# app.py
import os
import streamlit as st
import model  # Importing our structured backend module

# --- 1. Web Page Meta Architecture ---
st.set_page_config(
    page_title="Custom Multi-Agent Persona Sandbox",
    page_icon="🧬",
    layout="wide"
)

st.title("🧬 Custom Multi-Agent Persona Sandbox")
st.caption("Select your specialized AI expert, inject live files/source code, and preview reviewed data outputs.")

# --- 2. Sidebar Layout Engine (Token & LLM Settings) ---
st.sidebar.header("⚙️ LLM Infrastructure Configuration")

# Extract defaults from Environment Secrets if available
hf_token_env = os.getenv("HUGGINGFACEHUB_API_TOKEN") or os.getenv("HF_TOKEN")
openai_token_env = os.getenv("OPENAI_API_KEY")

engine_provider = st.sidebar.radio(
    "Choose LLM Engine Provider",
    options=["Hugging Face Serverless (Default)", "OpenAI (Optional Override)"]
)

# Conditional infrastructure state rendering
user_token = ""
if engine_provider == "Hugging Face Serverless (Default)":
    user_token = st.sidebar.text_input(
        "Hugging Face Access Token",
        value=hf_token_env if hf_token_env else "",
        type="password",
        placeholder="hf_..."
    )
    selected_model = st.sidebar.selectbox(
        "Model Option",
        options=["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.2"]
    )
else:
    user_token = st.sidebar.text_input(
        "OpenAI API Key",
        value=openai_token_env if openai_token_env else "",
        type="password",
        placeholder="sk-..."
    )
    selected_model = st.sidebar.selectbox(
        "Model Option",
        options=["gpt-4o-mini", "gpt-4-turbo", "gpt-3.5-turbo"]
    )

# --- 3. Interactive Main Canvas Columns ---
col_left, col_right = st.columns([1, 1])

with col_left:
    st.subheader("📋 Step 1: Design Agent Workspace")
    selected_persona = st.selectbox(
        "Choose Your Target AI Agent Persona:", 
        options=list(model.AGENT_PERSONAS.keys())
    )
    user_prompt = st.text_area(
        "✍️ Task Objective / Question:", 
        placeholder="Type your core question or instructions for the agent here...", 
        height=120
    )

with col_right:
    st.subheader("📎 Step 2: Inject Knowledge Base Assets")
    uploaded_files = st.file_uploader(
        "Upload reference materials (.py, .ipynb, .pdf, .txt)",
        accept_multiple_files=True
    )

# --- 4. Processing and Core Execution Trigger ---
if st.button("🚀 Run Agent Pipeline", type="primary"):
    if not user_prompt.strip():
        st.warning("Please specify a prompt instruction or goal before running.")
    else:
        # Build document knowledge context block
        injected_context = ""
        if uploaded_files:
            with st.spinner("Extracting multi-format source data elements..."):
                for uploaded_f in uploaded_files:
                    # Parse the bytes using our backend function
                    file_content = model.parse_uploaded_file_content(
                        uploaded_f.name, 
                        uploaded_f.read()
                    )
                    injected_context += file_content
                    
        # Trigger Multi-Agent Crew
        with st.status("🧠 Agents Collaborating... (Generating & Auditing Answers)", expanded=True) as status:
            try:
                final_output = model.run_agent_pipeline(
                    provider=engine_provider,
                    model_name=selected_model,
                    token=user_token,
                    persona_key=selected_persona,
                    user_prompt=user_prompt,
                    context_text=injected_context
                )
                status.update(label="✅ Review Passed & Processing Complete!", state="complete")
                
                # Render Clean Reviewed Content out to markdown canvas
                st.subheader(f"🏁 Final Output (Reviewed by QA Auditor)")
                st.markdown(final_output)
                
            except Exception as e:
                status.update(label="❌ Pipeline Execution Failed", state="error")
                st.error(f"An error occurred during agent orchestration: `{str(e)}`")