UPDATE

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+vectorStore/chroma.sqlite3 filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -0,0 +1,17 @@
+FROM python:3.10-slim
+
+COPY . /app
+
+WORKDIR /app
+
+USER root
+
+RUN pip install uv
+
+RUN uv sync
+
+RUN chmod -R 777 /app
+
+EXPOSE 7860
+
+CMD ["uv", "run", "main.py"]

config.py

@@ -0,0 +1,19 @@
+# config.py
+# This file stores all the configuration variables and constants for the application.
+
+# Embedding Model Configuration
+MODEL_NAME = "jinaai/jina-embeddings-v3-base-en"
+MODEL_KWARGS = {'device': 'cpu', 'trust_remote_code': True}
+ENCODE_KWARGS = {'normalize_embeddings': False}
+
+# Vector Store Configuration
+VECTOR_STORE_DIRECTORY = "vectorStore"
+
+# LLM Configuration
+LLM_MODEL = "meta-llama/llama-4-maverick-17b-128e-instruct"
+LLM_TEMPERATURE = 0.8
+
+# Retriever Configuration
+DENSE_RETRIEVER_K = 3
+KEYWORD_RETRIEVER_K = 3
+ENSEMBLE_WEIGHTS = [0.5, 0.5]

llm_chain.py

@@ -0,0 +1,69 @@
+# llm_chain.py
+# This file configures the language model, prompt template, and the final processing chain.
+
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_groq import ChatGroq
+from config import LLM_MODEL, LLM_TEMPERATURE
+
+def get_llm():
+    """Initializes and returns the ChatGroq LLM."""
+    return ChatGroq(
+        model=LLM_MODEL,
+        temperature=LLM_TEMPERATURE
+    )
+
+def get_prompt_template():
+    """Creates and returns the ChatPromptTemplate for the RAG chain."""
+    prompt_text = """
+    You are an AI assistant specialized exclusively in answering questions from **three mathematics books authored by Ice Venkatesh**.
+    Your role is to provide **clear, precise, and complete answers strictly based on the retrieved context from those books**.
+
+    ---
+
+    ### 🔹 Core Rules
+
+    1. **Strict Reliance on Context**: Only use the retrieved context (`context`) from Ice Venkatesh’s books. Never invent, guess, or rely on outside knowledge. If the context is insufficient, state this clearly.
+
+    2. **No External Sources**: Do not use prior training knowledge or the internet. All answers, definitions, proofs, and examples must come **only from the provided context**.
+
+    3. **Mathematical Rigor & Clarity**: Provide correct, step-by-step reasoning when solving problems. Keep explanations simple, precise, and mathematically flawless.
+
+    4. **Handling Off-Topic Queries**: If the user asks something unrelated to the books or mathematics, politely decline. Example: "I can only help with questions related to the three mathematics books by Ice Venkatesh. Unfortunately, I cannot assist with topics outside that scope."
+
+    5. **Style & Professionalism**: Always be polite, clear, and professional. Avoid filler or speculation.
+
+    ---
+
+    ### 🔹 Response Instructions
+
+    * Use the `context` to extract the necessary information.
+    * Answer the `query` directly, grounded **only in the retrieved book content**.
+    * If insufficient context is found, respond with: "The books by Ice Venkatesh do not provide enough information to answer this question."
+
+    ---
+
+    **Retrieved context:**
+    {context}
+
+    **User question:**
+    {query}
+    """
+    return ChatPromptTemplate.from_template(prompt_text)
+
+def create_rag_chain(retriever):
+    """Creates and returns the full RAG chain, accepting a retriever as an argument."""
+    print("Creating RAG chain...")
+    llm = get_llm()
+    prompt = get_prompt_template()
+    output_parser = StrOutputParser()
+
+    chain = (
+        {"context": retriever, "query": RunnablePassthrough()}
+        | prompt
+        | llm
+        | output_parser
+    )
+    print("RAG chain created successfully.")
+    return chain

main.py

@@ -0,0 +1,47 @@
+# main.py
+# This is the main file that runs the Sanic web server.
+
+from sanic import Sanic, response
+from retriever import get_ensemble_retriever
+from llm_chain import create_rag_chain
+
+app = Sanic("VibbaBackend")
+
+@app.before_server_start
+async def setup_model(app_instance, loop):
+    """
+    Initializes the retriever and RAG chain and attaches them
+    to the application context before the server starts.
+    """
+    print("Server starting up... Initializing model pipeline.")
+    retriever = get_ensemble_retriever()
+    rag_chain = create_rag_chain(retriever)
+    app_instance.ctx.rag_chain = rag_chain
+    print("Model pipeline is ready.")
+
+@app.get("/getResponse")
+async def get_response_endpoint(request):
+    """
+    Endpoint to get an answer to a question using the RAG chain.
+    Expects a 'question' query parameter.
+    """
+    question = request.args.get("question")
+    if not question:
+        return response.json(
+            {"error": "Please provide a 'question' query parameter."},
+            status=400
+        )
+
+    try:
+        chain = request.app.ctx.rag_chain
+        result = chain.invoke(question)
+        return response.text(result)
+    except Exception as e:
+        print(f"An error occurred during invocation: {e}")
+        return response.json(
+            {"error": "An internal error occurred while processing your request."},
+            status=500
+        )
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=7860)

pyproject.toml

@@ -0,0 +1,18 @@
+[project]
+name = "bookai"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = [
+    "einops>=0.8.1",
+    "langchain>=0.3.27",
+    "langchain-chroma>=0.2.6",
+    "langchain-community>=0.3.29",
+    "langchain-core>=0.3.76",
+    "langchain-groq>=0.3.8",
+    "langchain-huggingface>=0.3.1",
+    "rank-bm25>=0.2.2",
+    "sanic>=25.3.0",
+    "sentence-transformers>=5.1.1",
+]

retriever.py

@@ -0,0 +1,61 @@
+# retriever.py
+# This file handles the setup of embeddings, vector stores, and the ensemble retriever.
+
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_chroma import Chroma
+from langchain_community.retrievers import BM25Retriever
+from langchain.retrievers import EnsembleRetriever
+from config import (
+    MODEL_NAME, MODEL_KWARGS, ENCODE_KWARGS, VECTOR_STORE_DIRECTORY,
+    DENSE_RETRIEVER_K, KEYWORD_RETRIEVER_K, ENSEMBLE_WEIGHTS
+)
+
+def get_embedding_function():
+    """Initializes and returns the HuggingFace embedding model."""
+    return HuggingFaceEmbeddings(
+        model_name=MODEL_NAME,
+        model_kwargs=MODEL_KWARGS,
+        encode_kwargs=ENCODE_KWARGS
+    )
+
+def get_vector_store(embedding_function):
+    """Initializes and returns the Chroma vector store."""
+    return Chroma(
+        embedding_function=embedding_function,
+        persist_directory=VECTOR_STORE_DIRECTORY
+    )
+
+def get_ensemble_retriever():
+    """
+    Creates and returns an ensemble retriever combining dense and keyword-based search.
+    """
+    print("Initializing embeddings and vector store...")
+    embeddings = get_embedding_function()
+    vector_store = get_vector_store(embeddings)
+
+    dense_vector_retriever = vector_store.as_retriever(k=DENSE_RETRIEVER_K)
+
+    print("Loading documents for BM25 retriever...")
+    ids = vector_store.get().get("ids", [])
+
+    if not ids:
+        all_documents = []
+    else:
+        all_documents = vector_store.get_by_ids(ids)
+
+    keyword_search_retriever = BM25Retriever.from_documents(
+        documents=all_documents, k=KEYWORD_RETRIEVER_K
+    ) if all_documents else None
+    
+    if keyword_search_retriever:
+        print("Creating ensemble retriever...")
+        ensemble_retriever = EnsembleRetriever(
+            retrievers=[dense_vector_retriever, keyword_search_retriever],
+            weights=ENSEMBLE_WEIGHTS
+        )
+    else:
+        print("Creating dense-only retriever...")
+        ensemble_retriever = dense_vector_retriever
+
+    print("Retriever setup complete.")
+    return ensemble_retriever

vectorStore/chroma.sqlite3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7081f2a48908387f519e62c823ef7a9f3487b0e17a4f3a68d7ab81cd949a4a29
+size 5992448