Assessment Generation

def search_and_ask(user_query, collection_name: str, 
                   blooms_requirements: str = "5 remember, 3 understand, 4 apply, 3 analyze, 2 evaluate, 3 create", 
                   top_k = 5):
    
    vector_db = _vector_db(collection_name=collection_name)
    search_results = vector_db.similarity_search(query=user_query, k=top_k)
    
    if not search_results:
        print("No search result from vector DB.")
        return

context_blocks = []
for result in search_results:
    block = (
        f"--- ADMIN METADATA (DO NOT MENTION IN OUTPUT) ---\n"
        f"Source: {result.metadata['source']}\n"
        f"Page: {result.metadata['page_label']}\n"
        f"--- EDUCATIONAL CONTENT ---\n"
        f"{result.page_content}\n"
    )
    context_blocks.append(block)
    
context = "\n\n".join(context_blocks)

def prompt_modelling(context, blooms_requirements: str):
    SYSTEM_PROMPT = f"""
        You are a Subject Matter Expert designing a professional, standalone exam. 
        You have been provided with "Educational Content" and "Admin Metadata" for verification.

        ### THE RULES FOR YOUR OUTPUT:
        1. **STRICT BLIND EXAM MODE**: Write the questions as if the student has NO access to any documents. 
           - DO NOT mention "Page Numbers," "Lessons," "Sections," or "the PDF."
           - BAD: "According to the provided text on page 4, what is..."
           - GOOD: "What is the primary characteristic of..."
        2. **INTERNAL VERIFICATION ONLY**: Use the "Admin Metadata" only to ensure your answer is grounded in the correct chapter. DO NOT repeat this metadata in the question, the options, or the explanation.
        3. **EXPLANATION FORMAT**: Write the explanation as a factual teaching note. 
           - BAD: "This is found on page 10 of nodejs.pdf."
           - GOOD: "Promises are used to handle asynchronous operations more cleanly than callbacks."
        4. **BLOOM'S TAXONOMY**: Generate questions according to these counts: {blooms_requirements}.
           For each question, set the `bloom_level` field to exactly one of: remember, understand, apply, analyze, evaluate, create — matching the cognitive level of that question.

        ### PROVIDED DATA (FOR YOUR EYES ONLY):
        {context}
    """
    return SYSTEM_PROMPT

class SingleMCQ(BaseModel):
    question_no : str
    bloom_level : str  # e.g. "remember", "understand", "apply", "analyze", "evaluate", "create"
    question : str
    answer_options : List[str]
    correct_answer : str
    explaination : Optional[str]
    
class OutputFormat(BaseModel):
    mcqs : List[SingleMCQ]

response = open_ai_client.chat.completions.parse(
    model='gemini-2.5-flash-lite',
    response_format= OutputFormat,
    messages=[
        {"role":"system", "content" : SYSTEM_PROMPT},
        {"role":"user", "content":user_query},
    ],
)

parsed = response.choices[0].message.parsed
return parsed.model_dump() if hasattr(parsed, "model_dump") else parsed

blooms_requirements = "5 remember, 3 understand, 4 apply, 3 analyze, 2 evaluate, 3 create"

def search_and_ask(user_query, collection_name: str, 
                   blooms_requirements: str = "5 remember, 3 understand, 4 apply, 3 analyze, 2 evaluate, 3 create", 
                   top_k = 5):
    # 1. Retrieve context
    vector_db = _vector_db(collection_name=collection_name)
    search_results = vector_db.similarity_search(query=user_query, k=top_k)
    
    if not search_results:
        print("No search result from vector DB.")
        return
    
    # 2. Format context
    context_blocks = []
    for result in search_results:
        block = (
            f"--- ADMIN METADATA (DO NOT MENTION IN OUTPUT) ---\n"
            f"Source: {result.metadata['source']}\n"
            f"Page: {result.metadata['page_label']}\n"
            f"--- EDUCATIONAL CONTENT ---\n"
            f"{result.page_content}\n"
        )
        context_blocks.append(block)
    context = "\n\n".join(context_blocks)
    
    # 3. Build prompt
    SYSTEM_PROMPT = prompt_modelling(context, blooms_requirements)
    
    # 4. Generate structured output
    response = open_ai_client.chat.completions.parse(
        model='gemini-2.5-flash-lite',
        response_format= OutputFormat,
        messages=[
            {"role":"system", "content" : SYSTEM_PROMPT},
            {"role":"user", "content":user_query},
        ],
    )
    
    # 5. Return validated result
    parsed = response.choices[0].message.parsed
    return parsed.model_dump() if hasattr(parsed, "model_dump") else parsed

class Assessment(Base):
    __tablename__ = "assessments"
    
    id = Column(Integer, primary_key=True, index=True)
    user_id = Column(Integer, ForeignKey("users.id"))
    chapter_name = Column(String)
    bloom_factors = Column(JSONB)  # Stores {remember: 5, apply: 2, etc.}
    content_json = Column(JSONB)   # Stores the complete output from Gemini
    created_at = Column(DateTime(timezone=True), server_default=func.now())