import streamlit as st
import os
from crewai import Agent, Task, Crew, LLM
from pydantic import BaseModel, Field
from typing import List

# 1. DATA STRUCTURE 📐
# This ensures the AI gives us a clean "story" instead of raw JSON code.
class MatchReport(BaseModel):
    match_title: str = Field(description="The formal title of the match.")
    player_of_the_match: str = Field(description="The standout performer and why.")
    deep_narrative: str = Field(description="A 3-paragraph story of the match events.")
    key_highlights: List[str] = Field(description="3-5 bullet points of critical moments.")

# 2. SETUP & SECRETS 🔑
GROQ_API_KEY = os.environ.get("GROQ_API_KEY")

if not GROQ_API_KEY:
    st.error("Please set GROQ_API_KEY in your secrets.")
    st.stop()

# 3. LLM CONFIGURATION 🧠
# Updated LLM Configuration
cricket_llm = LLM(
    model="groq/llama-3.1-8b-instant",  # Updated supported ID
    api_key=GROQ_API_KEY
)

# 4. AGENT DEFINITIONS 🤖
# We define these FIRST so the tasks can find them.
scout = Agent(
    role="Cricket Match Scout",
    goal="Extract ball-by-ball data and Powerplay scores from {url}.",
    backstory="You are an expert at finding specific match facts.",
    llm=cricket_llm
)

analyst = Agent(
    role="Technical Match Analyst",
    goal="Identify momentum shifts and the Player of the Match.",
    backstory="You interpret data to find the 'why' behind the win.",
    llm=cricket_llm
)

writer = Agent(
    role="International Sports Journalist",
    goal="Write a deep, dramatic narration of the match.",
    backstory="You turn technical insights into a compelling story.",
    llm=cricket_llm
)

# 5. TASK DEFINITIONS 📋
scrape_task = Task(
    description="Analyze the commentary at {url}. Find key stats and partnerships.",
    agent=scout,
    expected_output="A list of match stats."
)

analyze_task = Task(
    description="Determine the turning points and the Player of the Match.",
    agent=analyst,
    expected_output="A technical summary of the game."
)

write_report_task = Task(
    description="Draft a deep narrative report. Focus on the drama and the Powerplay.",
    # We add this line to satisfy the Pydantic requirement:
    expected_output="A comprehensive match report following the MatchReport structure.",
    agent=writer,
    output_pydantic=MatchReport 
)

# 6. STREAMLIT UI 🖥️
st.title("🏏 Cricket Intelligence")

url_input = st.text_input("Enter Match Commentary URL:")

if st.button("Generate Report"):
    if url_input:
        with st.spinner("Analyzing match data..."):
            match_crew = Crew(
                agents=[scout, analyst, writer],
                tasks=[scrape_task, analyze_task, write_report_task],
                verbose=True, # This helps us see the agents working in real-time
                max_rpm=2     # Limits the crew to 2 requests every 60 seconds
            )
            
            result = match_crew.kickoff(inputs={'url': url_input})
            report = result.pydantic
            
            st.header(report.match_title)
            st.subheader(f"🏆 {report.player_of_the_match}")
            
            st.markdown("### ⚡ Highlights")
            for point in report.key_highlights:
                st.write(f"- {point}")
                
            st.divider()
            st.markdown("### 📖 Match Story")
            st.write(report.deep_narrative)
    else:
        st.warning("Please provide a URL.")