from __future__ import annotations

import io
import json
import logging
import os
from pathlib import Path
from typing import List

import gradio as gr
from PIL import Image

from llm_clients import GEMINI_3_VISION, OPENAI_GPT5, OPENAI_GPT5_MINI
from pipeline import DEFAULT_USER_PROMPT, process_images
from settings import load_settings

logging.basicConfig(level=logging.INFO)
LOGGER = logging.getLogger("app")


def _load_images(files: List[gr.File], downscale: bool) -> List[bytes]:
    images: List[bytes] = []
    for f in files or []:
        data = Path(f.name).read_bytes()
        if downscale:
            try:
                img = Image.open(io.BytesIO(data)).convert("RGB")
                img.thumbnail((1024, 1024))
                buf = io.BytesIO()
                img.save(buf, format="PNG")
                data = buf.getvalue()
            except Exception as exc:  # noqa: BLE001
                LOGGER.warning("Downscale failed for %s: %s; using original", f.name, exc)
        images.append(data)
    return images


def make_interface():
    settings = load_settings()
    settings.require_api_keys()

    def _infer(files, model, creativity, downscale_images):
        images = _load_images(files, downscale_images)
        if not images:
            raise gr.Error("Please upload at least one image.")

        try:
            result = process_images(
                images,
                model,
                settings,
                system_prompt_path=None,
                user_prompt=DEFAULT_USER_PROMPT,
            )
        except Exception as exc:  # noqa: BLE001
            LOGGER.exception("Inference failed")
            raise gr.Error(str(exc))

        trends = result["trends"]
        bullets = result["summary"]
        md = "\n\n".join(f"- {b}" for b in bullets) if bullets else "No summary available."
        trends_text = json.dumps(trends, indent=2)
        return trends_text, md

    def _on_files_change(files):
        """Update preview and clear outputs when files are removed."""
        if not files:
            return [], "", ""
        return files, gr.update(), gr.update()

    enable_gemini = os.environ.get("ENABLE_GEMINI", "true").lower() not in {"0", "false", "no"}

    with gr.Blocks(title="Garment Micro-Trend Detector") as demo:
        gr.Markdown(
            "Upload garment image(s), pick a model, and get structured micro-trend JSON plus a bullet summary."
        )

        with gr.Row():
            image_input = gr.Files(file_count="multiple", label="Garment images")
            with gr.Column():
                model_choices = [OPENAI_GPT5, OPENAI_GPT5_MINI]
                if enable_gemini:
                    model_choices.append(GEMINI_3_VISION)
                default_model = settings.openai_model if settings.openai_model in model_choices else OPENAI_GPT5_MINI
                model_dd = gr.Dropdown(
                    choices=model_choices,
                    value=default_model,
                    label="Model",
                    allow_custom_value=True,  # allow custom OpenAI model overrides like gpt-5.1
                )
                creativity = gr.Slider(
                    minimum=0.0,
                    maximum=1.0,
                    step=0.1,
                    value=0.2,
                    label="Creativity (temperature hint)",
                    info="Not all models use this directly; for now it is informational.",
                )
                downscale_chk = gr.Checkbox(value=True, label="Downscale images to 1024px for speed/cost")
                run_btn = gr.Button("Analyze", variant="primary")

        preview = gr.Gallery(
            label="Preview",
            show_label=True,
            object_fit="contain",  # preserve aspect ratio
            height="auto",
        )
        json_out = gr.Textbox(label="Micro-trend JSON", lines=20, interactive=False)
        summary_md = gr.Markdown(label="Summary")

        image_input.change(_on_files_change, inputs=image_input, outputs=[preview, json_out, summary_md], queue=False)

        run_btn.click(
            _infer,
            inputs=[image_input, model_dd, creativity, downscale_chk],
            outputs=[json_out, summary_md],
            queue=True,
        )

    return demo


if __name__ == "__main__":
    app = make_interface()
    app.queue()
    app.launch(
        server_name="0.0.0.0",
        server_port=int(os.environ.get("PORT", 7860)),
        share=False,  # Spaces provides its own public URL; share=True not supported there
        show_api=False,  # avoid schema generation that can crash on HF
    )