app.py

import os, re, json, traceback, subprocess
from pathlib import Path
from flask import Flask, request, jsonify, send_from_directory
from flask_cors import CORS

BASE_DIR = Path(__file__).parent
STATIC_DIR = BASE_DIR / "static"
VIDEOS_DIR = STATIC_DIR / "videos"
UPLOAD_DIR = BASE_DIR / "uploads"
MODELS_DIR = BASE_DIR / "models"

for p in [STATIC_DIR, VIDEOS_DIR, UPLOAD_DIR, MODELS_DIR, STATIC_DIR / "models"]:
    p.mkdir(exist_ok=True, parents=True)

app = Flask(__name__, static_folder=str(STATIC_DIR))
CORS(app)
whisper_model = None

@app.route("/")
def index():
    return send_from_directory(str(STATIC_DIR), "index.html")

@app.route("/static/<path:path>")
def static_files(path):
    return send_from_directory(str(STATIC_DIR), path)

@app.route("/api/videos")
def list_videos():
    videos = []
    for f in sorted(VIDEOS_DIR.glob("*.mp4")):
        title = f.stem.replace("_", " ").title()
        videos.append({"filename": f.name, "title": title, "url": f"/static/videos/{f.name}"})
    return jsonify({"videos": videos})

@app.route("/api/process-video", methods=["POST"])
def process_video():
    try:
        video_path = None
        local_video = request.form.get("local_video", "").strip()
        if local_video:
            candidate = (VIDEOS_DIR / local_video).resolve()
            if str(candidate).startswith(str(VIDEOS_DIR.resolve())) and candidate.exists():
                video_path = candidate

        if video_path is None and "file" in request.files and request.files["file"].filename:
            uploaded = request.files["file"]
            safe_name = re.sub(r"[^A-Za-z0-9_.-]", "_", uploaded.filename)
            video_path = UPLOAD_DIR / safe_name
            uploaded.save(video_path)

        if video_path is None:
            url = request.form.get("url", "").strip()
            if url:
                video_path = download_video_from_url(url)

        if not video_path:
            return jsonify({"error": "No video found. Choose an internal video or upload an MP4."}), 400

        stt = run_stt(str(video_path))
        transcript = stt["text"]
        segments = stt["segments"]
        tokens = clean_tokens(transcript)
        signs = text_to_asl_glosses(tokens)
        timeline = build_sign_timeline(segments, signs)
        summary = make_simple_summary(transcript, signs)

        return jsonify({
            "video_used": str(Path(video_path).name),
            "transcript": transcript,
            "segments": segments,
            "tokens": tokens[:120],
            "signs": signs,
            "timeline": timeline,
            "summary": summary,
            "metrics": {
                "stt": "Whisper base",
                "glosses": len(signs),
                "avatar": "Real GLB avatar + gloss animation"
            }
        })
    except Exception as e:
        traceback.print_exc()
        return jsonify({"error": str(e)}), 500

@app.route("/api/ask", methods=["POST"])
def ask():
    data = request.get_json(force=True) or {}
    q = data.get("question", "").lower().strip()
    transcript = data.get("transcript", "").strip()
    if not q:
        return jsonify({"answer": "Please write a question."})
    if not transcript:
        return jsonify({"answer": "Process a video first."})
    q_tokens = set(clean_tokens(q))
    sentences = re.split(r"(?<=[.!?])\s+", transcript)
    ranked = []
    for s in sentences:
        score = len(q_tokens.intersection(clean_tokens(s)))
        if score:
            ranked.append((score, s))
    ranked.sort(reverse=True)
    answer = ranked[0][1] if ranked else transcript[:350] + ("..." if len(transcript) > 350 else "")
    return jsonify({"answer": answer})

def download_video_from_url(url: str):
    out = UPLOAD_DIR / "downloaded_video.%(ext)s"
    cmd = ["yt-dlp", "--no-playlist", "--socket-timeout", "20", "--retries", "2", "-f", "best[ext=mp4]/best", "--merge-output-format", "mp4", "-o", str(out), url]
    try:
        r = subprocess.run(cmd, capture_output=True, text=True, timeout=180)
        if r.returncode != 0:
            print(r.stderr[:800])
            return None
        files = list(UPLOAD_DIR.glob("downloaded_video.*"))
        return files[0] if files else None
    except Exception as e:
        print("yt-dlp error:", e)
        return None

def run_stt(video_path: str) -> dict:
    global whisper_model
    try:
        import whisper
        if whisper_model is None:
            print("Loading Whisper base model. This can take time the first time...")
            whisper_model = whisper.load_model("base")
        print("Transcribing:", video_path)
        result = whisper_model.transcribe(video_path, language="en", fp16=False)
        text = result.get("text", "").strip()
        segments = []
        for seg in result.get("segments", []):
            segments.append({
                "start": float(seg.get("start", 0)),
                "end": float(seg.get("end", 0)),
                "text": seg.get("text", "").strip()
            })
        return {"text": text if text else "No speech detected in this video.", "segments": segments}
    except Exception as e:
        print("Whisper failed:", e)
        return {"text": "Automatic transcription failed. Install dependencies with: pip install -r requirements.txt. Also make sure ffmpeg is installed.", "segments": []}

def clean_tokens(text: str):
    text = text.lower()
    text = re.sub(r"[^a-z0-9\s]", " ", text)
    tokens = text.split()
    stop = {"the","a","an","and","or","to","of","in","on","for","with","is","are","was","were","be","been","being","this","that","it","we","you","your","our","i","they","them","as","at","by","from","can","will","would","should","about","into","than","then","so","if","not","do","does","have","has","had","there","their","what","when","where","why","how","also","just","very","really","more","most","one","two","three","first","second","now","today","here","like"}
    return [t for t in tokens if len(t) > 2 and t not in stop]

ASL_MAP = {
    "linkedin":"LINKEDIN", "profile":"PROFILE", "picture":"PICTURE", "photo":"PHOTO", "professional":"PROFESSIONAL", "identity":"IDENTITY", "headline":"HEADLINE", "summary":"SUMMARY", "resume":"RESUME", "cv":"RESUME", "career":"CAREER", "job":"JOB", "work":"WORK", "experience":"EXPERIENCE", "education":"EDUCATION", "skill":"SKILL", "skills":"SKILL", "network":"NETWORK", "connect":"CONNECT", "connection":"CONNECT", "message":"MESSAGE", "company":"COMPANY", "business":"BUSINESS", "interview":"INTERVIEW", "recruiter":"RECRUITER", "hire":"HIRE", "hiring":"HIRE",
    "math":"MATH", "mathematics":"MATH", "number":"NUMBER", "numbers":"NUMBER", "equation":"EQUATION", "equations":"EQUATION", "linear":"LINEAR", "system":"SYSTEM", "systems":"SYSTEM", "solve":"SOLVE", "solution":"SOLUTION", "variable":"VARIABLE", "matrix":"MATRIX", "addition":"ADDITION", "add":"ADDITION", "subtraction":"SUBTRACT", "subtract":"SUBTRACT", "multiply":"MULTIPLY", "division":"DIVIDE", "equal":"EQUAL", "graph":"GRAPH", "function":"FUNCTION",
    "science":"SCIENCE", "cell":"CELL", "energy":"ENERGY", "plant":"PLANT", "photosynthesis":"PHOTOSYNTHESIS", "learn":"LEARN", "learning":"LEARN", "lesson":"LESSON", "student":"STUDENT", "teacher":"TEACHER", "school":"SCHOOL", "course":"COURSE", "hello":"HELLO", "welcome":"WELCOME", "important":"IMPORTANT", "good":"GOOD", "great":"GREAT", "clear":"CLEAR", "example":"EXAMPLE", "result":"RESULT", "calculate":"CALCULATE", "understand":"UNDERSTAND", "explain":"EXPLAIN", "help":"HELP"
}

def text_to_asl_glosses(tokens):
    signs, seen = [], set()
    for t in tokens:
        gloss = ASL_MAP.get(t)
        if gloss and gloss not in seen:
            signs.append({"gloss": gloss, "source_word": t, "confidence": 0.88, "mode": "dictionary"})
            seen.add(gloss)
    if len(signs) < 5:
        for t in tokens[:12]:
            gloss = t.upper()
            if gloss not in seen:
                signs.append({"gloss": gloss, "source_word": t, "confidence": 0.60, "mode": "finger-spelling"})
                seen.add(gloss)
            if len(signs) >= 12:
                break
    return signs[:18]

def build_sign_timeline(segments, global_signs):
    timeline = []
    for seg in segments:
        tokens = clean_tokens(seg.get("text", ""))
        local_signs = text_to_asl_glosses(tokens)
        if not local_signs:
            continue
        start, end = float(seg["start"]), float(seg["end"])
        duration = max(end - start, 0.8)
        step = duration / len(local_signs)
        for i, sign in enumerate(local_signs):
            timeline.append({
                "start": start + i * step,
                "end": start + (i + 1) * step,
                "gloss": sign["gloss"],
                "source_word": sign["source_word"],
                "mode": sign["mode"]
            })
    if timeline:
        return timeline[:120]
    # fallback cycle when Whisper segments fail
    t = 0.0
    for sign in global_signs:
        timeline.append({"start": t, "end": t + 1.4, "gloss": sign["gloss"], "source_word": sign["source_word"], "mode": sign["mode"]})
        t += 1.4
    return timeline

def make_simple_summary(transcript, signs):
    return {
        "first_words": transcript[:220] + ("..." if len(transcript) > 220 else ""),
        "main_glosses": [s["gloss"] for s in signs[:10]],
        "note": "The 3D avatar uses a real GLB model. It follows glosses generated from Whisper transcript. This is an MVP rendering layer, not a full ASL production engine yet."
    }

if __name__ == "__main__":
    print("\n🚀 iLearning SignAI GLB avatar demo")
    print("Open: http://0.0.0.0:7860\n")
    app.run(host="0.0.0.0", port=7860, debug=False)