atc-agents/api/presentation_upload.py

"""Upload PPT/PPTX decks and convert to presentation JSON + HTML."""

from __future__ import annotations

import json
import os
import re
import uuid
from datetime import datetime, timezone
from pathlib import Path
from typing import Any

import httpx

PRESENTATIONS_DIR = Path(os.getenv("PRESENTATIONS_DIR", "/data/presentations"))
DOCLING_URL = os.getenv("DOCLING_URL", "http://docling-serve:5001").rstrip("/")


def _ensure_dir() -> Path:
    PRESENTATIONS_DIR.mkdir(parents=True, exist_ok=True)
    return PRESENTATIONS_DIR


def list_decks() -> list[dict[str, Any]]:
    _ensure_dir()
    decks = []
    for meta_path in sorted(PRESENTATIONS_DIR.glob("*/meta.json"), key=lambda p: p.stat().st_mtime, reverse=True):
        try:
            meta = json.loads(meta_path.read_text())
            decks.append(meta)
        except Exception:
            continue
    return decks


def _safe_name(name: str) -> str:
    return re.sub(r"[^a-zA-Z0-9._-]+", "_", name)[:80]


def pptx_to_slides(path: Path) -> list[dict[str, Any]]:
    from pptx import Presentation
    from pptx.enum.shapes import MSO_SHAPE_TYPE

    prs = Presentation(str(path))
    slides: list[dict[str, Any]] = []
    for idx, slide in enumerate(prs.slides, start=1):
        bullets: list[str] = []
        title = ""
        for shape in slide.shapes:
            if not hasattr(shape, "text"):
                continue
            text = (shape.text or "").strip()
            if not text:
                continue
            if shape.shape_type == MSO_SHAPE_TYPE.PLACEHOLDER and not title:
                title = text.split("\n")[0][:120]
            else:
                for line in text.split("\n"):
                    line = line.strip()
                    if line and line != title:
                        bullets.append(line[:240])
        if not title:
            title = f"Slide {idx}"
        slides.append({
            "id": f"upload-{idx}",
            "title": title,
            "subtitle": "",
            "bullets": bullets[:12] or ["(empty slide)"],
            "kind": "upload",
        })
    return slides


async def docling_enrich(path: Path) -> list[dict[str, Any]] | None:
    """Optional: parse via Docling for richer structure."""
    try:
        async with httpx.AsyncClient(timeout=120.0) as client:
            with path.open("rb") as f:
                r = await client.post(
                    f"{DOCLING_URL}/v1/convert/file",
                    files={"files": (path.name, f, "application/octet-stream")},
                    data={"to_formats": "md"},
                )
            if r.status_code >= 400:
                return None
            data = r.json()
            md = ""
            if isinstance(data, dict):
                doc = data.get("document") or data.get("result") or data
                if isinstance(doc, dict):
                    md = doc.get("md_content") or doc.get("markdown") or ""
                elif isinstance(doc, str):
                    md = doc
            if not md:
                return None
            slides = []
            chunks = [c.strip() for c in re.split(r"\n#{1,2}\s+", md) if c.strip()]
            for i, chunk in enumerate(chunks[:40], start=1):
                lines = [ln.strip() for ln in chunk.split("\n") if ln.strip()]
                title = lines[0][:120] if lines else f"Slide {i}"
                bullets = [ln.lstrip("-•* ").strip() for ln in lines[1:13] if ln.strip()]
                slides.append({
                    "id": f"docling-{i}",
                    "title": title,
                    "subtitle": "Docling parsed",
                    "bullets": bullets or ["—"],
                    "kind": "upload",
                })
            return slides if slides else None
    except Exception:
        return None


async def save_upload(filename: str, content: bytes) -> dict[str, Any]:
    _ensure_dir()
    deck_id = str(uuid.uuid4())[:8]
    deck_dir = PRESENTATIONS_DIR / deck_id
    deck_dir.mkdir(parents=True, exist_ok=True)
    safe = _safe_name(filename)
    dest = deck_dir / safe
    dest.write_bytes(content)

    slides: list[dict[str, Any]] = []
    source = "pptx"
    if safe.lower().endswith((".pptx", ".ppt")):
        slides = pptx_to_slides(dest)
        docling_slides = await docling_enrich(dest)
        if docling_slides and len(docling_slides) >= len(slides):
            slides = docling_slides
            source = "docling+pptx"
    else:
        docling_slides = await docling_enrich(dest)
        if docling_slides:
            slides = docling_slides
            source = "docling"

    if not slides:
        slides = [{
            "id": "upload-1",
            "title": safe,
            "subtitle": "Uploaded file",
            "bullets": [f"File stored at {dest.name}", "Could not auto-parse slides — open in editor or re-upload PPTX"],
            "kind": "upload",
        }]

    payload = {
        "id": deck_id,
        "filename": safe,
        "source": source,
        "ts": datetime.now(timezone.utc).isoformat(),
        "title": safe.rsplit(".", 1)[0],
        "subtitle": "Uploaded presentation",
        "slide_count": len(slides),
        "slides": slides,
    }
    (deck_dir / "meta.json").write_text(json.dumps(payload, indent=2, default=str))
    (deck_dir / "deck.json").write_text(json.dumps(payload, default=str))
    return payload


def get_deck(deck_id: str) -> dict[str, Any] | None:
    path = PRESENTATIONS_DIR / deck_id / "meta.json"
    if not path.exists():
        return None
    return json.loads(path.read_text())