Update app.py

app.py

@@ -13,6 +13,7 @@ import traceback
 
 import gradio as gr
 import spaces
+from gradio_client import Client
 
 
 # ============================================================
@@ -20,26 +21,62 @@ import spaces
 # ============================================================
 
 APP_NAME = "X-RUDRA"
-VERSION = "3.2.1"   # bumped version
+VERSION = "3.3.1"
 
+# M1 and M2 Spaces – change these to your own if needed
 M1_REPO = os.getenv("M1_REPO", "Shrijanagain/M1")
 M2_REPO = os.getenv("M2_REPO", "Shrijanagain/M2")
 PORT = int(os.getenv("PORT", "7860"))
 
+# ------------------------------------------------------------
+# HF_TOKEN is recommended to avoid rate limits
+# Set it as a Secret in your Space settings.
+# ------------------------------------------------------------
+HF_TOKEN = os.getenv("HF_TOKEN")
+if HF_TOKEN:
+    os.environ["HF_TOKEN"] = HF_TOKEN  # ensures the transformers lib uses it
+
+
+# ============================================================
+# GRADIO CLIENTS FOR M1 / M2 (LAZY)
+# ============================================================
+
+_M1_CLIENT = None
+_M2_CLIENT = None
+
+def get_m1_client():
+    global _M1_CLIENT
+    if _M1_CLIENT is None:
+        try:
+            # Construct the public URL of the Space
+            url = f"https://{M1_REPO.replace('/', '-')}.hf.space"
+            _M1_CLIENT = Client(url)
+        except Exception as e:
+            print(f"Could not connect to M1: {e}")
+            _M1_CLIENT = None
+    return _M1_CLIENT
+
+def get_m2_client():
+    global _M2_CLIENT
+    if _M2_CLIENT is None:
+        try:
+            url = f"https://{M2_REPO.replace('/', '-')}.hf.space"
+            _M2_CLIENT = Client(url)
+        except Exception as e:
+            print(f"Could not connect to M2: {e}")
+            _M2_CLIENT = None
+    return _M2_CLIENT
+
 
 # ============================================================
 # CASUAL QUERY DETECTOR
 # ============================================================
 
 def is_casual_query(text: str) -> bool:
-    """Return True if the query is casual/timepass, False if it's a serious research question."""
+    """Return True if the query is casual/timepass."""
     text = text.lower().strip()
-    
-    # Very short queries (1‑2 words) are almost always casual
     if len(text.split()) <= 2:
         return True
-    
-    # List of casual patterns (expand as needed)
     casual_patterns = [
         "hey", "hi", "hello", "yo", "what's up", "how are you",
         "good morning", "good evening", "good night",
@@ -47,17 +84,14 @@ def is_casual_query(text: str) -> bool:
         "tell me a joke", "sing a song", "what's your name",
         "who are you", "what can you do", "help", "thanks"
     ]
-    
-    # Check if the query starts with or contains any casual pattern
     for pattern in casual_patterns:
         if text.startswith(pattern) or pattern in text:
             return True
-    
     return False
 
 
 # ============================================================
-# LAZY ENGINE
+# LAZY ENGINE (for serious queries)
 # ============================================================
 
 _ENGINE = None
@@ -71,27 +105,7 @@ def get_engine():
 
 
 # ============================================================
-# SAFE HELPERS
-# ============================================================
-
-def safe_dict(value):
-    if isinstance(value, dict):
-        return value
-    if hasattr(value, "model_dump"):
-        try:
-            return value.model_dump()
-        except Exception:
-            pass
-    if hasattr(value, "dict"):
-        try:
-            return value.dict()
-        except Exception:
-            pass
-    return {"result": str(value)}
-
-
-# ============================================================
-# FORMATTERS
+# FORMATTERS (Sources, Evidence, Verification)
 # ============================================================
 
 def format_sources(sources):
@@ -161,7 +175,6 @@ def format_verification(contradictions):
 
 
 def extract_answer(data):
-    # Try common fields
     for key in ("final_answer", "answer", "response", "final", "synthesis", "summary"):
         val = data.get(key)
         if isinstance(val, str) and val.strip():
@@ -208,7 +221,71 @@ def build_activity(data, elapsed_ms):
 
 
 # ============================================================
-# RESEARCH FUNCTION – WITH CASUAL FILTER
+# SAFE DICT HELPER
+# ============================================================
+
+def safe_dict(value):
+    if isinstance(value, dict):
+        return value
+    if hasattr(value, "model_dump"):
+        try:
+            return value.model_dump()
+        except Exception:
+            pass
+    if hasattr(value, "dict"):
+        try:
+            return value.dict()
+        except Exception:
+            pass
+    return {"result": str(value)}
+
+
+# ============================================================
+# GET MODEL RESPONSE FOR CASUAL QUERIES
+# ============================================================
+
+def get_casual_model_response(query: str) -> str:
+    """Call M1 (or M2) to generate a friendly reply for casual queries."""
+    # Try M1 first
+    client = get_m1_client()
+    if client is not None:
+        try:
+            # Assuming the endpoint is /generate with inputs: prompt, max_tokens, temperature
+            result = client.predict(
+                prompt=f"User: {query}\nAssistant:",
+                max_tokens=64,
+                temperature=0.7,
+                api_name="/generate"
+            )
+            if result and isinstance(result, str) and result.strip():
+                return result.strip()
+        except Exception as e:
+            print(f"M1 casual call failed: {e}")
+
+    # Try M2
+    client = get_m2_client()
+    if client is not None:
+        try:
+            result = client.predict(
+                prompt=f"User: {query}\nAssistant:",
+                max_tokens=64,
+                temperature=0.7,
+                api_name="/generate"
+            )
+            if result and isinstance(result, str) and result.strip():
+                return result.strip()
+        except Exception as e:
+            print(f"M2 casual call failed: {e}")
+
+    # Ultimate fallback
+    return (
+        f"👋 Hi there! I'm X‑RUDRA, your research assistant. "
+        f"How can I help you today? (Your message `{query}` was casual, so I kept it light.)"
+    )
+
+
+# ============================================================
+# MAIN RESEARCH FUNCTION
 # ============================================================
 
 async def do_research(question, max_results, max_rounds, use_models, freshness):
@@ -218,24 +295,18 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
     question = str(question).strip()
 
     # ------------------------------------------------------------
-    # 1. CASUAL FILTER – skip expensive search for timepass
+    # CASUAL QUERY – get a model‑generated reply (no web search)
     # ------------------------------------------------------------
     if is_casual_query(question):
-        casual_response = (
-            f"👋 Hey there! I'm X‑RUDRA, your research assistant. "
-            f"I'm designed to help with serious questions, deep dives, and fact‑finding. "
-            f"If you have a specific topic you'd like me to research, just ask!\n\n"
-            f"*(Your message `{question}` seemed casual, so I skipped the heavy research.)*"
-        )
+        answer = get_casual_model_response(question)
         history = [
             {"role": "user", "content": question},
-            {"role": "assistant", "content": casual_response}
+            {"role": "assistant", "content": answer}
         ]
-        empty_activity = "⚡ Skipped research (casual query)."
-        return history, empty_activity, "", "", ""
+        return history, "⚡ Casual chat (model reply, no search).", "", "", ""
 
     # ------------------------------------------------------------
-    # 2. SERIOUS QUERY – run the engine
+    # SERIOUS QUERY – run the engine (web + models)
     # ------------------------------------------------------------
     started = time.perf_counter()
 
@@ -251,15 +322,13 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
         data = safe_dict(report)
         elapsed_ms = int((time.perf_counter() - started) * 1000)
 
-        # Debug: print raw data to logs
+        # Debug (optional)
         print("\n" + "="*60)
-        print("RAW ENGINE DATA:")
+        print("RAW ENGINE DATA (first 3000 chars):")
         print(json.dumps(data, indent=2, default=str)[:3000])
         print("="*60 + "\n")
 
-        # ------------------------------------------------------------
-        # 3. EXTRACT SOURCES – convert 'results' to 'sources' if needed
-        # ------------------------------------------------------------
+        # Convert 'results' to 'sources' if needed
         sources = data.get("sources", [])
         if not sources:
             results = data.get("results", [])
@@ -275,9 +344,7 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
                     })
             data["sources"] = sources
 
-        # ------------------------------------------------------------
-        # 4. EXTRACT ANSWER – if missing, synthesize from sources
-        # ------------------------------------------------------------
+        # Extract or synthesise answer
         answer = extract_answer(data)
         if answer is None:
             if sources:
@@ -291,9 +358,6 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
             else:
                 answer = "No information found. Try rephrasing your question."
 
-        # ------------------------------------------------------------
-        # 5. BUILD OUTPUTS
-        # ------------------------------------------------------------
         claims = data.get("claims", [])
         sources_md = format_sources(sources)
         evidence_md = format_evidence(claims)
@@ -344,7 +408,7 @@ def health_check():
 
 
 # ============================================================
-# CSS AND UI
+# CSS
 # ============================================================
 
 CSS = """
@@ -358,6 +422,11 @@ body { background: #f7f7f8; }
 footer { display: none !important; }
 """
 
+
+# ============================================================
+# GRADIO UI
+# ============================================================
+
 with gr.Blocks(title=APP_NAME) as demo:
     gr.HTML("""
         <div id="header">
@@ -433,4 +502,8 @@ if __name__ == "__main__":
     print("M1:", M1_REPO)
     print("M2:", M2_REPO)
     print("Lazy engine initialization: ON")
+    if HF_TOKEN:
+        print("HF_TOKEN set – rate limits reduced.")
+    else:
+        print("HF_TOKEN not set – you may experience rate limits. Set it as a Secret in your Space.")
     demo.launch(server_name="0.0.0.0", server_port=PORT, css=CSS, show_error=True)