Update app.py

app.py

@@ -1,16 +1,12 @@
 #!/usr/bin/env python3
 """
-Overthinker v24 — Gradio Server Backend
+Overthinker — Gradio Server Backend
 
 Dual Prompt Architecture:
 - Input nodes -> prompts to generate OPTIONS/CHOICES/DECISIONS
 - Outcome nodes -> prompts to generate OUTCOMES/CONSEQUENCES
-- Model: nvidia/nemotron-3-nano-30b-a3b
+- Model: nvidia/nemotron-3-nano-30b-a3b via OpenRouter
 Port: 7860
-
-Toggle:
-- USE_HUGGINGFACE = True  -> Uses HuggingFace Inference Client (HF_TOKEN from .env)
-- USE_HUGGINGFACE = False -> Uses OpenRouter/OpenAI fallback (original behavior)
 """
 
 import os
@@ -30,38 +26,20 @@ from gradio import Server
 from fastapi import HTTPException
 from starlette.responses import HTMLResponse
 
-# ────────────────────────────────────────────────
-# HuggingFace Inference Client (optional import)
-# ────────────────────────────────────────────────
-HF_AVAILABLE = False
-try:
-    from huggingface_hub import InferenceClient
-    HF_AVAILABLE = True
-except ImportError:
-    InferenceClient = None
-    print("[Warning] huggingface_hub not installed. HuggingFace mode disabled.")
-
 load_dotenv()
 
 OPENROUTER_API_KEY = os.getenv('OPENROUTER_API_KEY', '')
-OPENAI_API_KEY = os.getenv('OPENAI_API_KEY', '')
-HF_TOKEN = os.getenv('HF_TOKEN', '')
-
-# ────────────────────────────────────────────────
-# TOGGLE: Set True to use HuggingFace Inference Client
-# ────────────────────────────────────────────────
-USE_HUGGINGFACE = False  # <-- Set to False to use OpenRouter/OpenAI instead
 
+# ---------------------------------------------------------------------------
+# Application Setup
+# ---------------------------------------------------------------------------
 app = Server()
 PORT = 7860
-if USE_HUGGINGFACE:
-    DEFAULT_MODEL = "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4"
-else:
-    DEFAULT_MODEL = "nvidia/nemotron-3-nano-30b-a3b"
+DEFAULT_MODEL = "nvidia/nemotron-3-nano-30b-a3b"
 
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 # Node Manager
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 class NodeManager:
     def __init__(self):
         self.trees: Dict[str, Dict[str, Any]] = {}
@@ -123,9 +101,9 @@ class NodeManager:
 
 node_manager = NodeManager()
 
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 # History Manager
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 class HistoryManager:
     def __init__(self):
         self.history: Dict[str, Dict] = {}
@@ -155,105 +133,49 @@ class HistoryManager:
 
 history_manager = HistoryManager()
 
-# ────────────────────────────────────────────────
-# LLM API Call — with HuggingFace Toggle
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
+# LLM API Call — OpenRouter
+# ---------------------------------------------------------------------------
 def call_api(prompt: str, system_prompt: str = "You are a helpful assistant that generates decision trees.") -> Optional[str]:
-    
-    # ─── HuggingFace Inference Client Mode ───
-    if USE_HUGGINGFACE:
-        if not HF_TOKEN:
-            print("[HF Error] No HF_TOKEN found in environment variables")
-            return None
-        if not HF_AVAILABLE:
-            print("[HF Error] huggingface_hub not installed. Run: pip install huggingface-hub>=0.23.0")
-            return None
-        
-        try:
-            client = InferenceClient(token=HF_TOKEN)
-            
-            response = client.chat.completions.create(
-                model=DEFAULT_MODEL,
-                messages=[
-                    {"role": "system", "content": system_prompt},
-                    {"role": "user", "content": prompt}
-                ],
-                temperature=0.8,
-                max_tokens=2048
-            )
-            
-            return response.choices[0].message.content
-            
-        except Exception as e:
-            print(f"[HF Exception] {e}")
-            return None
+    if not OPENROUTER_API_KEY:
+        print("[OpenRouter Error] No API key configured")
+        return None
 
-    # ─── OpenRouter Mode (original fallback) ───
-    if OPENROUTER_API_KEY:
-        try:
-            headers = {
-                'Authorization': f'Bearer {OPENROUTER_API_KEY}',
-                'Content-Type': 'application/json',
-                'HTTP-Referer': 'http://localhost:7860',
-                'X-Title': 'Overthinker v20'
-            }
-            data = {
-                'model': DEFAULT_MODEL,
-                'messages': [
-                    {'role': 'system', 'content': system_prompt},
-                    {'role': 'user', 'content': prompt}
-                ],
-                'temperature': 0.8,
-                'max_tokens': 2048
-            }
-            response = requests.post(
-                'https://openrouter.ai/api/v1/chat/completions',
-                headers=headers,
-                json=data,
-                timeout=30
-            )
-            if response.status_code == 200:
-                result = response.json()
-                return result['choices'][0]['message']['content']
-            else:
-                print(f"[OpenRouter Error] {response.status_code}: {response.text}")
-        except Exception as e:
-            print(f"[OpenRouter Exception] {e}")
-
-    if OPENAI_API_KEY:
-        try:
-            headers = {
-                'Authorization': f'Bearer {OPENAI_API_KEY}',
-                'Content-Type': 'application/json'
-            }
-            data = {
-                'model': 'gpt-3.5-turbo',
-                'messages': [
-                    {'role': 'system', 'content': system_prompt},
-                    {'role': 'user', 'content': prompt}
-                ],
-                'temperature': 0.8,
-                'max_tokens': 2048
-            }
-            response = requests.post(
-                'https://api.openai.com/v1/chat/completions',
-                headers=headers,
-                json=data,
-                timeout=30
-            )
-            if response.status_code == 200:
-                result = response.json()
-                return result['choices'][0]['message']['content']
-            else:
-                print(f"[OpenAI Error] {response.status_code}: {response.text}")
-        except Exception as e:
-            print(f"[OpenAI Exception] {e}")
+    try:
+        headers = {
+            'Authorization': f'Bearer {OPENROUTER_API_KEY}',
+            'Content-Type': 'application/json',
+            'HTTP-Referer': 'http://localhost:7860',
+            'X-Title': 'Overthinker'
+        }
+        data = {
+            'model': DEFAULT_MODEL,
+            'messages': [
+                {'role': 'system', 'content': system_prompt},
+                {'role': 'user', 'content': prompt}
+            ],
+            'temperature': 0.8,
+            'max_tokens': 2048
+        }
+        response = requests.post(
+            'https://openrouter.ai/api/v1/chat/completions',
+            headers=headers,
+            json=data,
+            timeout=30
+        )
+        if response.status_code == 200:
+            result = response.json()
+            return result['choices'][0]['message']['content']
+        else:
+            print(f"[OpenRouter Error] {response.status_code}: {response.text}")
+    except Exception as e:
+        print(f"[OpenRouter Exception] {e}")
 
     return None
 
-# ────────────────────────────────────────────────
-# Fallback generators (two distinct fallbacks)
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
+# Fallback Generators
+# ---------------------------------------------------------------------------
 def _fallback_options(decision: str, context: str = "") -> dict:
     """Generate fallback children that read as options/choices."""
     import random
@@ -330,12 +252,12 @@ def _fallback_outcomes(decision: str, context: str = "") -> dict:
     labels = [pos, neu, neg]
     desc_map = {
         'positive': f'If this path unfolds favorably, {pos.lower()}. This represents a best-case scenario where your decision leads to growth and improvement.',
-        'neutral': f'On this path, {neu.lower()}. The outcome is neither clearly good nor bad \u2014 it requires careful monitoring.',
+        'neutral': f'On this path, {neu.lower()}. The outcome is neither clearly good nor bad — it requires careful monitoring.',
         'negative': f'In a challenging scenario, {neg.lower()}. This represents potential risks and difficulties that may arise.'
     }
     tip_map = {
         'positive': 'Nurture this positive outcome by staying engaged and proactive.',
-        'neutral': 'Monitor this neutral path closely \u2014 small changes can shift the outcome.',
+        'neutral': 'Monitor this neutral path closely — small changes can shift the outcome.',
         'negative': 'Prepare contingency plans to mitigate this risk if it materializes.'
     }
     children = []
@@ -353,9 +275,9 @@ def _fallback_outcomes(decision: str, context: str = "") -> dict:
         children.append(child)
     return {'children': children}
 
-# ────────────────────────────────────────────────
-# Prompt Builders \u2014 Dual prompts
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
+# Prompt Builders — Dual prompts
+# ---------------------------------------------------------------------------
 def build_root_prompt(decision: str) -> str:
     return f'''You are an AI that helps people explore decisions by generating decision trees.
 
@@ -380,14 +302,13 @@ def build_options_prompt(decision_label: str, decision_desc: str, count: int, co
     return f'''You are an AI that helps explore decisions by generating decision tree branches.
 
 Parent node: "{decision_label}"
-Description: "{decision_desc}"
-{comment_section}
+Description: "{decision_desc}"{comment_section}
 
-Generate EXACTLY {count} child nodes that represent different OPTIONS or CHOICES the person could take. Each child should be a distinct, realistic possibility \u2014 an actionable decision they could make at this point.
+Generate EXACTLY {count} child nodes that represent different OPTIONS or CHOICES the person could take. Each child should be a distinct, realistic possibility — an actionable decision they could make at this point.
 
 IMPORTANT: Frame each child as an OPTION or CHOICE, not as an outcome. For example:
 - GOOD: "Consult a financial advisor" (describes an action/choice)
-- BAD: "Financial situation improves" (describes an outcome \u2014 DO NOT use here)
+- BAD: "Financial situation improves" (describes an outcome — DO NOT use here)
 
 Return ONLY valid JSON with exactly this structure (no markdown, no backticks):
 {{
@@ -409,14 +330,13 @@ Ensure children have unique IDs like child_1, child_2, etc.'''
 def build_outcomes_prompt(decision_label: str, decision_desc: str, count: int, comment: str = "") -> str:
     """
     Prompt for generating OUTCOMES/CONSEQUENCES (for outcome-type nodes).
-    Now requests a diverse range: positive, neutral, and negative outcomes.
+    Requests a diverse range: positive, neutral, and negative outcomes.
     """
     comment_section = f'\nUser context/comment: "{comment}"' if comment else ''
     return f'''You are an AI that helps explore decisions by generating decision tree branches.
 
 Parent node: "{decision_label}"
-Description: "{decision_desc}"
-{comment_section}
+Description: "{decision_desc}"{comment_section}
 
 Generate EXACTLY {count} child nodes that represent a DIVERSE RANGE of possible OUTCOMES or CONSEQUENCES that could naturally follow from this decision path. Include a MIX of positive, neutral, and negative outcomes so the user sees the full spectrum of possibilities.
 
@@ -426,7 +346,7 @@ Generate EXACTLY {count} child nodes that represent a DIVERSE RANGE of possible
 
 IMPORTANT: Frame each child as an OUTCOME or CONSEQUENCE, not as a choice someone makes. For example:
 - GOOD: "Financial stability improves" (describes a result)
-- BAD: "Consider financial planning" (describes a choice \u2014 DO NOT do this)
+- BAD: "Consider financial planning" (describes a choice — DO NOT do this)
 
 Return ONLY valid JSON with exactly this structure (no markdown, no backticks):
 {{
@@ -463,10 +383,9 @@ def parse_json_response(text: str) -> Optional[dict]:
         print(f"[Raw text] {text[:500]}")
         return None
 
-
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 # Routes
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 
 @app.get("/")
 async def index():
@@ -474,7 +393,7 @@ async def index():
     if os.path.exists(html_path):
         with open(html_path, "r", encoding="utf-8") as f:
             return HTMLResponse(content=f.read(), status_code=200)
-    return HTMLResponse(content="<h1>Overthinker v24</h1><p>index.html not found</p>", status_code=404)
+    return HTMLResponse(content="<h1>Overthinker</h1><p>index.html not found</p>", status_code=404)
 
 
 @app.post("/create_tree")
@@ -522,27 +441,25 @@ async def get_children(node_id: str, count: int = 3, node_type: str = "outcome",
     parent = node_manager.get_node(tree_id, node_id)
     if not parent:
         raise HTTPException(status_code=404, detail="Parent node not found")
-    
+
     # Determine what type of children to generate (input -> options; outcome -> outcomes)
     next_type_map = {'root': 'input', 'input': 'outcome', 'outcome': 'input'}
     next_type = next_type_map.get(node_type, 'outcome')
-    
+
     parent_label = parent.get('label', 'Unknown')
     parent_desc = parent.get('description', '')
-    
+
     # Select the appropriate prompt and fallback based on next_type
     if next_type == 'input':
-        # Generate options/choices
         prompt = build_options_prompt(parent_label, parent_desc, count, comment or "")
         fallback_func = _fallback_options
     else:
-        # Generate outcomes
         prompt = build_outcomes_prompt(parent_label, parent_desc, count, comment or "")
         fallback_func = _fallback_outcomes
-    
+
     ai_response = call_api(prompt)
     parsed = parse_json_response(ai_response) if ai_response else None
-    
+
     if parsed and 'children' in parsed:
         children = parsed['children']
         for i, child in enumerate(children):
@@ -556,7 +473,7 @@ async def get_children(node_id: str, count: int = 3, node_type: str = "outcome",
         node_manager.add_children(tree_id, node_id, children)
         history_manager.push_state(tree_id, {'action': 'expand', 'node_id': node_id, 'children': children})
         return {'children': children}
-    
+
     # Fallback
     fallback = fallback_func(parent_label, parent_desc)
     children = fallback.get('children', [])
@@ -583,23 +500,23 @@ async def add_options(request: dict):
     parent = node_manager.get_node(tree_id, node_id)
     if not parent:
         raise HTTPException(status_code=404, detail="Parent node not found")
-    
+
     parent_label = parent.get('label', 'Unknown')
     parent_desc = parent.get('description', '')
     node_type = parent.get('type', 'root')
     next_type_map = {'root': 'input', 'input': 'outcome', 'outcome': 'input'}
     next_type = next_type_map.get(node_type, 'outcome')
-    
+
     if next_type == 'input':
         prompt = build_options_prompt(parent_label, parent_desc, count)
         fallback_func = _fallback_options
     else:
         prompt = build_outcomes_prompt(parent_label, parent_desc, count)
         fallback_func = _fallback_outcomes
-    
+
     ai_response = call_api(prompt)
     parsed = parse_json_response(ai_response) if ai_response else None
-    
+
     if parsed and 'children' in parsed:
         children = parsed['children']
         for i, child in enumerate(children):
@@ -613,7 +530,7 @@ async def add_options(request: dict):
         node_manager.add_children(tree_id, node_id, children)
         history_manager.push_state(tree_id, {'action': 'add', 'node_id': node_id, 'children': children})
         return {'children': children}
-    
+
     fallback = fallback_func(parent_label, parent_desc)
     children = fallback.get('children', [])
     for i, child in enumerate(children):
@@ -707,31 +624,18 @@ async def export_path_md(request: dict):
         md += '\n'
     return md
 
-
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 # Launch
-# ────────────────────────────────────────────────
+# ---------------------------------------------------------------------------
 if __name__ == "__main__":
-    mode_str = "HuggingFace Inference" if USE_HUGGINGFACE else "OpenRouter/OpenAI"
-    print(f"\U0001f9e0 Overthinker v24 \u2014 Starting on port {PORT}")
+    print(f"\U0001f9e0 Overthinker \u2014 Starting on port {PORT}")
     print(f"\U0001f916 Model: {DEFAULT_MODEL}")
-    print(f"\U0001f500 Inference Mode: {mode_str}")
     print(f"\U0001f310 Open http://localhost:{PORT} in your browser")
-    
-    if USE_HUGGINGFACE:
-        if not HF_TOKEN:
-            print("\u26a0\ufe0f  No HF_TOKEN found. Set it in your .env file.")
-        elif not HF_AVAILABLE:
-            print("\u26a0\ufe0f  huggingface_hub not installed. Run: pip install huggingface-hub>=0.23.0")
-        else:
-            print(f"\u2705 HuggingFace Inference Client ready (token: {HF_TOKEN[:6]}...{HF_TOKEN[-4:]})")
-    else:
-        if not OPENROUTER_API_KEY and not OPENAI_API_KEY:
-            print("\u26a0\ufe0f  No API key found. Using local fallback generation (limited).")
-    
+    if not OPENROUTER_API_KEY:
+        print("\u26a0\ufe0f  No OPENROUTER_API_KEY found. Add it to your .env file. Fallback generation will be used.")
     print(f"\U0001f4da API docs available at http://localhost:{PORT}/docs")
     app.launch(
         server_port=PORT,
         show_error=True,
         share=False
-    )
+    )