Upload 7 files

app.py

@@ -0,0 +1,41 @@
+import os
+import gradio as gr
+from scripts.router_simple import build_router_chain
+
+OPENAI_KEY = os.getenv("OPENAI_API_KEY", None)
+MODEL_NAME = os.getenv("OPENAI_MODEL", "gpt-4o-mini")
+
+if not OPENAI_KEY:
+    print("WARNING: OPENAI_API_KEY not set. The app may fail at runtime.")
+
+# Build the router once (keeps vectorstore & models in memory)
+router = build_router_chain(model_name=MODEL_NAME)
+
+def chat_fn(message, history):
+    if not message:
+        return history, ""
+    # call router
+    result = router.invoke({"input": message})
+    # RetrievalQA returns dict with 'result' key (and maybe 'source_documents')
+    answer = result.get("result") if isinstance(result, dict) else str(result)
+    # append sources if present
+    sources = None
+    if isinstance(result, dict) and "source_documents" in result and result["source_documents"]:
+        try:
+            sources = list({str(d.metadata.get("source", "unknown")) for d in result["source_documents"]})
+        except Exception:
+            sources = None
+    if sources:
+        answer = f"{answer}\n\n📚 Sources: {', '.join(sources)}"
+    history.append((message, answer))
+    return history, ""
+
+with gr.Blocks() as demo:
+    gr.Markdown("## 📚 Course Assistant — Chat with your course files")
+    chatbot = gr.Chatbot(elem_id="chatbot")
+    txt = gr.Textbox(show_label=False, placeholder="Ask about the course...")
+    txt.submit(chat_fn, [txt, chatbot], [chatbot, txt])
+    txt.submit(lambda: None, None, txt)  # clear input
+
+if __name__ == "__main__":
+    demo.launch(server_port=int(os.getenv("PORT", 7860)))

output/all_docs.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5f5adf7c679373919bad157bdf906af58e79d96197f5dc8d4a544b808ba9943
+size 6245290

output/chunks.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0401320f1374d0169a82a3f9e66fc2814bf3c3180074d90ccaffdd530c00240a
+size 6796736

requirements.txt

@@ -0,0 +1,10 @@
+langchain
+langchain-community
+langchain-openai
+langchain-chroma
+chromadb
+tiktoken
+gradio
+pickle5
+pydantic<2
+

scripts/rag_chat.py

@@ -0,0 +1,36 @@
+from langchain.chains import RetrievalQA
+from langchain_openai import ChatOpenAI
+from langchain_chroma import Chroma
+from langchain_openai import OpenAIEmbeddings
+from langchain.prompts import PromptTemplate
+from pathlib import Path
+
+BASE_DIR = Path(__file__).resolve().parent.parent
+DB_DIR = str(BASE_DIR / "db")
+
+def build_general_qa_chain(model_name=None):
+    embedding = OpenAIEmbeddings(model="text-embedding-3-small")
+    vectorstore = Chroma(persist_directory=DB_DIR, embedding_function=embedding)
+    
+    # Custom prompt with source attribution
+    template = """Use the following context to answer the question. 
+    If the answer isn't found in the context, use your general knowledge but say so.
+    Always cite your sources at the end with 'Source: <filename>' when using course materials.
+
+    Context: {context}
+    Question: {question}
+    Helpful Answer:"""
+    
+    QA_PROMPT = PromptTemplate(
+        template=template,
+        input_variables=["context", "question"]
+    )
+    
+    llm = ChatOpenAI(model_name=model_name or "gpt-4o-mini", temperature=0.0) 
+    qa_chain = RetrievalQA.from_chain_type(
+        llm=llm,
+        retriever=vectorstore.as_retriever(search_kwargs={"k": 4}),
+        chain_type_kwargs={"prompt": QA_PROMPT},
+        return_source_documents=True
+    )
+    return qa_chain

scripts/router_chain.py

@@ -0,0 +1,33 @@
+# scripts/router_simple.py
+from typing import Dict, Any
+from langchain.chat_models import ChatOpenAI
+from langchain.prompts import ChatPromptTemplate
+from scripts.rag_chat import build_general_qa_chain
+
+def build_router_chain(model_name=None):
+    general_qa = build_general_qa_chain(model_name=model_name)
+    llm = ChatOpenAI(model_name=model_name or "gpt-4o-mini", temperature=0.0)
+
+    class Router:
+        def invoke(self, input_dict: Dict[str, Any]):
+            text = input_dict.get("input", "").lower()
+            if "code" in text or "program" in text or "debug" in text:
+                prompt = ChatPromptTemplate.from_template(
+                    "As a coding assistant, help with this Python question.\nQuestion: {input}\nAnswer:"
+                )
+                chain = prompt | llm
+                return {"result": chain.invoke({"input": input_dict["input"]}).content}
+            elif "summarize" in text or "summary" in text:
+                prompt = ChatPromptTemplate.from_template(
+                    "Provide a concise summary about: {input}\nSummary:"
+                )
+                chain = prompt | llm
+                return {"result": chain.invoke({"input": input_dict["input"]}).content}
+            elif "calculate" in text or any(char.isdigit() for char in text):
+                return {"result": "For calculations, please ask a specific calculation or provide more context."}
+            else:
+                # Use RAG chain
+                result = general_qa({"query": input_dict["input"]})
+                return result
+
+    return Router()

scripts/summarizer.py

@@ -0,0 +1,7 @@
+from langchain.chains.summarize import load_summarize_chain
+from langchain_openai import ChatOpenAI
+
+def get_summarizer():
+    llm = ChatOpenAI(model_name="gpt-4o-mini", temperature=0)
+    chain = load_summarize_chain(llm, chain_type="map_reduce")
+    return chain