app.py

# ------------------------------
# Imports
# ------------------------------
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import Chroma
from langchain_core.messages import HumanMessage, AIMessage, ToolMessage
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langgraph.graph import END, StateGraph
from langgraph.prebuilt import ToolNode
from langgraph.graph.message import add_messages
from typing_extensions import TypedDict, Annotated
from typing import Sequence, List, Dict, Any
import chromadb
import re
import os
import streamlit as st
import requests
import time
import hashlib
from langchain.tools.retriever import create_retriever_tool
from datetime import datetime

# ------------------------------
# Data
# ------------------------------
research_texts = [
    "Research Report: Results of a New AI Model Improving Image Recognition Accuracy to 98%",
    "Academic Paper Summary: Why Transformers Became the Mainstream Architecture in Natural Language Processing",
    "Latest Trends in Machine Learning Methods Using Quantum Computing",
    "Advancements in Neuromorphic Computing for Energy-Efficient AI Systems",
    "Cross-Modal Learning: Integrating Visual and Textual Representations for Multimodal AI"
]

development_texts = [
    "Project A: UI Design Completed, API Integration in Progress",
    "Project B: Testing New Feature X, Bug Fixes Needed",
    "Product Y: In the Performance Optimization Stage Before Release",
    "Framework Z: Version 3.2 Released with Enhanced Distributed Training Support",
    "DevOps Pipeline: Automated CI/CD Implementation for ML Model Deployment"
]

# ------------------------------
# Configuration
# ------------------------------
class AppConfig:
    def __init__(self):
        self.DEEPSEEK_API_KEY = os.environ.get("DEEPSEEK_API_KEY")
        self.CHROMA_PATH = "chroma_db"
        self.MAX_RETRIES = 3
        self.RETRY_DELAY = 1.5
        self.DOCUMENT_CHUNK_SIZE = 300
        self.DOCUMENT_OVERLAP = 50
        self.SEARCH_K = 5
        self.SEARCH_TYPE = "mmr"

    def validate(self):
        if not self.DEEPSEEK_API_KEY:
            st.error("""
            **Configuration Error**  
            🔑 Missing DeepSeek API key.  
            Configure through Hugging Face Space secrets:
            1. Space Settings → Repository secrets
            2. Add secret: DEEPSEEK_API_KEY=your_key
            3. Rebuild Space
            """)
            st.stop()

class ChromaManager:
    def __init__(self, config: AppConfig):
        os.makedirs(config.CHROMA_PATH, exist_ok=True)
        self.client = chromadb.PersistentClient(path=config.CHROMA_PATH)
        self.embeddings = OpenAIEmbeddings(model="text-embedding-3-large")
        
        self.research_collection = self._create_collection(
            research_texts, 
            "research_collection",
            {"category": "research"}
        )
        self.dev_collection = self._create_collection(
            development_texts,
            "development_collection", 
            {"category": "development"}
        )

    def _create_collection(self, documents: List[str], name: str, metadata: dict) -> Chroma:
        splitter = RecursiveCharacterTextSplitter(
            chunk_size=300,
            chunk_overlap=50,
            separators=["\n\n", "\n", "。"]
        )
        docs = splitter.create_documents(documents)
        return Chroma.from_documents(
            documents=docs,
            embedding=self.embeddings,
            client=self.client,
            collection_name=name,
            collection_metadata=metadata
        )

# ------------------------------
# Document Processing
# ------------------------------
class DocumentProcessor:
    @staticmethod
    def deduplicate(docs: List[Any]) -> List[Any]:
        seen = set()
        return [doc for doc in docs 
                if not (hashlib.md5(doc.page_content.encode()).hexdigest() in seen 
                        or seen.add(hashlib.md5(doc.page_content.encode()).hexdigest()))]

    @staticmethod
    def extract_keypoints(docs: List[Any]) -> str:
        categories = {
            "quantum": ["quantum", "qubit"],
            "vision": ["image", "recognition"],
            "nlp": ["transformer", "language"]
        }
        return "\n".join(sorted({
            "- " + {
                "quantum": "Quantum computing breakthroughs",
                "vision": "Computer vision advancements",
                "nlp": "NLP architecture innovations"
            }[cat] 
            for doc in docs 
            for cat, kw in categories.items() 
            if any(k in doc.page_content.lower() for k in kw)
        }))

# ------------------------------
# Workflow
# ------------------------------
class AgentWorkflow:
    def __init__(self, chroma: ChromaManager):
        self.chroma = chroma
        self.workflow = StateGraph(AgentState)
        
        # Define nodes
        self.workflow.add_node("agent", self.agent)
        self.workflow.add_node("retrieve", ToolNode([
            create_retriever_tool(
                chroma.research_collection.as_retriever(),
                "research_tool",
                "Search research documents"
            ),
            create_retriever_tool(
                chroma.dev_collection.as_retriever(),
                "dev_tool",
                "Search development updates"
            )
        ]))
        self.workflow.add_node("generate", self.generate)
        self.workflow.add_node("rewrite", self.rewrite)

        # Define edges
        self.workflow.set_entry_point("agent")
        self.workflow.add_conditional_edges(
            "agent",
            self._tools_condition,
            {"retrieve": "retrieve", "end": END}
        )
        self.workflow.add_conditional_edges(
            "retrieve",
            self._grade_documents,
            {"generate": "generate", "rewrite": "rewrite"}
        )
        self.workflow.add_edge("generate", END)
        self.workflow.add_edge("rewrite", "agent")
        
        self.app = self.workflow.compile()

    class AgentState(TypedDict):
        messages: Annotated[Sequence[AIMessage | HumanMessage | ToolMessage], add_messages]

    def agent(self, state: AgentState):
        try:
            messages = state["messages"]
            query = messages[-1].content if isinstance(messages[-1], HumanMessage) else messages[-1]['content']
            
            response = requests.post(
                "https://api.deepseek.com/v1/chat/completions",
                headers={"Authorization": f"Bearer {config.DEEPSEEK_API_KEY}"},
                json={
                    "model": "deepseek-chat",
                    "messages": [{
                        "role": "user",
                        "content": f"""Analyze this query: "{query}"
                        Respond EXACTLY as:
                        - SEARCH_RESEARCH: <terms> (for research topics)
                        - SEARCH_DEV: <terms> (for development updates)
                        - DIRECT: <answer> (otherwise)"""
                    }]
                }
            ).json()
            
            content = response['choices'][0]['message']['content']
            if "SEARCH_RESEARCH:" in content:
                terms = content.split("SEARCH_RESEARCH:")[1].strip()
                results = self.chroma.research_collection.similarity_search(terms)
                return {"messages": [AIMessage(content=f"Research Results: {str(results)}")]}
            elif "SEARCH_DEV:" in content:
                terms = content.split("SEARCH_DEV:")[1].strip()
                results = self.chroma.dev_collection.similarity_search(terms)
                return {"messages": [AIMessage(content=f"Development Results: {str(results)}")]}
            return {"messages": [AIMessage(content=content)]}
            
        except Exception as e:
            return {"messages": [AIMessage(content=f"Error: {str(e)}")]}

    def generate(self, state: AgentState):
        docs = eval(state["messages"][-1].content.split("Results: ")[1])
        processed = "\n".join([d.page_content[:200] for d in DocumentProcessor.deduplicate(docs)])
        
        response = requests.post(
            "https://api.deepseek.com/v1/chat/completions",
            headers={"Authorization": f"Bearer {config.DEEPSEEK_API_KEY}"},
            json={
                "model": "deepseek-chat",
                "messages": [{
                    "role": "user",
                    "content": f"Summarize these findings:\n{processed}"
                }]
            }
        ).json()
        
        return {"messages": [AIMessage(content=response['choices'][0]['message']['content'])]}

    def rewrite(self, state: AgentState):
        original = state["messages"][0].content
        response = requests.post(
            "https://api.deepseek.com/v1/chat/completions",
            headers={"Authorization": f"Bearer {config.DEEPSEEK_API_KEY}"},
            json={
                "model": "deepseek-chat",
                "messages": [{
                    "role": "user",
                    "content": f"Rephrase this query: {original}"
                }]
            }
        ).json()
        return {"messages": [AIMessage(content=response['choices'][0]['message']['content'])]}

    def _tools_condition(self, state: AgentState):
        return "retrieve" if "Results:" in state["messages"][-1].content else "end"

    def _grade_documents(self, state: AgentState):
        return "generate" if len(eval(state["messages"][-1].content.split("Results: ")[1])) > 0 else "rewrite"

# ------------------------------
# Streamlit App
# ------------------------------
def apply_theme():
    st.markdown("""
    <style>
    .stApp { background: #1a1a1a; color: white; }
    .stTextArea textarea { background: #2d2d2d !important; color: white !important; }
    .stButton>button { background: #2E86C1; transition: 0.3s; }
    .stButton>button:hover { background: #1B4F72; transform: scale(1.02); }
    .data-box { background: #2d2d2d; border-left: 4px solid #2E86C1; padding: 15px; margin: 10px 0; }
    </style>
    """, unsafe_allow_html=True)

def main():
    apply_theme()
    
    with st.sidebar:
        st.header("📚 Databases")
        with st.expander("Research", expanded=True):
            for text in research_texts:
                st.markdown(f'<div class="data-box">{text}</div>', unsafe_allow_html=True)
        with st.expander("Development"):
            for text in development_texts:
                st.markdown(f'<div class="data-box">{text}</div>', unsafe_allow_html=True)

    st.title("🔍 AI Research Assistant")
    query = st.text_area("Enter your query:", height=100)
    
    if st.button("Analyze"):
        with st.spinner("Processing..."):
            workflow = AgentWorkflow(chroma_manager)
            results = workflow.app.invoke({"messages": [HumanMessage(content=query)]})
            
            with st.expander("Processing Details", expanded=True):
                st.write("### Raw Results", results)
                
            st.success("### Final Answer")
            st.markdown(results['messages'][-1].content)

# ------------------------------
# Initialization
# ------------------------------
if __name__ == "__main__":
    st.set_page_config(
        page_title="AI Research Assistant",
        layout="wide",
        initial_sidebar_state="expanded"
    )
    
    try:
        config = AppConfig()
        config.validate()
        chroma_manager = ChromaManager(config)
        main()
    except Exception as e:
        st.error(f"Initialization failed: {str(e)}")