Merge pull request #1 from run-llama/clelia/adding-document-chat

Adding document chat and moving to a multi-page app

Author: AstraBert

.env.example

@@ -0,0 +1,3 @@
+OPENAI_API_KEY="sk-***"
+LLAMACLOUD_API_KEY="llx-***"
+ELEVENLABS_API_KEY="sk_***"

LICENSE

@@ -1,6 +1,6 @@
 The MIT License
 
-Copyright (c) Clelia Astra Bertelli
+Copyright (c) Jerry Liu
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -1,3 +1,67 @@
-# NotebookLM clone
+# NotebookLlaMa🦙
 
-This project is aimed at producing a fully open-source, LlamaCloud-backed alternative to NotebookLM.
+## A fluffy and open-source alternative to NotebookLM!
+
+This project is aimed at producing a fully open-source, [**LlamaCloud**](https//cloud.llamaindex.ai)-backed alternative to NotebookLM.
+
+### Get it up and running!
+
+Get the GitHub repository:
+
+```bash
+git clone https://github.com/run-llama/notebooklm-clone
+```
+
+Install dependencies:
+
+```bash
+cd notebooklm-clone/
+uv sync
+```
+
+Modify the `.env.example` file with your API keys:
+
+- `OPENAI_API_KEY`: find it [on OpenAI Platform](https://platform.openai.com/api-keys)
+- `ELEVENLABS_API_KEY`: find it [on ElevenLabs Settings](https://elevenlabs.io/app/settings/api-keys)
+- `LLAMACLOUD_API_KEY`: find it [on LlamaCloud Dashboard](https://cloud.llamaindex.ai/)
+
+Rename the file to `.env`:
+
+```bash
+mv .env.example .env
+```
+
+Now, you will have to execute the following scripts:
+
+```bash
+uv run tools/create_llama_extract_agent.py
+uv run tools/create_llama_cloud_index.py
+```
+
+And you're ready to set up the app!
+
+Run the **MCP** server:
+
+```bash
+uv run src/notebooklm_clone/server.py
+```
+
+Now, launch the Streamlit app:
+
+```bash
+streamlit run src/notebooklm_clone/Home.py
+```
+
+> [!IMPORTANT]
+>
+> _You might need to install `ffmpeg` if you do not have it installed already_
+
+And start exploring the app at `http://localhost:8751/`.
+
+### Contributing
+
+Contribute to this project following the [guidelines](./CONTRIBUTING.md).
+
+### License
+
+This project is provided under an [MIT License](LICENSE).

src/notebooklm_clone/main.py renamed to src/notebooklm_clone/Home.py

@@ -53,16 +53,19 @@ def sync_create_podcast(file_content: str):
 
 # Display the network
 st.set_page_config(
-    page_title="NotebookLlaMa",
-    page_icon="🦙",
+    page_title="NotebookLlaMa - Home",
+    page_icon="🏠",
     layout="wide",
     menu_items={
         "Get Help": "https://github.com/run-llama/notebooklm-clone/discussions/categories/general",
         "Report a bug": "https://github.com/run-llama/notebooklm-clone/issues/",
         "About": "An OSS alternative to NotebookLM that runs with the power of a flully Llama!",
     },
 )
-st.title("NotebookLlaMa🦙")
+st.sidebar.header("Home🏠")
+st.sidebar.info("To switch to the Document Chat, select it from above!🔺")
+st.markdown("---")
+st.markdown("## NotebookLlaMa - Home🦙")
 
 file_input = st.file_uploader(
     label="Upload your source PDF file!", accept_multiple_files=False

src/notebooklm_clone/audio.py

@@ -57,10 +57,11 @@ def validate_conversation(self) -> Self:
 
 
 class PodcastGenerator(BaseModel):
-    model_config: ConfigDict = ConfigDict(arbitrary_types_allowed=True)
     llm: StructuredLLM
     client: AsyncElevenLabs
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
+
     @model_validator(mode="after")
     def validate_podcast(self) -> Self:
         try:
@@ -69,6 +70,7 @@ def validate_podcast(self) -> Self:
             raise ValueError(
                 f"The output class of the structured LLM must be {MultiTurnConversation.__qualname__}, your LLM has output class: {self.llm.output_cls.__qualname__}"
             )
+        return self
 
     async def _conversation_script(self, file_transcript: str) -> MultiTurnConversation:
         response = await self.llm.achat(

src/notebooklm_clone/pages/1_Document_Chat.py

@@ -0,0 +1,93 @@
+import streamlit as st
+import asyncio
+
+from llama_index.tools.mcp import BasicMCPClient
+
+MCP_CLIENT = BasicMCPClient(command_or_url="http://localhost:8000/mcp")
+
+
+async def chat(inpt: str):
+    result = await MCP_CLIENT.call_tool(
+        tool_name="query_index_tool", arguments={"question": inpt}
+    )
+    return result.content[0].text
+
+
+def sync_chat(inpt: str):
+    return asyncio.run(chat(inpt))
+
+
+# Chat Interface
+st.set_page_config(page_title="NotebookLlaMa - Document Chat", page_icon="🗣")
+
+st.sidebar.header("Document Chat🗣")
+st.sidebar.info("To switch to the Home page, select it from above!🔺")
+st.markdown("---")
+st.markdown("## NotebookLlaMa - Document Chat🗣")
+
+# Initialize chat history
+if "messages" not in st.session_state:
+    st.session_state.messages = []
+
+# Display chat messages from history on app rerun
+for i, message in enumerate(st.session_state.messages):
+    with st.chat_message(message["role"]):
+        if message["role"] == "assistant" and "sources" in message:
+            # Display the main response
+            st.markdown(message["content"])
+            # Add toggle for sources
+            with st.expander("Sources"):
+                st.markdown(message["sources"])
+        else:
+            st.markdown(message["content"])
+
+# React to user input
+if prompt := st.chat_input("Ask a question about your document"):
+    # Display user message in chat message container
+    st.chat_message("user").markdown(prompt)
+    # Add user message to chat history
+    st.session_state.messages.append({"role": "user", "content": prompt})
+
+    # Get bot response
+    with st.chat_message("assistant"):
+        with st.spinner("Thinking..."):
+            try:
+                response = sync_chat(prompt)
+
+                # Split response and sources if they exist
+                # Assuming your response format includes sources somehow
+                # You might need to modify this based on your actual response format
+                if "## Sources" in response:
+                    parts = response.split("## Sources", 1)
+                    main_response = parts[0].strip()
+                    sources = "## Sources" + parts[1].strip()
+                else:
+                    main_response = response
+                    sources = None
+
+                st.markdown(main_response)
+
+                # Add toggle for sources if they exist
+                if sources:
+                    with st.expander("Sources"):
+                        st.markdown(sources)
+                    # Add to history with sources
+                    st.session_state.messages.append(
+                        {
+                            "role": "assistant",
+                            "content": main_response,
+                            "sources": sources,
+                        }
+                    )
+                else:
+                    # Add to history without sources
+                    st.session_state.messages.append(
+                        {"role": "assistant", "content": main_response}
+                    )
+
+            except Exception as e:
+                error_msg = f"Error: {str(e)}"
+                st.markdown(error_msg)
+                st.session_state.messages.append(
+                    {"role": "assistant", "content": error_msg}
+                )

src/notebooklm_clone/utils.py

@@ -6,8 +6,6 @@
 
 from pydantic import BaseModel, Field, model_validator
 from llama_index.core.llms import ChatMessage
-from llama_index.core.query_engine.multistep_query_engine import MultiStepQueryEngine
-from llama_index.core.indices.query.query_transform import StepDecomposeQueryTransform
 from llama_cloud_services import LlamaExtract, LlamaParse
 from llama_cloud_services.extract import SourceText
 from llama_cloud.client import AsyncLlamaCloud
@@ -90,11 +88,9 @@ class MindMapCreationFailedWarning(Warning):
     )
     PARSER = LlamaParse(api_key=os.getenv("LLAMACLOUD_API_KEY"), result_type="markdown")
     PIPELINE_ID = os.getenv("LLAMACLOUD_PIPELINE_ID")
-    qe = LlamaCloudIndex(
+    QE = LlamaCloudIndex(
         api_key=os.getenv("LLAMACLOUD_API_KEY"), pipeline_id=PIPELINE_ID
     ).as_query_engine(llm=LLM)
-    step_decompose = StepDecomposeQueryTransform(llm=LLM)
-    MS_QE = MultiStepQueryEngine(query_engine=qe, query_transform=step_decompose)
     LLM_STRUCT = LLM.as_structured_llm(MindMap)
 
 
@@ -157,13 +153,13 @@ async def get_mind_map(summary: str, highlights: List[str]) -> Union[str, None]:
 
 
 async def query_index(question: str) -> Union[str, None]:
-    response = await MS_QE.aquery(question)
+    response = await QE.aquery(question)
     if not response.response:
         return None
     sources = [node.text for node in response.source_nodes]
     return (
         "## Answer\n\n"
         + response.response
-        + "\n\n##Sources\n\n- "
+        + "\n\n## Sources\n\n- "
         + "\n- ".join(sources)
     )