Update planner.py

planner.py

@@ -1,34 +1,48 @@
 # planner.py
+import os
 from huggingface_hub import InferenceClient
-import re
 
-# Get your HF token from secrets (set in HF Spaces)
-import os
+# Get HF token from environment (set as Secret in HF Spaces)
 HF_TOKEN = os.getenv("HF_TOKEN")
 
+# Use the SAME model as your working chatbot
 client = InferenceClient(
-    "Qwen/Qwen2.5-Coder-7B-Instruct",
+    model="Qwen/Qwen2.5-Coder-7B-Instruct",
     token=HF_TOKEN
 )
 
 def generate_task_plan(goal: str) -> str:
-    prompt = f"""
-    Break down the following goal into actionable tasks with:
-    - Clear task descriptions
-    - Realistic deadlines (relative to today)
-    - Dependencies between tasks (if any)
-
-    Format your response as a numbered list with this structure:
-    1. [Task name] - Due: [Day X] - Depends on: [Task # or "None"]
-       Description: [Brief explanation]
-
-    Goal: "{goal}"
     """
+    Uses Qwen2.5-Coder (instruct model) via chat API to break down a goal into tasks.
+    """
+    messages = [
+        {
+            "role": "user",
+            "content": (
+                "You are a smart project planner. Break down the following goal into a clear, actionable task list.\n\n"
+                "Requirements:\n"
+                "- Number each task (1., 2., 3., ...)\n"
+                "- Include a realistic deadline as 'Due: Day X' (start from Day 1)\n"
+                "- Specify dependencies as 'Depends on: Task N' or 'None'\n"
+                "- Add a short description for each task\n"
+                "- Keep it practical and time-bound\n\n"
+                "Example format:\n"
+                "1. Research market needs - Due: Day 1 - Depends on: None\n"
+                "   Description: Interview 5 potential users about pain points.\n\n"
+                f"Goal: \"{goal}\""
+            )
+        }
+    ]
 
-    response = client.text_generation(
-        prompt,
-        max_new_tokens=2048,
-        temperature=0.3,
-        repetition_penalty=1.2
-    )
-    return response.strip()
+    try:
+        response = client.chat.completions.create(
+            model="Qwen/Qwen2.5-Coder-7B-Instruct",
+            messages=messages,
+            max_tokens=600,
+            temperature=0.4,
+            top_p=0.95,
+            stream=False
+        )
+        return response.choices[0].message.content.strip()
+    except Exception as e:
+        raise Exception(f"LLM inference failed: {str(e)}")