app.py

from typing import Any

import gradio as gr
from langchain.chat_models import init_chat_model
from langchain_core.tools import tool
from langchain_aws import BedrockEmbeddings
from langchain_qdrant import QdrantVectorStore
from langgraph.checkpoint.memory import MemorySaver
from langgraph.graph import MessagesState, StateGraph, END
from langgraph.prebuilt import ToolNode, tools_condition

import structlog

from qdrant_client import QdrantClient
from qdrant_client.http.models import Distance, VectorParams, SparseVectorParams

import logging_config as _
# from conversation.main import graph
from conversation.generate import generate, trigger_ai_message_with_tool_call
from conversation.chat_source_history import prettify_source_history, build_source_history_object, clear_chat_and_source_history, create_chat_hash
from ingestion.main import ingest_document
from tools.langfuse_client import get_langfuse_handler

from config import app_settings

# Get Langfuse Callback handler
langfuse_handler = get_langfuse_handler()

# Create a logger instance
logger = structlog.get_logger(__name__)

embeddings = BedrockEmbeddings(
    model_id=app_settings.embedding_model,
    region_name=app_settings.llm_region,
    aws_access_key_id=app_settings.aws_access_key_id,
    aws_secret_access_key=app_settings.aws_secret_access_key,
)

llm = init_chat_model(
    app_settings.llm_model,
    model_provider=app_settings.model_provider,
    region_name=app_settings.llm_region,
    aws_access_key_id=app_settings.aws_access_key_id,
    aws_secret_access_key=app_settings.aws_secret_access_key,
)

client = QdrantClient(app_settings.vector_db_url)
if not client.collection_exists(app_settings.vector_db_collection_name):
    client.create_collection(
        collection_name=app_settings.vector_db_collection_name,
        vectors_config=VectorParams(size=app_settings.embedding_size, distance=Distance.COSINE),
        sparse_vectors_config={'langchain-sparse': SparseVectorParams(index=None, modifier=None)}
    )
# TODO: move to LLM files later
vector_store = QdrantVectorStore(
    client=client,
    collection_name=app_settings.vector_db_collection_name,
    embedding=embeddings,
)

# ------
# Move to `conversation/main`` later
@tool(response_format="content_and_artifact")
def retrieve(query: str):
    """Retrieve information related to a query."""
    retrieved_docs = vector_store.similarity_search(query, k=10)
    serialized = "\n\n".join(
        (f"Source: {doc.metadata}\n" f"Content: {doc.page_content}")
        for doc in retrieved_docs
    )
    return serialized, retrieved_docs


graph_builder = StateGraph(MessagesState)
memory = MemorySaver()

tool_node = ToolNode([retrieve])

graph_builder.add_node(trigger_ai_message_with_tool_call)
graph_builder.add_node("tool_node", tool_node)
graph_builder.add_node(generate)

graph_builder.set_entry_point("trigger_ai_message_with_tool_call")
graph_builder.add_edge("trigger_ai_message_with_tool_call", "tool_node")
graph_builder.add_edge("tool_node", "generate")
graph_builder.add_edge("generate", END)

graph = graph_builder.compile(checkpointer=memory)
# -----

def bot(message, history, source_history, chat_hash) -> list[Any]:
    """Generate bot response and history from message.
    
    With multi-modal inputs text and each file is treated as separate message.
    """
   
    logger.info("This is the history", history=history)
    
    # enable message edit
    if isinstance(message, str):
        message = {"text": message}
        
    # process files
    for file in message.get("files"):
        logger.info("Received file", file=file)
        ingest_document(file, vector_store)
    
    # create text response
    user_input_prompt = message.get("text")
    logger.info("This is the current chat_hash", chat_hash=chat_hash)

    config = {"configurable": {"thread_id": chat_hash}, "callbacks": [langfuse_handler], "stream": False}

    response = graph.invoke(
        {"messages": [{"role": "user", "content": user_input_prompt}]},
        config=config,
        )
    
    logger.info("Generated a response", response=response)

    logger.info("This is the source history before appending response", source_history=source_history)
    # Append source details of the response to source history
    source_history = build_source_history_object(response, source_history, user_input_prompt)

    return [response["messages"][-1].content], source_history, chat_hash


with gr.Blocks() as demo:
    # Initialize source history as session state variable
    source_history = gr.State([])
    chat_hash = gr.State(create_chat_hash)

    gr.Markdown("### Dies ist der aktuelle Prototyp der DEval AI4Kontextanalysen, mit dem Fokus auf dem Testen der Zitationsfunktion und der grundlegenden Konfiguration.")
    with gr.Tab("Chat"):
        chatbot = gr.Chatbot(type="messages")
        chatbot.clear(
            clear_chat_and_source_history,
            inputs=[source_history, chat_hash],
            outputs=[source_history, chat_hash],
        )
        gr.ChatInterface(
        bot,
        chatbot=chatbot,
        type="messages",
        flagging_mode="manual",
        editable=True, 
        flagging_options=["Like", "Dislike", "Spam", "Inappropriate", "Other"],
        multimodal=True,
        textbox=gr.MultimodalTextbox(file_count="multiple"),
        title="DEval Prototype 1",
        additional_inputs=[source_history, chat_hash],
        additional_outputs=[source_history, chat_hash],
        )
    with gr.Tab("Quellen"):
        prettified_source_history_md = gr.Markdown()
        # Update the source history UI immediately when the source_history value changes
        source_history.change(prettify_source_history, source_history, prettified_source_history_md)

if __name__ == "__main__":
    demo.launch(
        auth=(
            app_settings.test_user_name,
            app_settings.test_user_password
        ),
        ssr_mode=False
    )