Update app.py

app.py

@@ -1,40 +1,28 @@
-import uvicorn
-from fastapi import FastAPI, Request, Response, File, UploadFile, Form
-from fastapi.responses import JSONResponse, HTMLResponse
-from fastapi.middleware.cors import CORSMiddleware
-import httpx
+import os
 import random
 import re
-import os
-import base64
-import struct
 import shutil
 import tempfile
+import uvicorn
+import httpx
 
-# --- INISIALISASI NLP LOKAL (DistilBERT) ---
-MODEL_DIR = "/media/devmon/D A T A/project/models"
-DISTILBERT_PATH = os.path.join(MODEL_DIR, "distilbert")
+from fastapi import FastAPI, Request, UploadFile, File
+from fastapi.responses import JSONResponse
+from fastapi.middleware.cors import CORSMiddleware
 
-try:
-    from transformers import pipeline
-    if os.path.exists(DISTILBERT_PATH):
-        nlp_engine = pipeline("sentiment-analysis", model=DISTILBERT_PATH, local_files_only=True)
-        print("DistilBERT: Loaded from Disk.")
-    else:
-        nlp_engine = None
-        print("DistilBERT: Folder not found.")
-except Exception as e:
-    nlp_engine = None
-    print(f"Transformers error: {e}")
+from transformers import pipeline
+from faster_whisper import WhisperModel
+from dotenv import load_dotenv
 
-# --- INISIALISASI WHISPER LOKAL ---
-try:
-    from faster_whisper import WhisperModel
-    whisper_model = WhisperModel("tiny", device="cpu", compute_type="int8")
-    print("Whisper-1: Local Engine Ready.")
-except Exception as e:
-    whisper_model = None
-    print(f"Whisper Local error (Switching to Cloud-Only Mode): {e}")
+# =========================
+# LOAD ENV
+# =========================
+
+load_dotenv()
+
+# =========================
+# FASTAPI INIT
+# =========================
 
 app = FastAPI()
 
@@ -46,209 +34,272 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# ================= KONFIGURASI API =================
-OLLAMA_URL = "http://127.0.0.1:11434/api/chat"
-LOCAL_MODEL = "deepseek-r1:1.5b"
-TIMEOUT = 900.0 
-
-# 7 OpenRouter API Keys
-OR_KEYS = [
-    "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",
-    "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",
-    "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",
-    "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",
-    "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",
-    "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",
-    "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"
-]
-
-# 8 Cloudflare Accounts (ID & API Key)
-CF_ACCOUNTS = [
-    {"id": "a3f0db8c35cad604bf7572d0303cf930", "token": "fnCi0UTHbygWNDraZut8EV3N4Tzmw-wagP33gmU2"},
-    {"id": "ae24ffa3616016431a539f2c0ed3d884", "token": "nzpDfTUJw90I0cJQcePM58ZkGWrKruuudhR6L721"},
-    {"id": "f0bc52c5fcd64c76b883965add9fe64a", "token": "vuUIjX5kKH7k_QtYWkejIzJcdtfw-lFGsDzh6BR7"},
-    {"id": "6641f1bee075d1b12d3f5267b3df1424", "token": "BxKVqGpgkXtb2t0sKOguaurT4tkRGH4ZbRLGjG9t"},
-    {"id": "c70d184eb2aae17ad02af3c13d489de1", "token": "2dk5Oj7yClLMU_qpJgxeWSJPy8pebouPPvMVtQzT"},
-    {"id": "84039b443bb1e7dab62613cc39655c80", "token": "6VdcjF7_KtBPP46CEG2R82NzTSHwkBCV81rq0LVq"},
-    {"id": "84bc6c610667a3d8988003dc36b585a4", "token": "Y8h_qYcaXFTbEQ-r-amzzHV_eL37sp-ATZPxiLZz"},
-    {"id": "d8848c9ac069b6cc660a233ca3911331", "token": "ur_Jv30k7tPT6y-6O_TRODKIo05btJj_dfucUtkN"}
-]
-
-# ================= HELPER FUNCTIONS =================
+# =========================
+# MODEL PATH
+# =========================
+
+MODEL_DIR = "./models_cache"
+DISTILBERT_PATH = os.path.join(MODEL_DIR, "distilbert")
+
+# =========================
+# LOAD DISTILBERT
+# =========================
+
+nlp_engine = None
+
+try:
+    if os.path.exists(DISTILBERT_PATH):
+        nlp_engine = pipeline(
+            "sentiment-analysis",
+            model=DISTILBERT_PATH,
+            local_files_only=True
+        )
+        print("DistilBERT loaded")
+except Exception as e:
+    print("DistilBERT failed:", e)
+
+# =========================
+# LOAD WHISPER
+# =========================
+
+whisper_model = None
+
+try:
+    whisper_model = WhisperModel(
+        "tiny",
+        device="cpu",
+        compute_type="int8"
+    )
+    print("Whisper loaded")
+except Exception as e:
+    print("Whisper failed:", e)
+
+# =========================
+# ENV CONFIG
+# =========================
+
+TIMEOUT = 120.0
+
+OLLAMA_URL = os.getenv("OLLAMA_URL", "http://127.0.0.1:11434/api/chat")
+LOCAL_MODEL = os.getenv("LOCAL_MODEL", "deepseek-r1:1.5b")
+
+# =========================
+# OPENROUTER KEYS
+# =========================
+
+OR_KEYS = os.getenv("OPENROUTER_KEYS", "").split(",")
+
+# =========================
+# CLOUDFLARE ACCOUNTS
+# =========================
+
+CF_ACCOUNTS = []
+
+cf_env = os.getenv("CLOUDFLARE_ACCOUNTS", "")
+
+for acc in cf_env.split(","):
+    if ":" in acc:
+        acc_id, token = acc.split(":")
+        CF_ACCOUNTS.append({
+            "id": acc_id,
+            "token": token
+        })
+
+# =========================
+# HELPER
+# =========================
 
 def clean_ai_response(text):
-    text = re.sub(r'<think>.*?</think>', '', text, flags=re.DOTALL)
-    text = text.replace("```json", "").replace("```", "").strip()
-    return text
-
-def pcm_to_wav(pcm_data, sample_rate=24000):
-    num_channels = 1
-    sample_width = 2
-    wav_header = struct.pack('<4sI4s4sIHHIIHH4sI',
-        b'RIFF', 36 + len(pcm_data), b'WAVE',
-        b'fmt ', 16, 1, num_channels, sample_rate,
-        sample_rate * num_channels * sample_width,
-        num_channels * sample_width, sample_width * 8,
-        b'data', len(pcm_data))
-    return wav_header + pcm_data
-
-# ================= ROUTING =================
-
-@app.get("/", response_class=HTMLResponse)
+    text = re.sub(r"<think>.*?</think>", "", text, flags=re.DOTALL)
+    text = text.replace("```json", "").replace("```", "")
+    return text.strip()
+
+
+# =========================
+# ROOT
+# =========================
+
+@app.get("/")
 async def root():
-    return """
-    <html>
-        <body style="font-family:sans-serif; display:flex; justify-content:center; align-items:center; height:100vh; margin:0; background:#0f172a; color:white;">
-            <div style="background:#1e293b; padding:3rem; border-radius:2rem; box-shadow:0 25px 50px -12px rgba(0,0,0,0.5); text-align:center; border: 1px solid #334155;">
-                <h1 style="color:#818cf8; margin-bottom:0.5rem;">STB AI Node v2.6</h1>
-                <p style="color:#10b981; font-weight:bold;">● Hybrid Multi-Fallback Active</p>
-                <div style="font-size: 0.8rem; color: #94a3b8; margin-top: 1.5rem; display:grid; grid-template-cols: 1fr 1fr; gap: 10px;">
-                    <div style="background:#0f172a; padding:10px; border-radius:10px;">LLM: DeepSeek / 7x OR / 8x CF</div>
-                    <div style="background:#0f172a; padding:10px; border-radius:10px;">Whisper: Local / 8x Cloud</div>
-                    <div style="background:#0f172a; padding:10px; border-radius:10px;">TTS: Gemini Pro</div>
-                    <div style="background:#0f172a; padding:10px; border-radius:10px;">NLP: DistilBERT</div>
-                </div>
-            </div>
-        </body>
-    </html>
-    """
-
-@app.get("/v1/models")
-async def get_models():
-    return {
-        "data": [
-            {"id": "gpt-4o-mini", "object": "model"},
-            {"id": "distilbert-nlp", "object": "model"},
-            {"id": LOCAL_MODEL, "object": "model"},
-            {"id": "tts-1", "object": "model"},
-            {"id": "whisper-1", "object": "model"}
-        ]
-    }
+    return {"status": "AI Node running"}
+
+
+# =========================
+# CHAT COMPLETION
+# =========================
 
 @app.post("/v1/chat/completions")
-async def chat_completions(request: Request):
+async def chat(request: Request):
+
     body = await request.json()
-    model_requested = body.get("model", LOCAL_MODEL)
     messages = body.get("messages", [])
-    user_query = messages[-1].get("content", "") if messages else ""
-    
-    # 1. DISTILBERT LOCAL
-    if model_requested == "distilbert-nlp" and nlp_engine:
-        res = nlp_engine(user_query)
-        return {"choices": [{"message": {"role": "assistant", "content": f"NLP: {res}"}}], "model": "distilbert", "provider": "Local"}
 
     async with httpx.AsyncClient(timeout=TIMEOUT) as client:
-        # 2. OLLAMA LOCAL
+
+        # =========================
+        # LOCAL OLLAMA
+        # =========================
+
         try:
-            res = await client.post(OLLAMA_URL, json={"model": LOCAL_MODEL, "messages": messages, "stream": False})
+            res = await client.post(
+                OLLAMA_URL,
+                json={
+                    "model": LOCAL_MODEL,
+                    "messages": messages,
+                    "stream": False
+                }
+            )
+
             if res.status_code == 200:
-                content = clean_ai_response(res.json()['message']['content'])
-                return {"choices": [{"message": {"role": "assistant", "content": content}}], "model": LOCAL_MODEL, "provider": "Local Ollama"}
-        except: pass
 
-        # 3. OPENROUTER (7 KEYS)
+                data = res.json()
+
+                content = clean_ai_response(
+                    data["message"]["content"]
+                )
+
+                return {
+                    "choices": [
+                        {
+                            "message": {
+                                "role": "assistant",
+                                "content": content
+                            }
+                        }
+                    ]
+                }
+
+        except:
+            pass
+
+        # =========================
+        # OPENROUTER FALLBACK
+        # =========================
+
         shuffled_keys = list(OR_KEYS)
         random.shuffle(shuffled_keys)
+
         for key in shuffled_keys:
+
             try:
+
                 res = await client.post(
                     "https://openrouter.ai/api/v1/chat/completions",
-                    headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
-                    json={"model": "openai/gpt-4o-mini", "messages": messages}
+                    headers={
+                        "Authorization": f"Bearer {key}",
+                        "Content-Type": "application/json"
+                    },
+                    json={
+                        "model": "openai/gpt-4o-mini",
+                        "messages": messages
+                    }
                 )
+
                 if res.status_code == 200:
-                    data = res.json()
-                    content = clean_ai_response(data['choices'][0]['message']['content'])
-                    data['choices'][0]['message']['content'] = content
-                    return data
-            except: continue
-        
-        # 4. CLOUDFLARE (8 ACCOUNTS)
+                    return res.json()
+
+            except:
+                continue
+
+        # =========================
+        # CLOUDFLARE FALLBACK
+        # =========================
+
         shuffled_cf = list(CF_ACCOUNTS)
         random.shuffle(shuffled_cf)
+
         for acc in shuffled_cf:
+
             try:
+
+                url = f"https://api.cloudflare.com/client/v4/accounts/{acc['id']}/ai/run/@cf/meta/llama-3-8b-instruct"
+
                 res = await client.post(
-                    f"https://api.cloudflare.com/client/v4/accounts/{acc['id']}/ai/run/@cf/meta/llama-3-8b-instruct",
-                    headers={"Authorization": f"Bearer {acc['token']}"},
-                    json={"messages": messages}
-                )
-                if res.status_code == 200:
-                    data = res.json()
-                    content = clean_ai_response(data['result']['response'])
-                    return {"choices": [{"message": {"role": "assistant", "content": content}}], "model": "cf-llama-3", "provider": "Cloudflare"}
-            except: continue
-
-    return JSONResponse({"error": "All AI paths failed"}, status_code=500)
-
-@app.post("/v1/audio/speech")
-async def text_to_speech(request: Request):
-    """Mendukung API tts-1 dengan Fallback Retry"""
-    try:
-        body = await request.json()
-        text = body.get("input", "")
-        voice = body.get("voice", "Kore")
-        
-        api_key = "" # Diisi oleh lingkungan
-        url = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-preview-tts:generateContent?key={api_key}"
-        
-        async with httpx.AsyncClient(timeout=60.0) as client:
-            for attempt in range(2):
-                res = await client.post(url, json={
-                    "contents": [{"parts": [{"text": text}]}],
-                    "generationConfig": {
-                        "responseModalities": ["AUDIO"],
-                        "speechConfig": {"voiceConfig": {"prebuiltVoiceConfig": {"voiceName": voice}}}
+                    url,
+                    headers={
+                        "Authorization": f"Bearer {acc['token']}"
+                    },
+                    json={
+                        "messages": messages
                     }
-                })
+                )
+
                 if res.status_code == 200:
-                    audio_base64 = res.json()['candidates'][0]['content']['parts'][0]['inlineData']['data']
-                    return Response(content=pcm_to_wav(base64.b64decode(audio_base64)), media_type="audio/wav")
-                
-        return JSONResponse({"error": "TTS Gemini Failed"}, status_code=500)
-    except Exception as e:
-        return JSONResponse({"error": str(e)}, status_code=500)
+                    return res.json()
+
+            except:
+                continue
+
+    return JSONResponse(
+        {"error": "all AI providers failed"},
+        status_code=500
+    )
+
+
+# =========================
+# WHISPER TRANSCRIPTION
+# =========================
 
 @app.post("/v1/audio/transcriptions")
-async def transcribe_audio(
-    file: UploadFile = File(...),
-    model: str = Form("whisper-1")
-):
-    """Mendukung API whisper-1 (Lokal STB -> Fallback ke Cloudflare 8 Akun)"""
-    
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+async def transcribe_audio(file: UploadFile = File(...)):
+
+    if whisper_model is None:
+        return JSONResponse(
+            {"error": "whisper not loaded"},
+            status_code=500
+        )
+
+    with tempfile.NamedTemporaryFile(delete=False) as tmp:
+
         shutil.copyfileobj(file.file, tmp)
-        tmp_path = tmp.name
-    
-    # 1. LOKAL WHISPER
-    if whisper_model:
-        try:
-            segments, info = whisper_model.transcribe(tmp_path, beam_size=5)
-            text = "".join([segment.text for segment in segments])
-            os.remove(tmp_path)
-            return {"text": text.strip(), "language": info.language, "provider": "local-stb"}
-        except: pass
 
-    # 2. CLOUDFLARE WHISPER FALLBACK (8 ACCOUNTS)
-    try:
-        with open(tmp_path, "rb") as audio_file:
-            audio_data = audio_file.read()
+        path = tmp.name
 
-        shuffled_cf = list(CF_ACCOUNTS)
-        random.shuffle(shuffled_cf)
-        
-        async with httpx.AsyncClient(timeout=120.0) as client:
-            for acc in shuffled_cf:
-                cf_url = f"https://api.cloudflare.com/client/v4/accounts/{acc['id']}/ai/run/@cf/openai/whisper"
-                res = await client.post(cf_url, headers={"Authorization": f"Bearer {acc['token']}"}, content=audio_data)
-                if res.status_code == 200:
-                    data = res.json()
-                    os.remove(tmp_path)
-                    return {"text": data['result']['text'], "provider": "cloudflare-cloud"}
-    except: pass
+    segments, info = whisper_model.transcribe(path)
+
+    text = ""
+
+    for s in segments:
+        text += s.text
 
-    if os.path.exists(tmp_path): os.remove(tmp_path)
-    return JSONResponse({"error": "Transcription failed"}, status_code=500)
+    os.remove(path)
+
+    return {
+        "text": text,
+        "language": info.language
+    }
+
+
+# =========================
+# SENTIMENT ANALYSIS
+# =========================
+
+@app.post("/v1/sentiment")
+async def sentiment(request: Request):
+
+    if nlp_engine is None:
+        return JSONResponse(
+            {"error": "distilbert not loaded"},
+            status_code=500
+        )
+
+    body = await request.json()
+
+    text = body.get("text", "")
+
+    result = nlp_engine(text)
+
+    return result
+
+
+# =========================
+# SERVER
+# =========================
 
 if __name__ == "__main__":
-    uvicorn.run(app, host="0.0.0.0", port=8000)
+
+    uvicorn.run(
+        app,
+        host="0.0.0.0",
+        port=8000
+    )