Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -90,15 +90,13 @@ def serpapi_search(query: str, location=None, num_results=3):
|
|
| 90 |
|
| 91 |
# Set location and language based on query
|
| 92 |
if is_indonesian_query:
|
| 93 |
-
country = "id"
|
| 94 |
lang = "id"
|
| 95 |
search_location = location or "Indonesia"
|
| 96 |
-
print(f"[SEARCH] ๐ฎ๐ฉ Indonesian query detected - Using ID region")
|
| 97 |
else:
|
| 98 |
-
country = "us"
|
| 99 |
lang = "en"
|
| 100 |
search_location = location or ""
|
| 101 |
-
print(f"[SEARCH] ๐ Global query - Using worldwide search")
|
| 102 |
|
| 103 |
url = "https://serpapi.com/search.json"
|
| 104 |
params = {
|
|
@@ -115,25 +113,20 @@ def serpapi_search(query: str, location=None, num_results=3):
|
|
| 115 |
images = []
|
| 116 |
|
| 117 |
try:
|
| 118 |
-
#
|
| 119 |
-
print("[SEARCH] ๐ Fetching text results...")
|
| 120 |
r = requests.get(url, params=params, timeout=10)
|
| 121 |
r.raise_for_status()
|
| 122 |
data = r.json()
|
| 123 |
|
| 124 |
if "organic_results" in data:
|
| 125 |
-
print(f"[SEARCH] ๐ Found {len(data['organic_results'])} organic results")
|
| 126 |
for item in data["organic_results"][:num_results]:
|
| 127 |
results.append({
|
| 128 |
"title": item.get("title", ""),
|
| 129 |
"snippet": item.get("snippet", ""),
|
| 130 |
"link": item.get("link", "")
|
| 131 |
})
|
| 132 |
-
print(f"[SEARCH] - {item.get('title', 'No title')[:60]}")
|
| 133 |
-
else:
|
| 134 |
-
print("[SEARCH] โ ๏ธ No organic results found")
|
| 135 |
|
| 136 |
-
#
|
| 137 |
img_params = {
|
| 138 |
"q": query,
|
| 139 |
"engine": "google_images",
|
|
@@ -142,27 +135,23 @@ def serpapi_search(query: str, location=None, num_results=3):
|
|
| 142 |
"gl": country,
|
| 143 |
"hl": lang
|
| 144 |
}
|
| 145 |
-
|
| 146 |
-
print("[SEARCH] ๐ผ๏ธ Fetching images...")
|
| 147 |
img_r = requests.get(url, params=img_params, timeout=10)
|
| 148 |
img_r.raise_for_status()
|
| 149 |
img_data = img_r.json()
|
| 150 |
-
|
| 151 |
if "images_results" in img_data:
|
| 152 |
-
print(f"[SEARCH] ๐ผ๏ธ Found {len(img_data['images_results'])} images")
|
| 153 |
for img in img_data["images_results"][:3]:
|
| 154 |
img_url = img.get("original", img.get("thumbnail", ""))
|
| 155 |
if img_url:
|
| 156 |
images.append(img_url)
|
| 157 |
-
print(f"[SEARCH] - Image: {img_url[:60]}...")
|
| 158 |
-
else:
|
| 159 |
-
print("[SEARCH] โ ๏ธ No images found in image search")
|
| 160 |
|
| 161 |
-
|
|
|
|
|
|
|
|
|
|
| 162 |
return {"results": results, "images": images}
|
| 163 |
|
| 164 |
except Exception as e:
|
| 165 |
-
print(f"[SEARCH] โ Error: {e}
|
| 166 |
return {"results": [], "images": []}
|
| 167 |
|
| 168 |
# =========================
|
|
@@ -172,14 +161,11 @@ def stream_chat(prompt: str, history=None):
|
|
| 172 |
wib = timezone(timedelta(hours=7))
|
| 173 |
now = datetime.now(wib)
|
| 174 |
sys_prompt = SYSTEM_PROMPT + f"\nCurrent time: {now.strftime('%A, %d %B %Y โ %H:%M:%S WIB')}."
|
| 175 |
-
|
| 176 |
messages = [{"role": "system", "content": sys_prompt}]
|
| 177 |
if history:
|
| 178 |
messages += history
|
| 179 |
messages.append({"role": "user", "content": prompt})
|
| 180 |
|
| 181 |
-
print(f"[CHAT] ๐ฌ Sending request to LLM (history: {len(history) if history else 0} msgs)")
|
| 182 |
-
|
| 183 |
payload = {
|
| 184 |
"model": "moonshotai/kimi-k2-instruct-0905",
|
| 185 |
"messages": messages,
|
|
@@ -187,25 +173,23 @@ def stream_chat(prompt: str, history=None):
|
|
| 187 |
"max_tokens": 3600,
|
| 188 |
"stream": True,
|
| 189 |
}
|
| 190 |
-
|
| 191 |
headers = {"Authorization": f"Bearer {GROQ_API_KEY_1}"}
|
| 192 |
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 196 |
continue
|
| 197 |
-
line = line.decode()
|
| 198 |
-
if line.startswith("data: "):
|
| 199 |
-
data = line[6:]
|
| 200 |
-
if data == "[DONE]":
|
| 201 |
-
print("[CHAT] โ
Stream completed")
|
| 202 |
-
break
|
| 203 |
-
try:
|
| 204 |
-
delta = json.loads(data)["choices"][0]["delta"].get("content", "")
|
| 205 |
-
if delta:
|
| 206 |
-
yield delta
|
| 207 |
-
except:
|
| 208 |
-
continue
|
| 209 |
|
| 210 |
# =========================
|
| 211 |
# ๐ Chat Endpoint (Text + Voice)
|
|
@@ -215,89 +199,52 @@ def chat():
|
|
| 215 |
print("\n" + "="*60)
|
| 216 |
print(f"[REQUEST] ๐จ New request at {datetime.now().strftime('%H:%M:%S')}")
|
| 217 |
|
| 218 |
-
# ๐ค Voice Mode
|
| 219 |
if "audio" in request.files:
|
| 220 |
-
|
| 221 |
audio = request.files["audio"]
|
| 222 |
temp = f"/tmp/{time.time()}_{random.randint(1000,9999)}.wav"
|
| 223 |
audio.save(temp)
|
| 224 |
-
print(f"[AUDIO] ๐พ Saved to: {temp}")
|
| 225 |
-
|
| 226 |
user_text = transcribe_audio(temp)
|
| 227 |
-
if not user_text:
|
| 228 |
-
print("[ERROR] โ STT failed")
|
| 229 |
-
return jsonify({"error": "Failed STT"}), 500
|
| 230 |
-
|
| 231 |
-
print(f"[USER] ๐ค Transcribed: '{user_text}'")
|
| 232 |
|
| 233 |
keywords = ["hotel", "mall", "resort", "villa", "tempat wisata", "restaurant", "cafe"]
|
| 234 |
has_keyword = any(k in user_text.lower() for k in keywords)
|
| 235 |
-
|
| 236 |
if has_keyword:
|
| 237 |
-
print(f"[KEYWORD] โ
Detected keyword in: '{user_text}'")
|
| 238 |
serp = serpapi_search(user_text)
|
| 239 |
-
|
| 240 |
-
if serp["images"]:
|
| 241 |
-
print(f"[IMAGES] ๐ผ๏ธ Adding {len(serp['images'])} images to prompt")
|
| 242 |
-
else:
|
| 243 |
-
print("[IMAGES] โ ๏ธ No images to add")
|
| 244 |
-
|
| 245 |
text = "\n".join([f"{r['title']} โ {r['snippet']} โ {r['link']}" for r in serp["results"]])
|
| 246 |
imgs = " ".join(serp["images"])
|
| 247 |
user_text = f"{user_text}\n\nGoogle Results:\n{text}\n\nImages: {imgs}\n\nExplain & recommend."
|
| 248 |
-
print(f"[PROMPT] ๐ Enhanced prompt length: {len(user_text)} chars")
|
| 249 |
-
else:
|
| 250 |
-
print(f"[KEYWORD] โ No keyword detected in: '{user_text}'")
|
| 251 |
|
| 252 |
-
print("[AI] ๐ค Generating response...")
|
| 253 |
ai = "".join(chunk for chunk in stream_chat(user_text))
|
| 254 |
-
print(f"[AI] โ
Response generated ({len(ai)} chars)")
|
| 255 |
-
|
| 256 |
-
print("[TTS] ๐ Converting to audio...")
|
| 257 |
audio_bytes = text_to_speech(ai)
|
| 258 |
|
| 259 |
-
|
| 260 |
-
|
| 261 |
"mode": "voice",
|
| 262 |
"transcript": user_text,
|
| 263 |
"reply_text": ai,
|
| 264 |
"audio_base64": "data:audio/mp3;base64," + base64.b64encode(audio_bytes).decode()
|
| 265 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 266 |
|
| 267 |
# ๐ฌ Text Mode
|
| 268 |
-
print("[MODE] ๐ฌ Text mode detected")
|
| 269 |
data = request.get_json(force=True)
|
| 270 |
prompt = data.get("prompt", "")
|
| 271 |
history = data.get("history", [])
|
| 272 |
-
|
| 273 |
-
print(f"[USER] ๐ค Prompt: '{prompt}'")
|
| 274 |
-
print(f"[HISTORY] ๐ History length: {len(history)} messages")
|
| 275 |
|
| 276 |
keywords = ["hotel", "mall", "resort", "villa", "tempat wisata", "restaurant", "cafe"]
|
| 277 |
has_keyword = any(k in prompt.lower() for k in keywords)
|
| 278 |
-
|
| 279 |
if has_keyword:
|
| 280 |
-
print(f"[KEYWORD] โ
Detected keyword in: '{prompt}'")
|
| 281 |
serp = serpapi_search(prompt)
|
| 282 |
-
|
| 283 |
-
if serp["images"]:
|
| 284 |
-
print(f"[IMAGES] ๐ผ๏ธ Adding {len(serp['images'])} images to prompt")
|
| 285 |
-
else:
|
| 286 |
-
print("[IMAGES] โ ๏ธ No images to add")
|
| 287 |
-
|
| 288 |
text = "\n".join([f"{r['title']} โ {r['snippet']} โ {r['link']}" for r in serp["results"]])
|
| 289 |
imgs = " ".join(serp["images"])
|
| 290 |
prompt = f"{prompt}\n\nGoogle Results:\n{text}\n\nImages: {imgs}\n\nExplain & recommend."
|
| 291 |
-
print(f"[PROMPT] ๐ Enhanced prompt length: {len(prompt)} chars")
|
| 292 |
-
else:
|
| 293 |
-
print(f"[KEYWORD] โ No keyword detected in: '{prompt}'")
|
| 294 |
-
|
| 295 |
-
print("[AI] ๐ค Streaming response...")
|
| 296 |
|
| 297 |
def generate():
|
| 298 |
for chunk in stream_chat(prompt, history):
|
| 299 |
yield chunk
|
| 300 |
-
print("="*60 + "\n")
|
| 301 |
|
| 302 |
return Response(generate(), mimetype="text/plain")
|
| 303 |
|
|
|
|
| 90 |
|
| 91 |
# Set location and language based on query
|
| 92 |
if is_indonesian_query:
|
| 93 |
+
country = "id"
|
| 94 |
lang = "id"
|
| 95 |
search_location = location or "Indonesia"
|
|
|
|
| 96 |
else:
|
| 97 |
+
country = "us"
|
| 98 |
lang = "en"
|
| 99 |
search_location = location or ""
|
|
|
|
| 100 |
|
| 101 |
url = "https://serpapi.com/search.json"
|
| 102 |
params = {
|
|
|
|
| 113 |
images = []
|
| 114 |
|
| 115 |
try:
|
| 116 |
+
# Text results
|
|
|
|
| 117 |
r = requests.get(url, params=params, timeout=10)
|
| 118 |
r.raise_for_status()
|
| 119 |
data = r.json()
|
| 120 |
|
| 121 |
if "organic_results" in data:
|
|
|
|
| 122 |
for item in data["organic_results"][:num_results]:
|
| 123 |
results.append({
|
| 124 |
"title": item.get("title", ""),
|
| 125 |
"snippet": item.get("snippet", ""),
|
| 126 |
"link": item.get("link", "")
|
| 127 |
})
|
|
|
|
|
|
|
|
|
|
| 128 |
|
| 129 |
+
# Images search
|
| 130 |
img_params = {
|
| 131 |
"q": query,
|
| 132 |
"engine": "google_images",
|
|
|
|
| 135 |
"gl": country,
|
| 136 |
"hl": lang
|
| 137 |
}
|
|
|
|
|
|
|
| 138 |
img_r = requests.get(url, params=img_params, timeout=10)
|
| 139 |
img_r.raise_for_status()
|
| 140 |
img_data = img_r.json()
|
|
|
|
| 141 |
if "images_results" in img_data:
|
|
|
|
| 142 |
for img in img_data["images_results"][:3]:
|
| 143 |
img_url = img.get("original", img.get("thumbnail", ""))
|
| 144 |
if img_url:
|
| 145 |
images.append(img_url)
|
|
|
|
|
|
|
|
|
|
| 146 |
|
| 147 |
+
# Debug JSON
|
| 148 |
+
print("[DEBUG] ๐น SerpAPI JSON output:")
|
| 149 |
+
print(json.dumps({"results": results, "images": images}, indent=2))
|
| 150 |
+
|
| 151 |
return {"results": results, "images": images}
|
| 152 |
|
| 153 |
except Exception as e:
|
| 154 |
+
print(f"[SEARCH] โ Error: {e}")
|
| 155 |
return {"results": [], "images": []}
|
| 156 |
|
| 157 |
# =========================
|
|
|
|
| 161 |
wib = timezone(timedelta(hours=7))
|
| 162 |
now = datetime.now(wib)
|
| 163 |
sys_prompt = SYSTEM_PROMPT + f"\nCurrent time: {now.strftime('%A, %d %B %Y โ %H:%M:%S WIB')}."
|
|
|
|
| 164 |
messages = [{"role": "system", "content": sys_prompt}]
|
| 165 |
if history:
|
| 166 |
messages += history
|
| 167 |
messages.append({"role": "user", "content": prompt})
|
| 168 |
|
|
|
|
|
|
|
| 169 |
payload = {
|
| 170 |
"model": "moonshotai/kimi-k2-instruct-0905",
|
| 171 |
"messages": messages,
|
|
|
|
| 173 |
"max_tokens": 3600,
|
| 174 |
"stream": True,
|
| 175 |
}
|
|
|
|
| 176 |
headers = {"Authorization": f"Bearer {GROQ_API_KEY_1}"}
|
| 177 |
|
| 178 |
+
for line in requests.post(GROQ_URL_CHAT, headers=headers, json=payload, stream=True).iter_lines():
|
| 179 |
+
if not line:
|
| 180 |
+
continue
|
| 181 |
+
line = line.decode()
|
| 182 |
+
if line.startswith("data: "):
|
| 183 |
+
data = line[6:]
|
| 184 |
+
if data == "[DONE]":
|
| 185 |
+
break
|
| 186 |
+
try:
|
| 187 |
+
delta = json.loads(data)["choices"][0]["delta"].get("content", "")
|
| 188 |
+
if delta:
|
| 189 |
+
print("[DEBUG] ๐น AI chunk:", delta)
|
| 190 |
+
yield delta
|
| 191 |
+
except:
|
| 192 |
continue
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 193 |
|
| 194 |
# =========================
|
| 195 |
# ๐ Chat Endpoint (Text + Voice)
|
|
|
|
| 199 |
print("\n" + "="*60)
|
| 200 |
print(f"[REQUEST] ๐จ New request at {datetime.now().strftime('%H:%M:%S')}")
|
| 201 |
|
|
|
|
| 202 |
if "audio" in request.files:
|
| 203 |
+
# ๐ค Voice Mode
|
| 204 |
audio = request.files["audio"]
|
| 205 |
temp = f"/tmp/{time.time()}_{random.randint(1000,9999)}.wav"
|
| 206 |
audio.save(temp)
|
|
|
|
|
|
|
| 207 |
user_text = transcribe_audio(temp)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
|
| 209 |
keywords = ["hotel", "mall", "resort", "villa", "tempat wisata", "restaurant", "cafe"]
|
| 210 |
has_keyword = any(k in user_text.lower() for k in keywords)
|
|
|
|
| 211 |
if has_keyword:
|
|
|
|
| 212 |
serp = serpapi_search(user_text)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 213 |
text = "\n".join([f"{r['title']} โ {r['snippet']} โ {r['link']}" for r in serp["results"]])
|
| 214 |
imgs = " ".join(serp["images"])
|
| 215 |
user_text = f"{user_text}\n\nGoogle Results:\n{text}\n\nImages: {imgs}\n\nExplain & recommend."
|
|
|
|
|
|
|
|
|
|
| 216 |
|
|
|
|
| 217 |
ai = "".join(chunk for chunk in stream_chat(user_text))
|
|
|
|
|
|
|
|
|
|
| 218 |
audio_bytes = text_to_speech(ai)
|
| 219 |
|
| 220 |
+
# Debug final JSON
|
| 221 |
+
debug_json = {
|
| 222 |
"mode": "voice",
|
| 223 |
"transcript": user_text,
|
| 224 |
"reply_text": ai,
|
| 225 |
"audio_base64": "data:audio/mp3;base64," + base64.b64encode(audio_bytes).decode()
|
| 226 |
+
}
|
| 227 |
+
print("[DEBUG] ๐น Final JSON to frontend:")
|
| 228 |
+
print(json.dumps(debug_json, indent=2))
|
| 229 |
+
|
| 230 |
+
return jsonify(debug_json)
|
| 231 |
|
| 232 |
# ๐ฌ Text Mode
|
|
|
|
| 233 |
data = request.get_json(force=True)
|
| 234 |
prompt = data.get("prompt", "")
|
| 235 |
history = data.get("history", [])
|
|
|
|
|
|
|
|
|
|
| 236 |
|
| 237 |
keywords = ["hotel", "mall", "resort", "villa", "tempat wisata", "restaurant", "cafe"]
|
| 238 |
has_keyword = any(k in prompt.lower() for k in keywords)
|
|
|
|
| 239 |
if has_keyword:
|
|
|
|
| 240 |
serp = serpapi_search(prompt)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 241 |
text = "\n".join([f"{r['title']} โ {r['snippet']} โ {r['link']}" for r in serp["results"]])
|
| 242 |
imgs = " ".join(serp["images"])
|
| 243 |
prompt = f"{prompt}\n\nGoogle Results:\n{text}\n\nImages: {imgs}\n\nExplain & recommend."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 244 |
|
| 245 |
def generate():
|
| 246 |
for chunk in stream_chat(prompt, history):
|
| 247 |
yield chunk
|
|
|
|
| 248 |
|
| 249 |
return Response(generate(), mimetype="text/plain")
|
| 250 |
|