Update generator_engine.py

generator_engine.py

@@ -1,33 +1,73 @@
 from __future__ import annotations
 
-from typing import Optional
+from typing import List, Optional
 
-try:
-    from transformers import pipeline
-except Exception:
-    pipeline = None
+from models import RetrievedChunk
 
 
 class GeneratorEngine:
-    def __init__(self, model_name: str = "google/flan-t5-small"):
-        self.model_name = model_name
-        self.pipe = None
-        if pipeline is not None:
-            try:
-                self.pipe = pipeline("text2text-generation", model=model_name)
-            except Exception:
-                self.pipe = None
-
-    def available(self) -> bool:
-        return self.pipe is not None
-
-    def generate(self, prompt: str, max_new_tokens: int = 96) -> Optional[str]:
-        if self.pipe is None:
-            return None
-        try:
-            out = self.pipe(prompt, max_new_tokens=max_new_tokens, do_sample=False)
-            if out and isinstance(out, list):
-                return str(out[0].get("generated_text", "")).strip()
-        except Exception:
-            return None
-        return None
+
+    def __init__(self):
+        pass
+
+    def _build_prompt(
+        self,
+        user_text: str,
+        question_text: Optional[str],
+        topic: str,
+        intent: str,
+        retrieval_context: List[RetrievedChunk],
+    ) -> str:
+
+        context_lines = []
+        for chunk in retrieval_context:
+            context_lines.append(f"- {chunk.text}")
+
+        context_block = "\n".join(context_lines)
+
+        prompt = f"""
+You are a GMAT quantitative tutor.
+
+Topic: {topic}
+Intent: {intent}
+
+Question:
+{question_text or user_text}
+
+Relevant teaching notes:
+{context_block}
+
+Explain how to approach the problem without immediately revealing the answer.
+"""
+
+        return prompt.strip()
+
+    def generate(
+        self,
+        user_text: str,
+        question_text: Optional[str] = None,
+        topic: str = "",
+        intent: str = "answer",
+        retrieval_context: Optional[List[RetrievedChunk]] = None,
+        chat_history=None,
+        **kwargs,
+    ) -> str:
+
+        retrieval_context = retrieval_context or []
+
+        prompt = self._build_prompt(
+            user_text,
+            question_text,
+            topic,
+            intent,
+            retrieval_context,
+        )
+
+        # simple fallback explanation system
+        if intent == "hint":
+            return "Focus on identifying the relationship between the quantities before calculating."
+
+        if intent in {"method", "walkthrough", "instruction"}:
+            return "Start by translating the wording of the problem into an equation or numerical relationship."
+
+        return "Let’s work through the structure of the problem before jumping to the calculation."