Upload 2 files

app.py

@@ -308,14 +308,19 @@ def run_sympy(problem: str) -> dict:
     return default
 
 
-# ─── HF Inference API (completely free, no key needed) ──────────────────────────
+# ─── Groq API (100% free, fast, reliable) ───────────────────────────────────────
 def ask_claude(problem: str, sympy_info: dict, history: list) -> str:
-    """
-    Uses Hugging Face Inference API with Mistral-7B — 100% free, no API key needed.
-    HF_TOKEN is optional but recommended (get free at huggingface.co/settings/tokens).
-    """
     import requests
 
+    api_key = os.environ.get("GROQ_API_KEY", "")
+    if not api_key:
+        return (
+            "⚠️ **GROQ_API_KEY not set.**\n\n"
+            "1. Go to https://console.groq.com\n"
+            "2. Sign up free → API Keys → Create API Key\n"
+            "3. HF Space → Settings → Secrets → add GROQ_API_KEY"
+        )
+
     sympy_context = ""
     if (sympy_info.get("result")
             and sympy_info["result"] not in (None, "matrix_detected", "mod_detected")):
@@ -323,74 +328,51 @@ def ask_claude(problem: str, sympy_info: dict, history: list) -> str:
             f"SYMPY VERIFIED RESULT (your final answer MUST match this):\n"
             f"- Operation: {sympy_info.get('type', 'N/A')}\n"
             f"- Result: {sympy_info.get('result', 'N/A')}\n"
-            f"- LaTeX: {sympy_info.get('latex', 'N/A')}\n\n"
+            f"- LaTeX: {sympy_info.get('latex', 'N/A')}\n"
         )
 
-    system_text = (
+    system_prompt = (
         "You are SAAD AI, an elite BSc Mathematics tutor. "
-        "Solve every problem step by step. "
-        "Label every step as Step 1:, Step 2:, etc. "
+        "Solve every problem with perfect step-by-step working. "
+        "Label every step: Step 1:, Step 2:, etc. "
         "Use $...$ for inline math and $$...$$ for display equations. "
-        "State the method/theorem used at the start. "
-        "Show ALL intermediate algebra — never skip steps. "
-        "End with: **Final Answer:** $$...$$ "
+        "State the method or theorem used at the start. "
+        "Show ALL intermediate algebra, never skip steps. "
+        "End every solution with: **Final Answer:** $$...$$ "
         "NEVER make arithmetic errors. "
         "Topics: Calculus, Linear Algebra, Number Theory, ODEs, "
         "Numerical Methods, Differential Geometry, Hydro Mechanics, "
-        "Theory of Numbers, Real Analysis II, General Math. "
+        "Theory of Numbers, Real Analysis II, General Math.\n\n"
         + sympy_context
     )
 
-    # Build prompt with history (last 4 exchanges)
-    conversation = ""
+    messages = [{"role": "system", "content": system_prompt}]
     for msg in history[-8:]:
-        role = "User" if msg["role"] == "user" else "Assistant"
-        conversation += f"{role}: {msg['content']}\n"
-    conversation += f"User: {problem}\nAssistant:"
-
-    full_prompt = f"<s>[INST] {system_text}\n\n{conversation} [/INST]"
-
-    # HF_TOKEN is free — get at huggingface.co/settings/tokens
-    # Without it, rate limits are stricter but it still works
-    hf_token = os.environ.get("HF_TOKEN", "")
-    headers = {"Authorization": f"Bearer {hf_token}"} if hf_token else {}
-
-    # Mistral-7B-Instruct — excellent at math, free on HF
-    API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.3"
+        messages.append({"role": msg["role"], "content": msg["content"]})
+    messages.append({"role": "user", "content": problem})
 
+    headers = {
+        "Authorization": f"Bearer {api_key}",
+        "Content-Type": "application/json"
+    }
     payload = {
-        "inputs": full_prompt,
-        "parameters": {
-            "max_new_tokens": 1500,
-            "temperature": 0.2,       # low temp = more precise math
-            "repetition_penalty": 1.1,
-            "return_full_text": False
-        }
+        "model": "llama-3.3-70b-versatile",
+        "messages": messages,
+        "max_tokens": 2048,
+        "temperature": 0.2
     }
 
     try:
-        resp = requests.post(API_URL, headers=headers, json=payload, timeout=60)
+        resp = requests.post(
+            "https://api.groq.com/openai/v1/chat/completions",
+            headers=headers,
+            json=payload,
+            timeout=60
+        )
         resp.raise_for_status()
-        data = resp.json()
-
-        if isinstance(data, list) and len(data) > 0:
-            return data[0].get("generated_text", "No response generated.")
-        elif isinstance(data, dict) and "error" in data:
-            # Model is loading — happens first time, takes ~20 seconds
-            if "loading" in str(data["error"]).lower():
-                return (
-                    "⏳ **Model is loading** (first request takes ~20 seconds).\n\n"
-                    "Please click **SOLVE** again in a few seconds."
-                )
-            return f"⚠️ HF API Error: {data['error']}"
-        else:
-            return "⚠️ Unexpected response format from HF API."
-
+        return resp.json()["choices"][0]["message"]["content"]
     except requests.exceptions.Timeout:
-        return (
-            "⏳ **Request timed out.** The model may be loading.\n\n"
-            "Please click **SOLVE** again in a few seconds."
-        )
+        return "⏳ Request timed out. Please try again."
     except Exception as e:
         return f"⚠️ Error: {str(e)}"