": "record_user_details",
    "description": "Utilise cet outil pour enregistrer qu'un utilisateur souhaite être contacté et a fourni une adresse e-mail",
    "parameters": {
        "type": "object",
        "properties": {
            "email": {
                "type": "string",
                "description": "L'adresse e-mail de cet utilisateur"
            },
            "name": {
                "type": "string",
                "description": "Le nom de l'utilisateur, s'il l'a fourni"
            }
            ,
            "notes": {
                "type": "string",
                "description": "Toute information supplémentaire sur la conversation qui mérite d'être enregistrée pour donner du contexte"
            }
        },
        "required": ["email"],
        "additionalProperties": False
    }
}

record_unknown_question_json = {
    "name": "record_unknown_question",
    "description": "Utilise toujours cet outil pour enregistrer toute question à laquelle tu n'as pas pu répondre faute de connaissance",
    "parameters": {
        "type": "object",
        "properties": {
            "question": {
                "type": "string",
                "description": "La question à laquelle il n'a pas été possible de répondre"
            },
        },
        "required": ["question"],
        "additionalProperties": False
    }
}

tools = [{"type": "function", "function": record_user_details_json},
        {"type": "function", "function": record_unknown_question_json}]


ALLOWED_TOOLS = {
    "record_user_details": record_user_details,
    "record_unknown_question": record_unknown_question,
}


class Me:

    def __init__(self):
        self.openai = OpenAI()
        self.name = "Vikou Nelson"
        self.linkedin = self._load_knowledge_pdf()
        self.summary = self._load_knowledge_text("me/summary.txt")
        self.travaux = self._load_knowledge_text("me/Travaux.txt")
        remove_local_knowledge_copies()

    def _load_knowledge_text(self, path_in_repo: str) -> str:
        cached = _hf_knowledge_file_path(path_in_repo)
        if cached:
            print(
                f"[KNOWLEDGE] {path_in_repo} chargé depuis le dataset HF",
                flush=True,
            )
            return _read_text_file(cached)
        if os.path.isfile(path_in_repo):
            print(
                f"[KNOWLEDGE] {path_in_repo} chargé depuis me/ (local)",
                flush=True,
            )
            return _read_text_file(path_in_repo)
        print(f"[KNOWLEDGE] {path_in_repo} absent.", flush=True)
        return ""

    def _load_knowledge_pdf(self) -> str:
        path_in_repo = "me/cv.pdf"
        cached = _hf_knowledge_file_path(path_in_repo)
        if cached:
            print(
                f"[KNOWLEDGE] {path_in_repo} chargé depuis le dataset HF",
                flush=True,
            )
            return _read_pdf_text(cached)
        if os.path.isfile(path_in_repo):
            print(
                f"[KNOWLEDGE] {path_in_repo} chargé depuis me/ (local)",
                flush=True,
            )
            return _read_pdf_text(path_in_repo)
        print(
            "[KNOWLEDGE] me/cv.pdf absent — profil LinkedIn vide.",
            flush=True,
        )
        return ""


    def handle_tool_call(self, tool_calls):
        results = []
        for tool_call in tool_calls:
            tool_name = tool_call.function.name
            arguments = json.loads(tool_call.function.arguments)
            print(f"Tool called: {tool_name}", flush=True)
            tool = ALLOWED_TOOLS.get(tool_name)
            if tool is None:
                print(
                    f"[SECURITY] Tool non autorisé refusé : {tool_name}",
                    flush=True,
                )
                result = {"error": "tool_not_allowed"}
            else:
                result = tool(**arguments)
            results.append({
                "role": "tool",
                "content": json.dumps(result),
                "tool_call_id": tool_call.id,
            })
        return results
    
    def system_prompt(self):
        system_prompt = f"You are acting as {self.name}. You are answering questions on {self.name}'s website, \
        particularly questions related to {self.name}'s career, background, skills and experience. \
        Your responsibility is to represent {self.name} for interactions on the website as faithfully as possible. \
        You are given a summary of {self.name}'s background and LinkedIn profile which you can use to answer questions. \
        Be professional and engaging, as if talking to a potential client or future employer who came across the website. \
        If you don't know the answer to any question, use your record_unknown_question tool to record the question that you couldn't answer, even if it's about something trivial or unrelated to career. \
        If the user is engaging in discussion, try to steer them towards getting in touch via email; ask for their email and record it using your record_user_details tool. "

        system_prompt += f"\n\n## Summary:\n{self.summary}\n\n## LinkedIn Profile:\n{self.linkedin}\n\n## Portfolio & Projects:\n{self.travaux}\n\n"
        
        system_prompt += f"With this context, please chat with the user, always staying in character as {self.name}."
        return system_prompt
    
    def chat(self, message, history):
        push_question(message)
        history_messages = []
        for item in history:
            if isinstance(item, dict):
                history_messages.append(item)
            else:
                user_msg, assistant_msg = item[0], item[1]
                history_messages.append({"role": "user", "content": user_msg})
                if assistant_msg:
                    history_messages.append({"role": "assistant", "content": assistant_msg})
        messages = [{"role": "system", "content": self.system_prompt()}] + history_messages + [{"role": "user", "content": message}]
        try:
            done = False
            while not done:
                response = self.openai.chat.completions.create(model="gpt-4o-mini", messages=messages, tools=tools)
                if response.choices[0].finish_reason=="tool_calls":
                    message = response.choices[0].message
                    tool_calls = message.tool_calls
                    results = self.handle_tool_call(tool_calls)
                    messages.append(message)
                    messages.extend(results)
                else:
                    done = True
            return response.choices[0].message.content
        except requests.RequestException:
            return "Notre service est momentanément indisponible. Veuillez réessayer dans quelques instants."
        except (KeyError, IndexError, ValueError):
            return "Une erreur inattendue s'est produite lors du traitement de votre demande. Veuillez reformuler votre question ou réessayer."


load_dotenv(override=True)
init_knowledge_dataset()
init_logs_dataset()
init_questions_dataset()
me = Me()

app = FastAPI()

limiter = Limiter(key_func=get_remote_address)
app.state.limiter = limiter
app.add_exception_handler(
    RateLimitExceeded, _rate_limit_exceeded_handler,
)


class SecurityHeadersMiddleware(BaseHTTPMiddleware):
    """Ajoute des en-têtes de sécurité HTTP standards.

    Note : on n'utilise pas ``X-Frame-Options`` car cet en-tête ne
    supporte pas plusieurs origines. On utilise à la place la directive
    CSP ``frame-ancestors`` qui prend le pas dans les navigateurs
    modernes et autorise Hugging Face Spaces à embarquer l'app dans
    un iframe.
    """

    async def dispatch(self, request, call_next):
        response = await call_next(request)
        response.headers["X-Content-Type-Options"] = "nosniff"
        response.headers["Referrer-Policy"] = (
            "strict-origin-when-cross-origin"
        )
        response.headers["Content-Security-Policy"] = (
            "default-src 'self'; "
            "script-src 'self' 'unsafe-inline' "
            "https://cdn.jsdelivr.net; "
            "style-src 'self' 'unsafe-inline'; "
            "img-src 'self' data:; "
            "connect-src 'self'; "
            "frame-ancestors 'self' https://huggingface.co "
            "https://*.hf.space"
        )
        return response


app.add_middleware(SecurityHeadersMiddleware)

_origins_env = os.getenv("ALLOWED_ORIGINS", "*")
if _origins_env.strip() == "*":
    _allowed_origins = ["*"]
else:
    _allowed_origins = [
        o.strip() for o in _origins_env.split(",") if o.strip()
    ]

app.add_middleware(
    CORSMiddleware,
    allow_origins=_allowed_origins,
    allow_methods=["GET", "POST", "OPTIONS"],
    allow_headers=["*"],
)


class ChatRequest(BaseModel):
    message: str = Field(..., min_length=1, max_length=2000)
    history: list = Field(default_factory=list)
    consent: bool = False

    @field_validator("history")
    @classmethod
    def limit_history(cls, v: list) -> list:
        if len(v) > 50:
            return v[-50:]
        return v


@app.post("/chat")
@limiter.limit("20/minute")
async def chat(request: Request, req: ChatRequest):
    reply = me.chat(req.message, req.history)
    return {"response": reply}


STUDENTS_UNAVAILABLE_MESSAGE = (
    "The student assistant is currently unavailable. "
    "Please check back later."
)


@app.post("/chat/students")
@limiter.limit("20/minute")
async def chat_students(request: Request, req: ChatRequest):
    return {"response": STUDENTS_UNAVAILABLE_MESSAGE}


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