import os
import asyncio
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain_chroma import Chroma
from langchain_core.documents import Document
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain.tools import tool
from deepagents import create_deep_agent
from deepagents.backends import FilesystemBackend, LocalShellBackend, CompositeBackend
from langchain_core.messages import HumanMessage
from tavily import TavilySearchResults

# ---------------------
# Configuration
# ---------------------
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
TAVILY_API_KEY = os.getenv("TAVILY_API_KEY")

# ---------------------
# Vector store utilities
# ---------------------

def create_vectorstore(persist_directory: str = "./chroma_db") -> Chroma:
    """Create or load a Chroma vector store with OpenAI embeddings."""
    embeddings = OpenAIEmbeddings(
        model="text-embedding-3-small",
        base_url="https://openrouter.ai/api/v1",
        api_key=OPENAI_API_KEY,
    )
    return Chroma(
        collection_name="knowledge",
        embedding_function=embeddings,
        persist_directory=persist_directory,
    )


def load_documents(directory: str, vectorstore: Chroma) -> None:
    """Load .txt and .md files from *directory* into *vectorstore* using chunking."""
    splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    docs = []
    for root, _, files in os.walk(directory):
        for fname in files:
            if fname.lower().endswith(('.txt', '.md')):
                path = os.path.join(root, fname)
                with open(path, "r", encoding="utf-8") as f:
                    text = f.read()
                # Create Document objects
                docs.extend(
                    [Document(page_content=chunk, metadata={"source": path}) for chunk in splitter.split_text(text)]
                )
    if docs:
        vectorstore.add_documents(docs)
        vectorstore.persist()

# ---------------------
# Tools
# ---------------------

vectorstore = create_vectorstore()

# Ensure we have some data loaded – load from ./documents if collection empty
if len(vectorstore.get_all_documents()) == 0:
    load_documents("./documents", vectorstore)

@tool
def search_local_kb(query: str, top_k: int = 3) -> str:
    """Search the local knowledge base for relevant passages."""
    docs = vectorstore.similarity_search(query, k=top_k)
    if not docs:
        return "[Local KB] No relevant information found."
    result = "\n\n".join([f"{d.metadata.get('source', 'unknown')}\n{d.page_content}" for d in docs])
    return f"[Local KB]\n{result}"

@tool
def web_search(query: str) -> str:
    """Perform a web search using Tavily and return top results."""
    results = TavilySearchResults(query=query, max_results=3, api_key=TAVILY_API_KEY)
    if not results.results:
        return "[Web Search] No results found."
    snippets = []
    for r in results.results:
        snippets.append(f"{r.get('title', 'No title')}\n{r.get('content', 'No content')}\nURL: {r.get('url', '')}")
    return f"[Web Search]\n\n".join(snippets)

# ---------------------
# Agent setup
# ---------------------

llm = ChatOpenAI(
    model="openai/gpt-oss-20b:free",
    base_url="https://openrouter.ai/api/v1",
    api_key=OPENAI_API_KEY,
    temperature=0.0,
)

backend = CompositeBackend([
    LocalShellBackend(workspace_dir="./workspace"),
    FilesystemBackend(),
])

system_prompt = (
    "You are a helpful RAG agent. For questions about local documents, use the tool `search_local_kb`. "
    "For current news or facts that may not be in your local knowledge base, use `web_search`. "
    "Return the answer prefixed with either `[Local KB]` or `[Web Search]` to indicate the source."
)

agent = create_deep_agent(
    model=llm,
    tools=[search_local_kb, web_search],
    backend=backend,
    system_prompt=system_prompt,
)

# ---------------------
# CLI loop
# ---------------------

async def main():
    print("RAG Agent ready. Type your question (or 'exit' to quit).")
    while True:
        user_input = input("\nQuery: ")
        if user_input.lower() in {"exit", "quit", "q"}:
            print("Goodbye!")
            break
        # Invoke agent
        result = await agent.ainvoke(
            {"messages": [HumanMessage(content=user_input)]},
            {"configurable": {"thread_id": "session-1"}},
        )
        # The agent returns a list of messages; get the last one
        reply = result["messages"][-1].content
        print(f"\n{reply}")

if __name__ == "__main__":
    asyncio.run(main())