Upload 15 files

.dockerignore

@@ -0,0 +1,13 @@
+venv/
+__pycache__/
+.env
+.ipynb_checkpoints/
+*.pyc
+
+# Exclude any local model weights if you accidentally put them in the folder
+*.bin
+*.bin.*
+*.pth
+*.pt
+model/
+checkpoint/

.gitattributes

@@ -33,3 +33,5 @@ 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
+chroma_db/chroma.sqlite3 filter=lfs diff=lfs merge=lfs -text
+data/MedicalBook.pdf filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,4 @@
+venv/
+.env
+chroma_db/
+__pycache__/

app.py

@@ -0,0 +1,41 @@
+import streamlit as st
+import requests  # Standard library for sending API requests
+
+st.set_page_config(page_title="Medical AI Assistant", layout="centered")
+st.title("⚕️ General Medicine Chatbot")
+
+# This is the address of your FastAPI "Brain"
+API_URL = "http://127.0.0.1:8000/ask"
+
+# Initialize chat history
+if "messages" not in st.session_state:
+    st.session_state.messages = []
+
+# Display previous messages
+for message in st.session_state.messages:
+    with st.chat_message(message["role"]):
+        st.markdown(message["content"])
+
+if prompt := st.chat_input("Ask a medical question..."):
+    # 1. Display user message
+    st.chat_message("user").markdown(prompt)
+    st.session_state.messages.append({"role": "user", "content": prompt})
+
+    # 2. Call the API
+    with st.chat_message("assistant"):
+        with st.spinner("Thinking..."):
+            try:
+                # We send the query to your FastAPI server
+                # Note: 'query' must match the Pydantic model in main.py
+                payload = {"query": prompt}
+                response = requests.post(API_URL, json=payload)
+                
+                if response.status_code == 200:
+                    answer = response.json().get("answer", "No answer found.")
+                    st.markdown(answer)
+                    st.session_state.messages.append({"role": "assistant", "content": answer})
+                else:
+                    st.error(f"API Error {response.status_code}: {response.text}")
+            
+            except Exception as e:
+                st.error(f"Could not connect to the API. Is Uvicorn running? Error: {e}")

chroma_db/3680a8e6-9770-4cfc-9ae4-18ad2a369d9b/data_level0.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7d011826d12a6ecb524c46ede71a830b5c48fd6f1cce9fccc4769ab833c3f7a
+size 30484764

chroma_db/3680a8e6-9770-4cfc-9ae4-18ad2a369d9b/header.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cc1e2b6bdf1eee6eeddf6cb9aa7e20c69033a8a929d902cd56859392e703544
+size 100

chroma_db/3680a8e6-9770-4cfc-9ae4-18ad2a369d9b/index_metadata.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cddbf2ed05e967b82f0aa08af3fc6d86d674784c779fc12911e6b4c0513a534
+size 1673588

chroma_db/3680a8e6-9770-4cfc-9ae4-18ad2a369d9b/length.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e94d1a8852d94aa7c25b56188ab929e437a50bcce95ffd81eacfc29596156ec
+size 72756

chroma_db/3680a8e6-9770-4cfc-9ae4-18ad2a369d9b/link_lists.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a0204e0094ab2daaecd241d7808b915e3c05cf6f38abed8cba205a33303304e
+size 156260

chroma_db/chroma.sqlite3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:004ed53cdb169fca6e2874342756070a0d218cf9d3856ddd4395176bb58f83c9
+size 118001664

data/MedicalBook.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cef47d2bcde526111d3d29f77876ab139efdf7d7e5e3180838fff9c335c64dc1
+size 10764110

engine.py

@@ -0,0 +1,135 @@
+from operator import itemgetter
+import os
+from dotenv import load_dotenv
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_chroma import Chroma
+
+from utils import load_docs, split_docs
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.runnables import RunnablePassthrough
+from langchain_huggingface import ChatHuggingFace
+
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+from langchain_huggingface import HuggingFacePipeline
+
+
+import warnings
+warnings.filterwarnings("ignore", category=UserWarning)
+
+load_dotenv()
+HF_TOKEN = os.getenv("HF_TOKEN")
+file_path = "data/MedicalBook.pdf"
+
+def downlode_hugging_face_embeddings():
+    embeddings=HuggingFaceEmbeddings(model_name='sentence-transformers/msmarco-MiniLM-L6-v3')
+    return embeddings
+
+CHROMA_PATH="chroma_db"
+#embeddings = downlode_hugging_face_embeddings()
+
+def create_vectorstore(embeddings):
+    if os.path.exists(CHROMA_PATH):
+        print(f"--- Loading existing Vector DB from {CHROMA_PATH}... ---")
+        # FIX: To LOAD, use Chroma() directly. Do NOT use .from_documents()
+        return Chroma(
+            persist_directory=CHROMA_PATH, 
+            embedding_function=embeddings
+        )  
+    else: 
+        print("Creating a new one...")
+        documents = load_docs(file_path)
+        text_chunks = split_docs(documents)
+        # Use .from_documents ONLY when you have new text_chunks to process
+        return Chroma.from_documents(
+            documents=text_chunks, 
+            embedding=embeddings, 
+            persist_directory=CHROMA_PATH 
+        )
+
+# vector_db=create_vectorstore()
+# retriever=vector_db.as_retriever(search_kwargs={"k": 3})
+
+def get_llm():
+    # The correct ID for the 1B Instruct model
+    model_path = "ibm-granite/granite-3.0-1b-a400m-instruct"
+    
+    device = -1 # Force CPU
+    print(f"--- Loading local model: {model_path} ---")
+
+    tokenizer = AutoTokenizer.from_pretrained(model_path)
+    
+    # We use low_cpu_mem_usage to keep the RAM footprint small
+    model = AutoModelForCausalLM.from_pretrained(
+        model_path,
+        torch_dtype=torch.float32, 
+        low_cpu_mem_usage=True,
+        device_map=None 
+    )
+    
+    gen_pipeline = pipeline(
+        "text-generation",
+        model=model,
+        tokenizer=tokenizer,
+        max_new_tokens=256, 
+        temperature=0.3,
+        do_sample=True,
+        return_full_text=False
+    )
+    
+    llm = HuggingFacePipeline(pipeline=gen_pipeline)
+    return ChatHuggingFace(llm=llm)
+
+
+def get_llm_response(user_input, retriever, llm):
+    prompt = ChatPromptTemplate.from_messages([
+        ("system", "You are a medical specialist. Use the context to answer. If you don't know, say 'I don't know'."),
+        ("human", "Context: {context}\n\nQuestion: {input}")
+    ])
+    
+    chain = (
+        {
+            "context": itemgetter("input") | retriever,
+            "input": itemgetter("input"),
+        }
+        | prompt
+        | llm
+        | StrOutputParser()
+    )
+    return chain.invoke({"input": user_input})
+
+def load_system():
+    """
+    Orchestrates the loading of the full RAG pipeline.
+    This can be called by both FastAPI and Streamlit.
+    """
+    print("--- Initializing Medical AI System ---")
+    embeddings = downlode_hugging_face_embeddings()
+    vector_db = create_vectorstore(embeddings)
+    retriever = vector_db.as_retriever(search_kwargs={"k": 3})
+    llm = get_llm()
+    print("--- System Ready ---")
+    return retriever, llm       
+
+if __name__=="__main__":
+
+    embeddings = downlode_hugging_face_embeddings()
+    vector_db=create_vectorstore(embeddings)
+    retriever=vector_db.as_retriever(search_kwargs={"k": 3})
+    llm=get_llm()
+
+#     print("\n -- LLM Testing -- \n")
+#     while True:
+#         User_query=input("User: ")
+#         if User_query.lower() in ["quit","exit","q"]:
+#             print("Exiting..")
+#             break
+#         response=get_llm_response(User_query, retriever, llm)
+#         print(f"Assistant: {response}")
+
+
+
+    

main.py

@@ -0,0 +1,34 @@
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+from engine import load_system, get_llm_response
+
+
+app=FastAPI(
+    title="Medical RAG API",
+    description="Backend API for IBM Granite chat model",
+    version="1.0.0")
+
+#gloabal model loading 
+print("Loading Medical AI Engine... Please wait.")
+retriever, llm = load_system()
+print("AI loaded successfully.")
+
+class ChatRequest(BaseModel):
+    query: str
+
+class ChatResponse(BaseModel):
+    answer: str
+
+@app.get("/")
+def read_root():
+    return {"status":"online","message": "Medical API is Ready"}
+
+@app.post("/ask", response_model=ChatResponse)
+async def ask_question(request: ChatRequest):
+    try:
+        # Using the globally loaded retriever and llm
+        response = get_llm_response(request.query, retriever, llm)
+        return ChatResponse(answer=response)
+    except Exception as e:
+        # Standard API error handling
+        raise HTTPException(status_code=500, detail=str(e))

requirements.txt

@@ -0,0 +1,18 @@
+fastapi
+uvicorn
+Pydentic
+python-dotenv
+
+langchain
+langchain-community
+langchain-huggingface
+sentence-transformers
+
+pypdf
+langchain-chroma
+streamlit
+
+requests
+torch --index-url https://download.pytorch.org/whl/cpu
+transformers
+accelerate