Update app.py
Browse files
app.py
CHANGED
|
@@ -20,13 +20,42 @@ import spaces
|
|
| 20 |
# ============================================================
|
| 21 |
|
| 22 |
APP_NAME = "X-RUDRA"
|
| 23 |
-
VERSION = "3.2.
|
| 24 |
|
| 25 |
M1_REPO = os.getenv("M1_REPO", "Shrijanagain/M1")
|
| 26 |
M2_REPO = os.getenv("M2_REPO", "Shrijanagain/M2")
|
| 27 |
PORT = int(os.getenv("PORT", "7860"))
|
| 28 |
|
| 29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
# ============================================================
|
| 31 |
# LAZY ENGINE
|
| 32 |
# ============================================================
|
|
@@ -62,7 +91,7 @@ def safe_dict(value):
|
|
| 62 |
|
| 63 |
|
| 64 |
# ============================================================
|
| 65 |
-
# FORMATTERS
|
| 66 |
# ============================================================
|
| 67 |
|
| 68 |
def format_sources(sources):
|
|
@@ -137,7 +166,6 @@ def extract_answer(data):
|
|
| 137 |
val = data.get(key)
|
| 138 |
if isinstance(val, str) and val.strip():
|
| 139 |
return val.strip()
|
| 140 |
-
# If no answer, we'll build one from results in the caller
|
| 141 |
return None
|
| 142 |
|
| 143 |
|
|
@@ -180,7 +208,7 @@ def build_activity(data, elapsed_ms):
|
|
| 180 |
|
| 181 |
|
| 182 |
# ============================================================
|
| 183 |
-
# RESEARCH FUNCTION –
|
| 184 |
# ============================================================
|
| 185 |
|
| 186 |
async def do_research(question, max_results, max_rounds, use_models, freshness):
|
|
@@ -188,6 +216,27 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
|
|
| 188 |
return [], "⚪ Enter a question to start.", "", ""
|
| 189 |
|
| 190 |
question = str(question).strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191 |
started = time.perf_counter()
|
| 192 |
|
| 193 |
try:
|
|
@@ -202,18 +251,17 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
|
|
| 202 |
data = safe_dict(report)
|
| 203 |
elapsed_ms = int((time.perf_counter() - started) * 1000)
|
| 204 |
|
| 205 |
-
# Debug: print raw data to logs
|
| 206 |
print("\n" + "="*60)
|
| 207 |
print("RAW ENGINE DATA:")
|
| 208 |
-
print(json.dumps(data, indent=2, default=str)[:3000])
|
| 209 |
print("="*60 + "\n")
|
| 210 |
|
| 211 |
# ------------------------------------------------------------
|
| 212 |
-
#
|
| 213 |
# ------------------------------------------------------------
|
| 214 |
sources = data.get("sources", [])
|
| 215 |
if not sources:
|
| 216 |
-
# Engine uses 'results' – convert each to source format
|
| 217 |
results = data.get("results", [])
|
| 218 |
for res in results:
|
| 219 |
if isinstance(res, dict):
|
|
@@ -225,15 +273,13 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
|
|
| 225 |
"source_score": res.get("rank", "N/A"),
|
| 226 |
"description": res.get("snippet", ""),
|
| 227 |
})
|
| 228 |
-
# Store back for activity and formatters
|
| 229 |
data["sources"] = sources
|
| 230 |
|
| 231 |
# ------------------------------------------------------------
|
| 232 |
-
#
|
| 233 |
# ------------------------------------------------------------
|
| 234 |
answer = extract_answer(data)
|
| 235 |
if answer is None:
|
| 236 |
-
# Build a simple answer from the top sources
|
| 237 |
if sources:
|
| 238 |
top = sources[:5]
|
| 239 |
parts = [f"Based on the top results:"]
|
|
@@ -246,15 +292,9 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
|
|
| 246 |
answer = "No information found. Try rephrasing your question."
|
| 247 |
|
| 248 |
# ------------------------------------------------------------
|
| 249 |
-
#
|
| 250 |
# ------------------------------------------------------------
|
| 251 |
claims = data.get("claims", [])
|
| 252 |
-
# (Optional) You could generate simple claims from each source's snippet,
|
| 253 |
-
# but we'll leave it empty for now.
|
| 254 |
-
|
| 255 |
-
# ------------------------------------------------------------
|
| 256 |
-
# 4. BUILD OUTPUTS
|
| 257 |
-
# ------------------------------------------------------------
|
| 258 |
sources_md = format_sources(sources)
|
| 259 |
evidence_md = format_evidence(claims)
|
| 260 |
verification_md = format_verification(data.get("contradictions", []))
|
|
@@ -304,7 +344,7 @@ def health_check():
|
|
| 304 |
|
| 305 |
|
| 306 |
# ============================================================
|
| 307 |
-
# CSS AND UI
|
| 308 |
# ============================================================
|
| 309 |
|
| 310 |
CSS = """
|
|
|
|
| 20 |
# ============================================================
|
| 21 |
|
| 22 |
APP_NAME = "X-RUDRA"
|
| 23 |
+
VERSION = "3.2.1" # bumped version
|
| 24 |
|
| 25 |
M1_REPO = os.getenv("M1_REPO", "Shrijanagain/M1")
|
| 26 |
M2_REPO = os.getenv("M2_REPO", "Shrijanagain/M2")
|
| 27 |
PORT = int(os.getenv("PORT", "7860"))
|
| 28 |
|
| 29 |
|
| 30 |
+
# ============================================================
|
| 31 |
+
# CASUAL QUERY DETECTOR
|
| 32 |
+
# ============================================================
|
| 33 |
+
|
| 34 |
+
def is_casual_query(text: str) -> bool:
|
| 35 |
+
"""Return True if the query is casual/timepass, False if it's a serious research question."""
|
| 36 |
+
text = text.lower().strip()
|
| 37 |
+
|
| 38 |
+
# Very short queries (1‑2 words) are almost always casual
|
| 39 |
+
if len(text.split()) <= 2:
|
| 40 |
+
return True
|
| 41 |
+
|
| 42 |
+
# List of casual patterns (expand as needed)
|
| 43 |
+
casual_patterns = [
|
| 44 |
+
"hey", "hi", "hello", "yo", "what's up", "how are you",
|
| 45 |
+
"good morning", "good evening", "good night",
|
| 46 |
+
"lol", "haha", "just testing", "timepass", "nothing",
|
| 47 |
+
"tell me a joke", "sing a song", "what's your name",
|
| 48 |
+
"who are you", "what can you do", "help", "thanks"
|
| 49 |
+
]
|
| 50 |
+
|
| 51 |
+
# Check if the query starts with or contains any casual pattern
|
| 52 |
+
for pattern in casual_patterns:
|
| 53 |
+
if text.startswith(pattern) or pattern in text:
|
| 54 |
+
return True
|
| 55 |
+
|
| 56 |
+
return False
|
| 57 |
+
|
| 58 |
+
|
| 59 |
# ============================================================
|
| 60 |
# LAZY ENGINE
|
| 61 |
# ============================================================
|
|
|
|
| 91 |
|
| 92 |
|
| 93 |
# ============================================================
|
| 94 |
+
# FORMATTERS
|
| 95 |
# ============================================================
|
| 96 |
|
| 97 |
def format_sources(sources):
|
|
|
|
| 166 |
val = data.get(key)
|
| 167 |
if isinstance(val, str) and val.strip():
|
| 168 |
return val.strip()
|
|
|
|
| 169 |
return None
|
| 170 |
|
| 171 |
|
|
|
|
| 208 |
|
| 209 |
|
| 210 |
# ============================================================
|
| 211 |
+
# RESEARCH FUNCTION – WITH CASUAL FILTER
|
| 212 |
# ============================================================
|
| 213 |
|
| 214 |
async def do_research(question, max_results, max_rounds, use_models, freshness):
|
|
|
|
| 216 |
return [], "⚪ Enter a question to start.", "", ""
|
| 217 |
|
| 218 |
question = str(question).strip()
|
| 219 |
+
|
| 220 |
+
# ------------------------------------------------------------
|
| 221 |
+
# 1. CASUAL FILTER – skip expensive search for timepass
|
| 222 |
+
# ------------------------------------------------------------
|
| 223 |
+
if is_casual_query(question):
|
| 224 |
+
casual_response = (
|
| 225 |
+
f"👋 Hey there! I'm X‑RUDRA, your research assistant. "
|
| 226 |
+
f"I'm designed to help with serious questions, deep dives, and fact‑finding. "
|
| 227 |
+
f"If you have a specific topic you'd like me to research, just ask!\n\n"
|
| 228 |
+
f"*(Your message `{question}` seemed casual, so I skipped the heavy research.)*"
|
| 229 |
+
)
|
| 230 |
+
history = [
|
| 231 |
+
{"role": "user", "content": question},
|
| 232 |
+
{"role": "assistant", "content": casual_response}
|
| 233 |
+
]
|
| 234 |
+
empty_activity = "⚡ Skipped research (casual query)."
|
| 235 |
+
return history, empty_activity, "", "", ""
|
| 236 |
+
|
| 237 |
+
# ------------------------------------------------------------
|
| 238 |
+
# 2. SERIOUS QUERY – run the engine
|
| 239 |
+
# ------------------------------------------------------------
|
| 240 |
started = time.perf_counter()
|
| 241 |
|
| 242 |
try:
|
|
|
|
| 251 |
data = safe_dict(report)
|
| 252 |
elapsed_ms = int((time.perf_counter() - started) * 1000)
|
| 253 |
|
| 254 |
+
# Debug: print raw data to logs
|
| 255 |
print("\n" + "="*60)
|
| 256 |
print("RAW ENGINE DATA:")
|
| 257 |
+
print(json.dumps(data, indent=2, default=str)[:3000])
|
| 258 |
print("="*60 + "\n")
|
| 259 |
|
| 260 |
# ------------------------------------------------------------
|
| 261 |
+
# 3. EXTRACT SOURCES – convert 'results' to 'sources' if needed
|
| 262 |
# ------------------------------------------------------------
|
| 263 |
sources = data.get("sources", [])
|
| 264 |
if not sources:
|
|
|
|
| 265 |
results = data.get("results", [])
|
| 266 |
for res in results:
|
| 267 |
if isinstance(res, dict):
|
|
|
|
| 273 |
"source_score": res.get("rank", "N/A"),
|
| 274 |
"description": res.get("snippet", ""),
|
| 275 |
})
|
|
|
|
| 276 |
data["sources"] = sources
|
| 277 |
|
| 278 |
# ------------------------------------------------------------
|
| 279 |
+
# 4. EXTRACT ANSWER – if missing, synthesize from sources
|
| 280 |
# ------------------------------------------------------------
|
| 281 |
answer = extract_answer(data)
|
| 282 |
if answer is None:
|
|
|
|
| 283 |
if sources:
|
| 284 |
top = sources[:5]
|
| 285 |
parts = [f"Based on the top results:"]
|
|
|
|
| 292 |
answer = "No information found. Try rephrasing your question."
|
| 293 |
|
| 294 |
# ------------------------------------------------------------
|
| 295 |
+
# 5. BUILD OUTPUTS
|
| 296 |
# ------------------------------------------------------------
|
| 297 |
claims = data.get("claims", [])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 298 |
sources_md = format_sources(sources)
|
| 299 |
evidence_md = format_evidence(claims)
|
| 300 |
verification_md = format_verification(data.get("contradictions", []))
|
|
|
|
| 344 |
|
| 345 |
|
| 346 |
# ============================================================
|
| 347 |
+
# CSS AND UI
|
| 348 |
# ============================================================
|
| 349 |
|
| 350 |
CSS = """
|