from typing import Dict, Any
from pydantic import BaseModel
from groq import Groq
import os

LLM_MODEL = os.getenv("GROQ_MODEL", "llama-3.1-8b-instant")  # fast + free-tier friendly

class StrategyContext(BaseModel):
    match_title: str
    phase: str                 # "Powerplay" | "Middle" | "Death"
    striker: str
    non_striker: str
    batting_team: str
    bowling_team: str
    candidate_bowlers: list    # from scoreboard / last used bowlers
    striker_recent: Dict[str, Any]   # simple summary, e.g., dots, 4s, 6s vs pace/spin
    innings_state: Dict[str, Any]    # score, over, req rate, wickets, venue (if available)

def build_prompt(ctx: StrategyContext) -> str:
    return f"""
You are an elite T20/ODI bowling strategist and fielding coach.

Match: {ctx.match_title}
Phase: {ctx.phase}
Batting: {ctx.batting_team}   Striker: {ctx.striker}, Non-striker: {ctx.non_striker}
Bowling: {ctx.bowling_team}
State: {ctx.innings_state}
Striker recent profile (last 2-3 overs): {ctx.striker_recent}
Candidate bowlers available: {", ".join(ctx.candidate_bowlers) if ctx.candidate_bowlers else "not provided"}

TASKS:
1) Recommend the next OVER BOWLER (1-2 options) with reasoning.
2) Give a BALL-BY-BALL plan for 6 balls: {phase_plan(ctx.phase)} lengths & lines (yorker, hard length, slowers, wide lines).
3) Provide FIELD PLACEMENT (11 positions) tailored to the plan.
4) Add "IF/THEN" adaptations (e.g., if batter charges, switch to ___).
5) Keep it concise and actionable.

Output format:
- Bowler:
- Plan (Balls 1-6):
- Field Setting:
- If/Then:
""".strip()

def phase_plan(phase: str) -> str:
    return {
        "Powerplay": "attack with swing, protect square boundaries",
        "Middle": "mix of pace-off and tight lines, choke singles",
        "Death": "yorkers/slow bouncers, avoid slot"
    }.get(phase, "balanced plans")

def call_groq(prompt: str) -> str:
    api_key = os.getenv("GROQ_API_KEY")
    if not api_key:
        return "[Missing GROQ_API_KEY] Please set an environment variable."
    client = Groq(api_key=api_key)
    resp = client.chat.completions.create(
        model=LLM_MODEL,
        messages=[{"role": "system", "content": "Be precise, tactical, and match-aware."},
                  {"role": "user", "content": prompt}],
        temperature=0.5,
        max_tokens=600
    )
    return resp.choices[0].message.content

def suggest_strategy(ctx: StrategyContext) -> str:
    prompt = build_prompt(ctx)
    return call_groq(prompt)