Upload folder using huggingface_hub

modules/quiz/answer_validator.py

@@ -0,0 +1,44 @@
+from openai import OpenAI
+from modules.quiz.models import AnswerValidation
+from modules.quiz.quiz_helper import Question
+
+client = OpenAI()
+
+
+def validate_answer(
+    question: Question, user_answer: str, preferred_language: str = "English"
+) -> AnswerValidation:
+    """
+    Validate a user's answer against a Question object.
+    Uses LLM reasoning to account for synonyms, variations, and explanation.
+    """
+    print("validating answer ...")
+    prompt = f"""
+    You are an answer validator for a scripture-based quiz.
+
+    Question:
+    {question.question}
+
+    Choices (if any):
+    {question.choices}
+
+    Expected Answer:
+    {question.expected_answer}
+
+    User's Answer:
+    {user_answer}
+
+    Rules:
+    - Compare strictly against expected_answer and choices.
+    - Accept semantically equivalent answers (e.g., synonyms, transliterations).
+    - Respond in {question.preferred_language}.
+    - Do not invent a new correct answer. Only check against expected_answer.
+    """
+
+    response = client.chat.completions.parse(
+        model="gpt-5-nano",
+        messages=[{"role": "user", "content": prompt}],
+        response_format=AnswerValidation,
+    )
+
+    return response.choices[0].message.parsed

modules/quiz/models.py

@@ -0,0 +1,65 @@
+from typing import List, Optional, Literal
+from pydantic import BaseModel, Field
+from sanatan_assistant import allowedScriptureTitles
+
+
+class Question(BaseModel):
+    preferred_language: str = Field(
+        description="User's preferred language. Respond in this language"
+    )
+    question: str = Field(description="The generated question to be asked to the user")
+    scripture: allowedScriptureTitles = Field(
+        description="The scripture title this question is sourced from (e.g., Bhagavad Gita, Divya Prabandham)"
+    )
+    native_lyrics: str = Field(
+        description="The lyrics in its native/original script (e.g., Sanskrit/Tamil). Sanitize this if there are garbled characters."
+    )
+    transliteration: str = Field(
+        description="The transliterated text of the verse in English/the user's preferred language"
+    )
+    translation: str = Field(
+        description="The translated meaning of the verse in English/the user's preferred language"
+    )
+    detailed_explanation: Optional[str] = Field(
+        default=None,
+        description="A detailed explanation of the verse or context, if available in the context",
+    )
+    word_by_word_meaning: Optional[str] = Field(
+        default=None,
+        description="Word-by-word meaning (in both native and english) or breakdown of the verse, if available in the context",
+    )
+    verse_number: Optional[int] = Field(
+        default=None,
+        description="The absolute verse number from the context if available.",
+    )
+    page_number: Optional[int] = Field(
+        default=None, description="The page number from the context if available."
+    )
+    verse_reference: Optional[str] = Field(
+        default=None,
+        description="The reference identifier of the verse or the title (e.g., Bhagavad Gita 2.47)",
+    )
+    reference_link: Optional[str] = Field(
+        default=None, description="The html_url if available"
+    )
+    complexity: Literal["beginner", "intermediate", "advanced"] = Field(
+        description="Difficulty level of the question"
+    )
+    choices: List[str] = Field(
+        description="List of multiple-choice options. Leave empty for open-text questions"
+    )
+    expected_answer: str = Field(
+        description="The correct or expected answer to the question"
+    )
+    answer_explanation: str = Field(
+        description="Explanation why the expected answer is correct"
+    )
+    expected_choice_index: Optional[int] = Field(
+        default=None,
+        description="Index of the correct choice in 'choices'. Null if open-text question",
+    )
+
+class AnswerValidation(BaseModel):
+    is_correct: bool = Field(description="Whether the provided answer is correct")
+    reasoning: str = Field(description="Explanation of why the answer is correct or not")
+    correct_answer: str = Field(description="The correct/expected answer from the question object")

modules/quiz/quiz_helper.py

@@ -1,83 +1,23 @@
 import random
-from typing import List, Optional, Literal
-from pydantic import BaseModel, Field
+from typing import Literal
 from config import SanatanConfig
-from sanatan_assistant import allowedScriptureTitles, query, allowedCollections
+from modules.quiz.models import Question
+from sanatan_assistant import query, allowedCollections
 from openai import OpenAI
 
 client = OpenAI()
 
 
-class Question(BaseModel):
-    preferred_language : str = Field(description="User's preferred language. Respond in this language")
-    question: str = Field(
-        description="The generated question to be asked to the user"
-    )
-    scripture: allowedScriptureTitles = Field(
-        description="The scripture title this question is sourced from (e.g., Bhagavad Gita, Divya Prabandham)"
-    )
-    native_lyrics: str = Field(
-        description="The lyrics in its native/original script (e.g., Sanskrit/Tamil). Sanitize this if there are garbled characters."
-    )
-    transliteration: str = Field(
-        description="The transliterated text of the verse in English/the user's preferred language"
-    )
-    translation: str = Field(
-        description="The translated meaning of the verse in English/the user's preferred language"
-    )
-    detailed_explanation: Optional[str] = Field(
-        default=None,
-        description="A detailed explanation of the verse or context, if available in the context"
-    )
-    word_by_word_meaning: Optional[str] = Field(
-        default=None,
-        description="Word-by-word meaning (in both native and english) or breakdown of the verse, if available in the context"
-    )
-    verse_number: Optional[int] = Field(
-        default=None,
-        description="The absolute verse number from the context if available."
-    )
-    page_number: Optional[int] = Field(
-        default=None,
-        description="The page number from the context if available."
-    )
-    verse_reference: Optional[str] = Field(
-        default=None,
-        description="The reference identifier of the verse or the title (e.g., Bhagavad Gita 2.47)"
-    )
-    reference_link: Optional[str] = Field(
-        default=None,
-        description="The html_url if available"
-    )
-    complexity: Literal["beginner", "intermediate", "advanced"] = Field(
-        description="Difficulty level of the question"
-    )
-    choices: List[str] = Field(
-        description="List of multiple-choice options. Leave empty for open-text questions"
-    )
-    expected_answer: str = Field(
-        description="The correct or expected answer to the question"
-    )
-    answer_explanation: str = Field(
-        description="Explanation why the expected answer is correct"
-    )
-    expected_choice_index: Optional[int] = Field(
-        default=None,
-        description="Index of the correct choice in 'choices'. Null if open-text question"
-    )
-
-
-
 def generate_question(
     collection: allowedCollections,
-    complexity: Literal["beginner", "intermediate", "advanced"] = "beginner",
-    mode: Literal["mcq", "open"] = "mcq",
-    preferred_lamguage : str = "English"
+    complexity: Literal["beginner", "intermediate", "advanced"],
+    mode: Literal["mcq", "open"],
+    preferred_lamguage: str = "English",
 ) -> Question:
     """
     Fetch a random scripture record and have the LLM generate a structured Question.
     """
-
+    print("Generating question ...", collection, complexity,mode, preferred_lamguage)
     # 1. Fetch random scripture record
     context = query(
         collection_name=collection,
@@ -107,9 +47,9 @@ def generate_question(
 
     # 3. Structured response with Pydantic class reference
     response = client.chat.completions.parse(
-        model="gpt-5-nano", 
+        model="gpt-5-nano",
         messages=[{"role": "user", "content": prompt}],
-        response_format=Question, 
+        response_format=Question,
     )
 
     # print(response)
@@ -120,11 +60,17 @@ def generate_question(
 # Example usage
 if __name__ == "__main__":
     for i in range(3):
-        scripture = random.choice([s for s in SanatanConfig.scriptures if s["collection_name"] != "yt_metadata"])
-        mode = random.choice(["mcq", "open"])
-        complexity = random.choice(["beginner", "intermediate", "advanced"])
         q = generate_question(
-            scripture["collection_name"], complexity=complexity, mode=mode
+            collection=random.choice(
+                [
+                    s["collection_name"]
+                    for s in SanatanConfig.scriptures
+                    if s["collection_name"] != "yt_metadata"
+                ]
+            ),
+            complexity=random.choice(["beginner", "intermediate", "advanced"]),
+            mode=random.choice(["mcq", "open"]),
+            preferred_lamguage="Tamil",
         )
         print(q.model_dump_json(indent=1))
         print("_______________________")

tests/test_quiz.py

@@ -0,0 +1,36 @@
+import random
+from config import SanatanConfig
+from modules.quiz.answer_validator import validate_answer
+from modules.quiz.quiz_helper import generate_question
+
+
+if __name__ == "__main__":
+    while True:
+        q = generate_question(
+            collection=random.choice(
+                [
+                    s["collection_name"]
+                    for s in SanatanConfig.scriptures
+                    if s["collection_name"] != "yt_metadata"
+                ]
+            ),
+            complexity=random.choice(["beginner", "intermediate", "advanced"]),
+            mode=random.choice(["mcq", "open"]),
+            preferred_lamguage="English",
+        )
+        print(q.model_dump_json(indent=1))
+
+        print("Q:", q.question)
+        print("Choices:", q.choices)
+        print("Expected:", q.expected_answer)
+
+        # Simulate user input
+        user_ans = input("Your Answer: ")
+
+        result = validate_answer(q, user_ans, preferred_language="English")
+        print(result.model_dump_json(indent=2))
+
+        want_more = input("Want to keep playing? Y/N: ")
+        if want_more == "N":
+            break
+    print("Game over!")