Update app.py

app.py

@@ -5,15 +5,7 @@
 
 # ──────────────────────────────────────────────────────────────
 #  REASONING‑ENFORCED AGENT CONTRACT (AGENT.md)
-#  This file is the operating policy for the entire assistant.
-#  For every request, the agent MUST:
-#    - Understand the goal and requirements.
-#    - Check constraints and missing information.
-#    - Decompose complex tasks and track dependencies.
-#    - Plan, execute, verify, and adapt.
-#  Private chain‑of‑thought is NEVER exposed.
-#  Only concise reasoning summaries are shown.
-#  See the full policy in the multi‑line comment below.
+#  See full policy in the multi‑line comment below.
 # ──────────────────────────────────────────────────────────────
 
 """
@@ -55,7 +47,7 @@ execution, verification, sources, or completion.
 
 ## 6. REASONING VISIBILITY
 The agent MUST reason internally but NEVER expose private chain‑of‑thought.
-Provide concise summaries (Approach, Key assumptions, Decision, Blocker, Verification, Result) only when useful.
+Provide concise summaries only when useful.
 
 ## 7. FINAL CONTRACT
 UNDERSTAND → PLAN → ACT → VERIFY → ADAPT → DELIVER
@@ -84,7 +76,7 @@ from gradio_client import Client
 # ============================================================
 
 APP_NAME = "X-RUDRA"
-VERSION = "3.7.0"   # bumped version
+VERSION = "3.7.1"   # bumped
 
 M1_REPO = os.getenv("M1_REPO", "Shrijanagain/M1")
 M2_REPO = os.getenv("M2_REPO", "Shrijanagain/M2")
@@ -226,15 +218,15 @@ Answer:"""
                 title = src.get("title", "Untitled")
                 snippet = src.get("snippet", src.get("description", ""))
                 parts.append(f"{i}. {title}: {snippet[:200]}..." if snippet else f"{i}. {title}")
-            return "\n\n".join(parts), ""
+            return "\n\n".join(parts)
         else:
-            return "I couldn't find specific information on that topic. Could you rephrase?", ""
+            return "I couldn't find specific information on that topic. Could you rephrase?"
 
     # If only one draft exists, use that as final
     if not draft_m1:
-        return draft_m2, ""
+        return draft_m2
     if not draft_m2:
-        return draft_m1, ""
+        return draft_m1
 
     # ----- 2. Merge both drafts using M2 -----
     merge_prompt = f"""Question: {question}
@@ -251,18 +243,12 @@ Final answer:"""
 
     merged = call_model(m2_client, merge_prompt, max_tokens, temperature)
     if not merged:
-        # If merging fails, fallback to draft_m1 (or draft_m2)
+        # If merging fails, fallback to draft_m1
         merged = draft_m1
 
-    # Extract thinking from <think> tags (if any)
-    thinking_content = ""
-    clean_answer = merged
-    think_match = re.search(r"<think>(.*?)</think>", merged, re.DOTALL)
-    if think_match:
-        thinking_content = think_match.group(1).strip()
-        clean_answer = re.sub(r"<think>.*?</think>", "", merged, flags=re.DOTALL).strip()
-
-    return clean_answer, thinking_content
+    # Strip any remaining <think> tags from the final answer
+    clean_answer = re.sub(r"<think>.*?</think>", "", merged, flags=re.DOTALL).strip()
+    return clean_answer
 
 
 # ============================================================
@@ -433,7 +419,7 @@ def safe_dict(value):
 
 
 # ============================================================
-# MAIN RESEARCH FUNCTION (always returns 6 outputs)
+# MAIN RESEARCH FUNCTION (returns 5 outputs – no thinking)
 # ============================================================
 
 async def do_research(question, max_results, max_rounds, use_models, freshness):
@@ -442,10 +428,9 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
     empty_sources = ""
     empty_evidence = ""
     empty_verification = ""
-    empty_thinking = ""
 
     if not question or not str(question).strip():
-        return empty_history, empty_activity, empty_sources, empty_evidence, empty_verification, empty_thinking
+        return empty_history, empty_activity, empty_sources, empty_evidence, empty_verification
 
     question = str(question).strip()
 
@@ -456,7 +441,7 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
             {"role": "user", "content": question},
             {"role": "assistant", "content": answer}
         ]
-        return history, "⚡ Casual chat (model reply, no search).", "", "", "", ""
+        return history, "⚡ Casual chat (model reply, no search).", "", "", ""
 
     # ---- Serious query ----
     started = time.perf_counter()
@@ -488,22 +473,20 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
                     })
             data["sources"] = sources
 
-        # Generate final answer and thinking (using new dual‑draft + merge)
-        final_answer, thinking_content = get_combined_model_answer(question, sources)
+        # Generate final answer (no thinking)
+        final_answer = get_combined_model_answer(question, sources)
 
         sources_md = format_sources(sources)
         evidence_md = format_evidence(data.get("claims", []))
         verification_md = format_verification(data.get("contradictions", []))
         activity_md = build_activity(data, elapsed_ms)
 
-        thinking_md = f"### 🧠 Thinking\n\n{thinking_content}" if thinking_content else ""
-
         history = [
             {"role": "user", "content": question},
             {"role": "assistant", "content": final_answer}
         ]
 
-        return history, activity_md, sources_md, evidence_md, verification_md, thinking_md
+        return history, activity_md, sources_md, evidence_md, verification_md
 
     except Exception as exc:
         error = f"❌ **X-RUDRA Error**\n\n`{type(exc).__name__}: {exc}`"
@@ -515,7 +498,7 @@ async def do_research(question, max_results, max_rounds, use_models, freshness):
             {"role": "user", "content": question},
             {"role": "assistant", "content": error}
         ]
-        return history, "❌ Research failed.", "", "", "", ""
+        return history, "❌ Research failed.", "", "", ""
 
 
 # ============================================================
@@ -542,7 +525,7 @@ def health_check():
 
 
 # ============================================================
-# CSS – includes spinner animation
+# CSS – no thinking spinner needed
 # ============================================================
 
 CSS = """
@@ -554,36 +537,11 @@ body { background: #f7f7f8; }
 #chat { border-radius: 18px; }
 #send { min-height: 52px; font-size: 18px; font-weight: 700; }
 footer { display: none !important; }
-
-@keyframes think-pulse {
-    0% { opacity: 0.3; transform: scale(0.95); }
-    50% { opacity: 1; transform: scale(1.05); }
-    100% { opacity: 0.3; transform: scale(0.95); }
-}
-.thinking-spinner {
-    display: inline-block;
-    width: 12px;
-    height: 12px;
-    border-radius: 50%;
-    background: #6b7280;
-    margin-right: 8px;
-    animation: think-pulse 1.2s ease-in-out infinite;
-}
-.thinking-container {
-    background: #f3f4f6;
-    border-left: 4px solid #6366f1;
-    padding: 12px 16px;
-    border-radius: 8px;
-    margin: 12px 0;
-    font-family: monospace;
-    white-space: pre-wrap;
-    word-wrap: break-word;
-}
 """
 
 
 # ============================================================
-# GRADIO UI – 6 outputs
+# GRADIO UI – 5 outputs (no thinking)
 # ============================================================
 
 with gr.Blocks(title=APP_NAME) as demo:
@@ -604,7 +562,6 @@ with gr.Blocks(title=APP_NAME) as demo:
         with gr.Column(scale=4):
             gr.Markdown("## 🔬 Live Research")
             activity = gr.Markdown("⚪ Waiting for your question.")
-            thinking = gr.Markdown("", visible=True)
             gr.Markdown("---")
             gr.Markdown(f"""
 ### Model Spaces
@@ -646,7 +603,7 @@ with gr.Blocks(title=APP_NAME) as demo:
     )
 
     inputs = [question, max_results, max_rounds, use_models, freshness]
-    outputs = [chatbot, activity, sources, evidence, verification, thinking]
+    outputs = [chatbot, activity, sources, evidence, verification]
 
     send.click(fn=run_research, inputs=inputs, outputs=outputs)
     question.submit(fn=run_research, inputs=inputs, outputs=outputs)