Optional[Path]:
+ raw = (path_value or "").strip()
+ if not raw:
+ return None
+
+ candidate = Path(raw).expanduser()
+ if not candidate.is_absolute():
+ candidate = (PROJECT_ROOT / candidate).resolve()
+
+ if candidate.is_file():
+ return candidate
+
+ log.warning("[paper2poster] output file missing on disk: %s", raw)
+ return None
+
@staticmethod
def _validate_poster_dimensions(width: float, height: float) -> None:
if width <= 0 or height <= 0:
@@ -85,6 +102,16 @@ async def generate(
api_key,
scope="paper2poster",
)
+ model = resolve_model_name(
+ model,
+ managed_default=settings.PAPER2POSTER_DEFAULT_MODEL,
+ fallback_default="gpt-4o",
+ )
+ vision_model = resolve_model_name(
+ vision_model,
+ managed_default=settings.PAPER2POSTER_VISION_MODEL,
+ fallback_default="gpt-4o",
+ )
self._validate_poster_dimensions(poster_width, poster_height)
run_dir = self._create_run_dir(email)
@@ -138,13 +165,16 @@ async def generate(
raise HTTPException(status_code=500, detail=result.get("message") or "Failed to generate poster")
pptx_path = (result.get("output_pptx_path") or "").strip()
- if not pptx_path:
- raise HTTPException(status_code=500, detail="Poster workflow finished without a PPTX output")
+ pptx_file = self._resolve_existing_output_file(pptx_path)
+ if pptx_file is None:
+ detail = result.get("message") or "Poster workflow finished without a valid PPTX output"
+ raise HTTPException(status_code=500, detail=detail)
png_path = (result.get("output_png_path") or "").strip()
+ png_file = self._resolve_existing_output_file(png_path) if png_path else None
return {
"success": True,
- "pptx_url": _to_outputs_url(pptx_path),
- "png_url": _to_outputs_url(png_path) if png_path else None,
+ "pptx_url": _to_outputs_url(str(pptx_file)),
+ "png_url": _to_outputs_url(str(png_file)) if png_file else None,
"message": "Poster generated successfully",
}
diff --git a/fastapi_app/services/paper2ppt_frontend_service.py b/fastapi_app/services/paper2ppt_frontend_service.py
index 61b6e421..7315835a 100644
--- a/fastapi_app/services/paper2ppt_frontend_service.py
+++ b/fastapi_app/services/paper2ppt_frontend_service.py
@@ -30,6 +30,7 @@
from fastapi_app.services.managed_api_service import (
resolve_image_generation_credentials,
resolve_llm_credentials,
+ resolve_model_name,
)
from fastapi_app.utils import _from_outputs_url, _to_outputs_url, resolve_outputs_path
@@ -100,7 +101,11 @@ async def generate_slides(
pagecontent=pagecontent,
chat_api_url=resolved_chat_api_url,
api_key=resolved_api_key,
- model=req.model,
+ model=resolve_model_name(
+ req.model,
+ managed_default=settings.PAPER2PPT_CONTENT_MODEL,
+ fallback_default=settings.PAPER2PPT_DEFAULT_MODEL,
+ ),
language=req.language,
style=req.style,
)
@@ -116,12 +121,20 @@ async def generate_slides(
slide_index=req.page_id,
chat_api_url=resolved_chat_api_url,
api_key=resolved_api_key,
- model=req.model,
+ model=resolve_model_name(
+ req.model,
+ managed_default=settings.PAPER2PPT_CONTENT_MODEL,
+ fallback_default=settings.PAPER2PPT_DEFAULT_MODEL,
+ ),
language=req.language,
style=req.style,
include_images=req.include_images,
image_style=req.image_style,
- image_model=req.image_model,
+ image_model=resolve_model_name(
+ req.image_model,
+ managed_default=settings.PAPER2PPT_IMAGE_GEN_MODEL,
+ fallback_default=settings.PAPER2PPT_DEFAULT_IMAGE_MODEL,
+ ),
image_api_url=resolved_image_api_url,
image_api_key=resolved_image_api_key,
edit_prompt=req.edit_prompt,
@@ -178,12 +191,20 @@ async def generate_slides(
slide_index=index,
chat_api_url=resolved_chat_api_url,
api_key=resolved_api_key,
- model=req.model,
+ model=resolve_model_name(
+ req.model,
+ managed_default=settings.PAPER2PPT_CONTENT_MODEL,
+ fallback_default=settings.PAPER2PPT_DEFAULT_MODEL,
+ ),
language=req.language,
style=req.style,
include_images=req.include_images,
image_style=req.image_style,
- image_model=req.image_model,
+ image_model=resolve_model_name(
+ req.image_model,
+ managed_default=settings.PAPER2PPT_IMAGE_GEN_MODEL,
+ fallback_default=settings.PAPER2PPT_DEFAULT_IMAGE_MODEL,
+ ),
image_api_url=resolved_image_api_url,
image_api_key=resolved_image_api_key,
edit_prompt=None,
@@ -846,16 +867,12 @@ def _build_visual_asset_prompt(
"sci_fi": "restrained sci-fi research visual with clean lighting",
"flat_infographic": "flat infographic-style illustration with simple shapes",
}
- key_points = [
- str(item).strip()
- for item in (outline_item.get("key_points") or [])
- if str(item).strip()
- ][:4]
+ key_points = self._normalize_outline_points(outline_item.get("key_points"), limit=4, item_limit=120)
palette = theme.get("palette") or {}
return (
"Create one supporting image for an academic presentation slide. "
- f"Page topic: {str(outline_item.get('title') or f'Slide {slide_index + 1}').strip()}. "
- f"Layout intent: {str(outline_item.get('layout_description') or '').strip()}. "
+ f"Page topic: {self._clean_text_content(outline_item.get('title'), f'Slide {slide_index + 1}', 220)}. "
+ f"Layout intent: {self._clean_text_content(outline_item.get('layout_description'), '', 220)}. "
f"Key points: {'; '.join(key_points) if key_points else 'keep it concise and presentation-friendly'}. "
f"Visual style: {style_map.get(image_style, image_style or 'academic illustration')}. "
f"Preferred palette anchors: background {palette.get('bg', '#0b1020')}, accent {palette.get('accent', '#f59e0b')}, text contrast {palette.get('text', '#e2e8f0')}. "
@@ -1145,6 +1162,7 @@ def _build_theme_messages(
{
"theme_name": "short id",
"visual_mood": "one sentence",
+ "style_family": "modern | business | academic | creative",
"palette": {
"bg": "#0b1020",
"panel": "rgba(15,23,42,0.92)",
@@ -1183,6 +1201,7 @@ def _build_theme_messages(
6. The theme_lock must be concrete enough to prevent per-slide drift during later regeneration.
7. If style_prompt contains explicit color or material directions, translate them into the palette instead of ignoring them.
8. Do not default to cyan/teal accents unless the style_prompt clearly asks for them.
+9. style_family must be one of modern, business, academic, creative and should match the tone implied by style_prompt.
""".strip()
user_payload = {
@@ -1218,39 +1237,51 @@ def _build_messages(
visual_assets: List[Dict[str, Any]],
) -> List[Dict[str, Any]]:
system_prompt = """
-You are an expert academic slide frontend engineer.
-Generate a single 16:9 presentation slide as HTML/CSS for a browser-based PPT editor.
+You are an expert academic presentation designer.
+Generate one strictly structured 16:9 slide for a browser PPT editor and true editable PPT export.
Hard requirements:
1. Return JSON only. No markdown fences. No explanation.
2. Output schema:
{
"title": "short string",
- "html_template": "HTML string",
- "css_code": "CSS string",
- "editable_fields": [
- {"key": "title", "label": "Title", "type": "text", "value": "..."},
- {"key": "summary", "label": "Summary", "type": "textarea", "value": "..."},
- {"key": "key_points", "label": "Key Points", "type": "list", "items": ["...", "..."]}
- ],
+ "layout_type": "cover | section | bullets | two_column | cards_2x2 | image_focus | comparison | timeline",
+ "content": {
+ "...": "layout-specific content"
+ },
"generation_note": "one short sentence"
}
-3. Every visible text in html_template must come from placeholders only:
- - text/textarea fields: {{field:key}}
- - list fields: {{list:key}}
- - controlled images, when required: {{image:key}}
-4. css_code must only target .slide-root and its descendants.
-5. Do not use external assets, remote fonts, raw image URLs, svg, canvas, script, iframe, video or img tags.
-6. The slide must fit inside a 1600x900 canvas with safe margins and no overflow.
-7. Use the supplied deck theme so every page looks like the same presentation family.
-8. Treat theme_lock as non-negotiable. Do not invent a new palette family, component language, or typography system.
-9. Keep titles within 2 lines, with title font 42-60px and body text 18-28px.
-10. Prefer grid/flex layouts over brittle absolute positioning.
-11. If visual_assets are supplied, reserve layout space and place them using {{image:key}} placeholders. Never write a raw ![]()
tag yourself.
-12. If visual_assets are empty, build a text-first slide using editable text blocks and CSS decoration only.
-13. The HTML must contain a single .slide-root root element.
-14. If reference deck slides are provided, preserve their shared component grammar, spacing rhythm, and card treatment.
-15. Never put {{field:...}}, {{list:...}}, or {{image:...}} placeholders inside HTML attributes like aria-label, title, alt, data-*, href, or style. Placeholders may only appear in element content.
+3. Never return HTML, CSS, SVG, coordinates, raw style code, or arbitrary DOM.
+4. Use only the allowed layout_type values.
+5. Keep the slide strictly editable:
+ - all visible text must live in `content`
+ - images must be referenced only through the provided visual_assets slots
+6. Use the supplied deck theme so every page looks like the same presentation family.
+7. Treat theme_lock as non-negotiable. Do not invent a new palette family, component language, or typography system.
+8. Keep titles within 2 lines, body content concise, and list lengths <= 6.
+9. Use `image_focus` only when the slide genuinely benefits from a dominant supporting visual. If no visual_assets are present, do not choose `image_focus`.
+10. `cards_2x2` must contain exactly 4 cards.
+11. `timeline` must contain 3 to 5 items.
+12. `comparison` must contain left and right sections with short bullet lists.
+13. Do not overuse one layout type across the deck. Reuse the shared theme, but vary page structure according to the content.
+
+Layout content schema:
+- cover:
+ eyebrow, title, subtitle, presenter, footer
+- section:
+ eyebrow, title, summary, quote, footer
+- bullets:
+ eyebrow, title, summary, bullets[], takeaway, footer
+- two_column:
+ eyebrow, title, summary, left_heading, left_body, left_points[], right_heading, right_body, right_points[], footer
+- cards_2x2:
+ eyebrow, title, summary, cards[{title, body} x4], footer
+- image_focus:
+ eyebrow, title, summary, bullets[], visual_caption, footer
+- comparison:
+ eyebrow, title, summary, left_title, left_points[], right_title, right_points[], footer
+- timeline:
+ eyebrow, title, summary, timeline[{label, body}], footer
""".strip()
outline_payload = {
@@ -1309,7 +1340,7 @@ def _build_messages(
user_sections.append(f"Revision request: {edit_prompt}")
user_sections.append(
- "Ensure the editable_fields fully cover all meaningful visible text shown on the slide."
+ "Return a compact structured slide. Do not emit arbitrary layout code."
)
return [
@@ -1387,13 +1418,9 @@ def _summarize_slide_for_review(self, slide: Dict[str, Any]) -> Dict[str, Any]:
"type": field_type,
}
if field_type == "list":
- entry["items"] = [
- str(item).strip()
- for item in (field.get("items") or [])
- if str(item).strip()
- ][:5]
+ entry["items"] = self._normalize_outline_points(field.get("items"), limit=5, item_limit=140)
else:
- entry["value"] = str(field.get("value") or "").strip()[:280]
+ entry["value"] = self._clean_text_content(field.get("value"), "", 280)
summarized_fields.append(entry)
visual_assets = slide.get("visual_assets") or slide.get("visualAssets") or []
@@ -1434,63 +1461,305 @@ def _normalize_slide_payload(
theme=theme,
visual_assets=visual_assets,
)
- html_template = payload.get("html_template") or payload.get("html") or ""
- css_code = payload.get("css_code") or payload.get("css") or ""
- if not isinstance(html_template, str) or not isinstance(css_code, str):
- return fallback_slide
- if len(html_template) > 16000 or len(css_code) > 20000:
+ layout_type = str(payload.get("layout_type") or payload.get("layoutType") or "").strip()
+ content = payload.get("content") or {}
+ if not isinstance(content, dict):
return fallback_slide
- if _FORBIDDEN_HTML_RE.search(html_template) or _FORBIDDEN_CSS_RE.search(css_code):
+ if layout_type not in {
+ "cover",
+ "section",
+ "bullets",
+ "two_column",
+ "cards_2x2",
+ "image_focus",
+ "comparison",
+ "timeline",
+ }:
return fallback_slide
-
- normalized_html = self._sanitize_html_template(html_template)
- normalized_css = self._sanitize_css(css_code, theme=theme)
- editable_fields = self._normalize_fields(
- payload.get("editable_fields"),
- outline_item=outline_item,
- slide_index=slide_index,
- )
- if not editable_fields:
+ if not visual_assets and layout_type == "image_focus":
return fallback_slide
- normalized_html, attribute_warnings = self._sanitize_attribute_placeholders(
- normalized_html,
- editable_fields,
- )
- if attribute_warnings:
+ try:
+ return self._build_structured_slide(
+ layout_type=layout_type,
+ content=content,
+ outline_item=outline_item,
+ slide_index=slide_index,
+ slide_count=slide_count,
+ theme=theme,
+ visual_assets=visual_assets,
+ generation_note=str(payload.get("generation_note") or "").strip(),
+ )
+ except Exception as exc: # noqa: BLE001
log.warning(
- "[Paper2PPTFrontendService] Sanitized attribute placeholders for page %s: %s",
+ "[Paper2PPTFrontendService] Failed to normalize structured slide payload for page %s: %s",
slide_index + 1,
- ", ".join(attribute_warnings),
+ exc,
)
-
- field_keys = {field["key"] for field in editable_fields}
- placeholders = set(_FIELD_PLACEHOLDER_RE.findall(normalized_html))
- image_placeholders = set(_IMAGE_PLACEHOLDER_RE.findall(normalized_html))
- asset_keys = {str(asset.get("key") or "").strip() for asset in visual_assets if str(asset.get("key") or "").strip()}
- if not placeholders:
- return fallback_slide
- if not placeholders.issubset(field_keys):
- return fallback_slide
- if image_placeholders and not image_placeholders.issubset(asset_keys):
- return fallback_slide
- if visual_assets and not image_placeholders:
return fallback_slide
- title_value = (
- self._find_field_value(editable_fields, "title")
- or outline_item.get("title")
- or f"Slide {slide_index + 1}"
+ def _field_entry(
+ self,
+ *,
+ key: str,
+ label: str,
+ field_type: str,
+ value: str = "",
+ items: Optional[List[str]] = None,
+ ) -> Dict[str, Any]:
+ return {
+ "key": key,
+ "label": label,
+ "type": field_type,
+ "value": value,
+ "items": items or [],
+ }
+
+ def _clean_text_content(self, value: Any, default: str = "", limit: int = 280) -> str:
+ text = self._extract_outline_text(value)
+ text = re.sub(r"\s+", " ", text)
+ return (text or default)[:limit]
+
+ def _extract_outline_text(self, value: Any) -> str:
+ if value is None:
+ return ""
+ if isinstance(value, str):
+ return value.strip()
+ if isinstance(value, (int, float, bool)):
+ return str(value).strip()
+ if isinstance(value, dict):
+ preferred_keys = (
+ "text",
+ "value",
+ "content",
+ "summary",
+ "title",
+ "label",
+ "body",
+ "description",
+ "reason",
+ "point",
+ )
+ for key in preferred_keys:
+ extracted = self._extract_outline_text(value.get(key))
+ if extracted:
+ return extracted
+ parts = [self._extract_outline_text(item) for item in value.values()]
+ joined = " ".join(part for part in parts if part)
+ return joined.strip()
+ if isinstance(value, list):
+ parts = [self._extract_outline_text(item) for item in value]
+ joined = " ".join(part for part in parts if part)
+ return joined.strip()
+ return str(value).strip()
+
+ def _normalize_outline_points(
+ self,
+ value: Any,
+ *,
+ limit: int = 6,
+ item_limit: int = 120,
+ ) -> List[str]:
+ normalized: List[str] = []
+
+ def _append(item: Any) -> None:
+ text = self._clean_text_content(item, "", item_limit)
+ if text and text not in normalized:
+ normalized.append(text)
+
+ if isinstance(value, list):
+ for item in value:
+ if isinstance(item, list):
+ for nested in item:
+ _append(nested)
+ else:
+ _append(item)
+ elif value is not None:
+ _append(value)
+ return normalized[:limit]
+
+ def _clean_list_content(
+ self,
+ value: Any,
+ *,
+ defaults: Optional[List[str]] = None,
+ limit: int = 6,
+ item_limit: int = 120,
+ ) -> List[str]:
+ cleaned: List[str] = []
+ if isinstance(value, list):
+ for item in value:
+ text = self._clean_text_content(item, "", item_limit)
+ if text:
+ cleaned.append(text)
+ elif isinstance(value, str) and value.strip():
+ cleaned = [self._clean_text_content(value, "", item_limit)]
+ if cleaned:
+ return cleaned[:limit]
+ return (defaults or [])[:limit]
+
+ def _build_structured_slide(
+ self,
+ *,
+ layout_type: str,
+ content: Dict[str, Any],
+ outline_item: Dict[str, Any],
+ slide_index: int,
+ slide_count: int,
+ theme: Dict[str, Any],
+ visual_assets: List[Dict[str, Any]],
+ generation_note: str,
+ ) -> Dict[str, Any]:
+ fallback_title = str(outline_item.get("title") or f"Slide {slide_index + 1}").strip()
+ section_template = str(theme.get("section_label_template") or "Slide {page_num:02d}/{slide_count:02d}")
+ try:
+ default_eyebrow = section_template.format(page_num=slide_index + 1, slide_count=slide_count)
+ except Exception: # noqa: BLE001
+ default_eyebrow = f"Slide {slide_index + 1:02d}/{slide_count:02d}"
+ key_points = self._normalize_outline_points(outline_item.get("key_points"), limit=6, item_limit=120)
+ default_summary = key_points[0] if key_points else self._clean_text_content(
+ outline_item.get("layout_description"),
+ "",
+ 280,
+ )
+ default_footer = str(theme.get("footer_text") or "Paper2Any Structured PPT").strip()
+
+ editable_fields: List[Dict[str, Any]] = []
+ layout_data: Dict[str, Any] = {"type": layout_type}
+
+ def add_text(key: str, label: str, default: str, *, field_type: str = "text", limit: int = 280) -> str:
+ value = self._clean_text_content(content.get(key), default, limit)
+ editable_fields.append(
+ self._field_entry(
+ key=key,
+ label=label,
+ field_type="textarea" if field_type == "textarea" else "text",
+ value=value,
+ )
+ )
+ return key
+
+ def add_list(key: str, label: str, default_items: List[str], *, limit: int = 6, item_limit: int = 120) -> str:
+ items = self._clean_list_content(
+ content.get(key),
+ defaults=default_items,
+ limit=limit,
+ item_limit=item_limit,
+ )
+ editable_fields.append(
+ self._field_entry(
+ key=key,
+ label=label,
+ field_type="list",
+ items=items,
+ )
+ )
+ return key
+
+ layout_data["eyebrow_key"] = add_text("eyebrow", "Eyebrow", default_eyebrow)
+ layout_data["title_key"] = add_text("title", "Title", fallback_title, limit=120)
+ layout_data["footer_key"] = add_text("footer", "Footer", default_footer, limit=80)
+
+ if layout_type == "cover":
+ layout_data["subtitle_key"] = add_text("subtitle", "Subtitle", default_summary, field_type="textarea", limit=220)
+ layout_data["presenter_key"] = add_text("presenter", "Presenter", "Presenter / Team", limit=80)
+ elif layout_type == "section":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=220)
+ layout_data["quote_key"] = add_text("quote", "Quote", key_points[1] if len(key_points) > 1 else default_summary, field_type="textarea", limit=200)
+ elif layout_type == "bullets":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=220)
+ layout_data["bullets_key"] = add_list("bullets", "Bullets", key_points[:5] or ["Add key points"])
+ layout_data["takeaway_key"] = add_text("takeaway", "Takeaway", key_points[-1] if key_points else default_summary, field_type="textarea", limit=180)
+ elif layout_type == "two_column":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=220)
+ layout_data["left_heading_key"] = add_text("left_heading", "Left Heading", "Core Idea", limit=80)
+ layout_data["left_body_key"] = add_text("left_body", "Left Body", key_points[0] if key_points else default_summary, field_type="textarea", limit=180)
+ layout_data["left_points_key"] = add_list("left_points", "Left Points", key_points[:3], limit=4)
+ layout_data["right_heading_key"] = add_text("right_heading", "Right Heading", "Implication", limit=80)
+ layout_data["right_body_key"] = add_text("right_body", "Right Body", key_points[1] if len(key_points) > 1 else default_summary, field_type="textarea", limit=180)
+ layout_data["right_points_key"] = add_list("right_points", "Right Points", key_points[2:5] or key_points[:2], limit=4)
+ elif layout_type == "cards_2x2":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=200)
+ raw_cards = content.get("cards")
+ cards = raw_cards if isinstance(raw_cards, list) else []
+ card_refs: List[Dict[str, str]] = []
+ for index in range(4):
+ item = cards[index] if index < len(cards) and isinstance(cards[index], dict) else {}
+ title_key = f"card_{index + 1}_title"
+ body_key = f"card_{index + 1}_body"
+ editable_fields.append(self._field_entry(
+ key=title_key,
+ label=f"Card {index + 1} Title",
+ field_type="text",
+ value=self._clean_text_content(item.get("title"), f"Point {index + 1}", 80),
+ ))
+ editable_fields.append(self._field_entry(
+ key=body_key,
+ label=f"Card {index + 1} Body",
+ field_type="textarea",
+ value=self._clean_text_content(
+ item.get("body"),
+ key_points[index] if index < len(key_points) else default_summary,
+ 140,
+ ),
+ ))
+ card_refs.append({"title_key": title_key, "body_key": body_key})
+ layout_data["cards"] = card_refs
+ elif layout_type == "image_focus":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=180)
+ layout_data["bullets_key"] = add_list("bullets", "Bullets", key_points[:4], limit=4)
+ layout_data["visual_caption_key"] = add_text("visual_caption", "Visual Caption", "Supporting visual", limit=90)
+ layout_data["visual_key"] = str((visual_assets[0].get("key") if visual_assets else _DEFAULT_VISUAL_KEY) or _DEFAULT_VISUAL_KEY)
+ elif layout_type == "comparison":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=180)
+ layout_data["left_title_key"] = add_text("left_title", "Left Title", "Track A", limit=80)
+ layout_data["left_points_key"] = add_list("left_points", "Left Points", key_points[:3], limit=4)
+ layout_data["right_title_key"] = add_text("right_title", "Right Title", "Track B", limit=80)
+ layout_data["right_points_key"] = add_list("right_points", "Right Points", key_points[3:6] or key_points[:3], limit=4)
+ elif layout_type == "timeline":
+ layout_data["summary_key"] = add_text("summary", "Summary", default_summary, field_type="textarea", limit=180)
+ raw_timeline = content.get("timeline")
+ timeline_items = raw_timeline if isinstance(raw_timeline, list) else []
+ timeline_refs: List[Dict[str, str]] = []
+ count = max(3, min(5, len(timeline_items) or 3))
+ for index in range(count):
+ item = timeline_items[index] if index < len(timeline_items) and isinstance(timeline_items[index], dict) else {}
+ label_key = f"timeline_{index + 1}_label"
+ body_key = f"timeline_{index + 1}_body"
+ editable_fields.append(self._field_entry(
+ key=label_key,
+ label=f"Timeline {index + 1} Label",
+ field_type="text",
+ value=self._clean_text_content(item.get("label"), f"Phase {index + 1}", 60),
+ ))
+ editable_fields.append(self._field_entry(
+ key=body_key,
+ label=f"Timeline {index + 1} Body",
+ field_type="textarea",
+ value=self._clean_text_content(
+ item.get("body"),
+ key_points[index] if index < len(key_points) else default_summary,
+ 120,
+ ),
+ ))
+ timeline_refs.append({"label_key": label_key, "body_key": body_key})
+ layout_data["timeline"] = timeline_refs
+ else:
+ raise ValueError(f"unsupported layout_type: {layout_type}")
+
+ title_value = next(
+ (field.get("value") for field in editable_fields if field.get("key") == "title"),
+ fallback_title,
)
return {
- "slide_id": str(payload.get("slide_id") or slide_index + 1),
+ "slide_id": str(slide_index + 1),
"page_num": slide_index + 1,
- "title": str(payload.get("title") or title_value),
- "html_template": normalized_html,
- "css_code": normalized_css,
+ "title": str(title_value or fallback_title),
+ "layout_type": layout_type,
+ "layout_data": layout_data,
"editable_fields": editable_fields,
"visual_assets": visual_assets,
- "generation_note": str(payload.get("generation_note") or ""),
+ "generation_note": generation_note or "Structured slide generated",
"status": "done",
}
@@ -1501,11 +1770,7 @@ def _normalize_review_payload(
slide: Dict[str, Any],
local_layout_issues: List[str],
) -> Dict[str, Any]:
- issues = [
- str(item).strip()
- for item in (payload.get("issues") or [])
- if str(item).strip()
- ]
+ issues = self._normalize_outline_points(payload.get("issues"), limit=12, item_limit=220)
combined_issues: List[str] = []
for issue in [*local_layout_issues, *issues]:
if issue and issue not in combined_issues:
@@ -1542,11 +1807,7 @@ def _normalize_fields(
) -> List[Dict[str, Any]]:
normalized: List[Dict[str, Any]] = []
seen_keys: set[str] = set()
- outline_points = [
- str(item).strip()
- for item in (outline_item.get("key_points") or [])
- if str(item).strip()
- ]
+ outline_points = self._normalize_outline_points(outline_item.get("key_points"), limit=6, item_limit=120)
if isinstance(raw_fields, list):
for raw_field in raw_fields:
@@ -1560,11 +1821,7 @@ def _normalize_fields(
field_type = "text"
label = str(raw_field.get("label") or key.replace("_", " ").title())
if field_type == "list":
- items = [
- str(item).strip()
- for item in (raw_field.get("items") or [])
- if str(item).strip()
- ]
+ items = self._normalize_outline_points(raw_field.get("items"), limit=8, item_limit=140)
if not items:
items = outline_points[:4]
normalized.append(
@@ -1577,7 +1834,7 @@ def _normalize_fields(
}
)
else:
- value = str(raw_field.get("value") or "").strip()
+ value = self._clean_text_content(raw_field.get("value"), "", 280)
normalized.append(
{
"key": key,
@@ -1595,7 +1852,7 @@ def _normalize_fields(
"key": "title",
"label": "Title",
"type": "text",
- "value": str(outline_item.get("title") or f"Slide {slide_index + 1}"),
+ "value": self._clean_text_content(outline_item.get("title"), f"Slide {slide_index + 1}", 220),
"items": [],
}
)
@@ -1605,9 +1862,11 @@ def _normalize_fields(
"key": "summary",
"label": "Summary",
"type": "textarea",
- "value": str(
- (outline_points[0] if outline_points else outline_item.get("layout_description") or "")
- ).strip(),
+ "value": self._clean_text_content(
+ outline_points[0] if outline_points else outline_item.get("layout_description"),
+ "",
+ 280,
+ ),
"items": [],
}
)
@@ -1624,6 +1883,7 @@ def _normalize_fields(
return normalized
def _build_fallback_theme(self, *, language: str, style: str) -> Dict[str, Any]:
+ style_family = self._infer_style_family(style)
footer_text = "Paper2Any Frontend PPT"
section_label_template = (
"第 {page_num:02d}/{slide_count:02d} 页"
@@ -1637,9 +1897,11 @@ def _build_fallback_theme(self, *, language: str, style: str) -> Dict[str, Any]:
)
)
palette = self._resolve_palette_from_style(style)
+ family_rules = self._build_family_rules(style_family)
return {
"theme_name": "scholarly_signal",
"visual_mood": visual_mood,
+ "style_family": style_family,
"palette": palette,
"typography": {
"title_font_stack": 'Georgia, "Times New Roman", serif',
@@ -1649,40 +1911,130 @@ def _build_fallback_theme(self, *, language: str, style: str) -> Dict[str, Any]:
"summary_size": 26,
"body_size": 24,
},
- "layout_rules": [
- "Keep 72px+ safe margins around major content.",
- "Prefer one dominant text area plus one supporting card or metrics block.",
- "Avoid more than two visual columns in a single slide.",
- "Reserve a quiet footer area for page identity or takeaway.",
- ],
- "component_rules": [
- "Use rounded cards with subtle borders and a restrained glow.",
- "Use one accent color only for emphasis, not for large fills.",
- "Keep text hierarchy clear with title, summary, and supporting bullets.",
- ],
+ "layout_rules": family_rules["layout_rules"],
+ "component_rules": family_rules["component_rules"],
"theme_lock": {
"must_keep": [
"Use only the deck palette colors for fills, borders, and emphasis.",
"Keep the same serif title style and sans body style across the deck.",
- "Keep rounded translucent cards and a quiet footer treatment.",
- ],
- "preferred_layout_patterns": [
- "hero_with_side_card",
- "split_insight_grid",
- "stacked_cards",
- "timeline_overview",
+ family_rules["must_keep"],
],
- "component_signature": "Rounded refined cards, restrained accent usage, thin borders, and quiet academic spacing.",
+ "preferred_layout_patterns": family_rules["preferred_layout_patterns"],
+ "component_signature": family_rules["component_signature"],
"avoid": [
"Do not introduce unrelated bright color families.",
"Do not use more than two main columns.",
- "Do not use oversized billboard titles or poster-like full-bleed blocks.",
+ family_rules["avoid"],
],
},
"footer_text": footer_text,
"section_label_template": section_label_template,
}
+ def _infer_style_family(self, style: str) -> str:
+ style_text = (style or "").strip().lower()
+ if any(keyword in style_text for keyword in ("academic", "report", "paper", "research", "严谨", "学术", "报告")):
+ return "academic"
+ if any(keyword in style_text for keyword in ("business", "brand", "corporate", "executive", "商务", "商业", "品牌")):
+ return "business"
+ if any(keyword in style_text for keyword in ("creative", "illustration", "warm", "friendly", "playful", "soft", "创意", "插画", "柔和")):
+ return "creative"
+ return "modern"
+
+ def _build_family_rules(self, style_family: str) -> Dict[str, Any]:
+ family = (style_family or "modern").strip().lower()
+ if family == "academic":
+ return {
+ "layout_rules": [
+ "Keep generous white or paper-like breathing room and stable reading rhythm.",
+ "Prefer section, bullets, two-column, and comparison layouts over showy hero frames.",
+ "Use image_focus only for genuinely visual pages.",
+ "Reserve a quiet footer for provenance or page identity.",
+ ],
+ "component_rules": [
+ "Use restrained panels, subtle dividers, and report-like hierarchy.",
+ "Keep decoration secondary to text structure and evidence density.",
+ "Avoid billboard marketing blocks or exaggerated hero cards.",
+ ],
+ "preferred_layout_patterns": [
+ "section_break",
+ "split_report_grid",
+ "comparison_columns",
+ "timeline_overview",
+ ],
+ "component_signature": "Refined report-style panels, paper-like spacing, and calm academic hierarchy.",
+ "must_keep": "Keep the visual language rigorous, airy, and report-like rather than glossy.",
+ "avoid": "Do not use neon glow, oversized promo badges, or playful sticker motifs.",
+ }
+ if family == "business":
+ return {
+ "layout_rules": [
+ "Favor crisp comparison, KPI-card, and executive-summary patterns.",
+ "Use strong alignment and clear block grouping with moderate density.",
+ "Keep image_focus to showcase slides only.",
+ "Prefer horizontal momentum and strong title anchoring.",
+ ],
+ "component_rules": [
+ "Use sharp, decisive cards with stronger contrast and cleaner edges.",
+ "Accent color should be used sparingly for strategic emphasis.",
+ "Make hierarchy feel presentation-room ready rather than article-like.",
+ ],
+ "preferred_layout_patterns": [
+ "executive_hero",
+ "split_insight_grid",
+ "kpi_cards",
+ "decision_comparison",
+ ],
+ "component_signature": "Crisp executive cards, strong title anchors, and controlled business contrast.",
+ "must_keep": "Keep the deck polished, decisive, and boardroom-oriented.",
+ "avoid": "Do not use whimsical illustration accents or soft scrapbook styling.",
+ }
+ if family == "creative":
+ return {
+ "layout_rules": [
+ "Allow more asymmetry, softer framing, and stronger hero moments.",
+ "Alternate between image_focus, section, cards, and timeline layouts to keep the deck lively.",
+ "Use comparison and two-column layouts only when the content clearly calls for them.",
+ "Let accent shapes support the narrative without overpowering text.",
+ ],
+ "component_rules": [
+ "Use soft panels, expressive color accents, and warmer visual transitions.",
+ "Preserve readability, but allow more character in backgrounds and separators.",
+ "Favor friendly, presentation-forward composition over report density.",
+ ],
+ "preferred_layout_patterns": [
+ "hero_spotlight",
+ "soft_cards",
+ "story_timeline",
+ "image_caption_feature",
+ ],
+ "component_signature": "Warm expressive panels, softer geometry, and more atmospheric deck motion.",
+ "must_keep": "Keep the deck warm, expressive, and visibly more playful than academic or business presets.",
+ "avoid": "Do not collapse the deck back into a uniform report grid on every page.",
+ }
+ return {
+ "layout_rules": [
+ "Keep 72px+ safe margins around major content.",
+ "Prefer one dominant text area plus one supporting card or metrics block.",
+ "Avoid more than two visual columns in a single slide.",
+ "Reserve a quiet footer area for page identity or takeaway.",
+ ],
+ "component_rules": [
+ "Use refined rounded cards with controlled glow and layered depth.",
+ "Use one accent color only for emphasis, not for large fills.",
+ "Keep text hierarchy clear with title, summary, and supporting bullets.",
+ ],
+ "preferred_layout_patterns": [
+ "hero_with_side_card",
+ "split_insight_grid",
+ "stacked_cards",
+ "timeline_overview",
+ ],
+ "component_signature": "Modern layered cards, restrained glow, and polished presentation spacing.",
+ "must_keep": "Keep the deck sleek, layered, and contemporary without drifting into plain report style.",
+ "avoid": "Do not flatten everything into plain white report blocks unless the prompt explicitly asks for that.",
+ }
+
def _resolve_palette_from_style(self, style: str) -> Dict[str, str]:
style_text = (style or "").strip().lower()
@@ -1802,27 +2154,26 @@ def _clean_int(value: Any, default: int, min_value: int, max_value: int) -> int:
return default
return max(min_value, min(max_value, parsed))
+ def _clean_style_family(value: Any, default: str) -> str:
+ candidate = str(value or "").strip().lower()
+ if candidate in {"modern", "business", "academic", "creative"}:
+ return candidate
+ return default
+
def _clean_list(value: Any, defaults: List[str], limit: int = 6) -> List[str]:
if isinstance(value, list):
- cleaned = [str(item).strip() for item in value if str(item).strip()]
+ cleaned = self._normalize_outline_points(value, limit=limit, item_limit=140)
if cleaned:
return cleaned[:limit]
return defaults[:limit]
- layout_rules = [
- str(item).strip()
- for item in (payload.get("layout_rules") or [])
- if str(item).strip()
- ][:6]
- component_rules = [
- str(item).strip()
- for item in (payload.get("component_rules") or [])
- if str(item).strip()
- ][:6]
+ layout_rules = self._normalize_outline_points(payload.get("layout_rules"), limit=6, item_limit=180)
+ component_rules = self._normalize_outline_points(payload.get("component_rules"), limit=6, item_limit=180)
return {
"theme_name": _clean_text(payload.get("theme_name"), fallback["theme_name"]),
"visual_mood": _clean_text(payload.get("visual_mood"), fallback["visual_mood"]),
+ "style_family": _clean_style_family(payload.get("style_family"), fallback["style_family"]),
"palette": {
"bg": _clean_color(palette_raw.get("bg"), fallback["palette"]["bg"]),
"panel": _clean_color(palette_raw.get("panel"), fallback["palette"]["panel"]),
@@ -1898,25 +2249,25 @@ def _build_theme_lock(self, theme: Dict[str, Any]) -> Dict[str, Any]:
theme_lock = theme.get("theme_lock")
if isinstance(theme_lock, dict):
return {
- "must_keep": [
- str(item).strip()
- for item in (theme_lock.get("must_keep") or [])
- if str(item).strip()
- ] or fallback["theme_lock"]["must_keep"],
- "preferred_layout_patterns": [
- str(item).strip()
- for item in (theme_lock.get("preferred_layout_patterns") or [])
- if str(item).strip()
- ] or fallback["theme_lock"]["preferred_layout_patterns"],
+ "must_keep": self._normalize_outline_points(
+ theme_lock.get("must_keep"),
+ limit=8,
+ item_limit=180,
+ ) or fallback["theme_lock"]["must_keep"],
+ "preferred_layout_patterns": self._normalize_outline_points(
+ theme_lock.get("preferred_layout_patterns"),
+ limit=8,
+ item_limit=180,
+ ) or fallback["theme_lock"]["preferred_layout_patterns"],
"component_signature": str(
theme_lock.get("component_signature")
or fallback["theme_lock"]["component_signature"]
).strip(),
- "avoid": [
- str(item).strip()
- for item in (theme_lock.get("avoid") or [])
- if str(item).strip()
- ] or fallback["theme_lock"]["avoid"],
+ "avoid": self._normalize_outline_points(
+ theme_lock.get("avoid"),
+ limit=8,
+ item_limit=180,
+ ) or fallback["theme_lock"]["avoid"],
}
return fallback["theme_lock"]
@@ -1927,6 +2278,7 @@ def _build_deck_identity_summary(self, theme: Dict[str, Any]) -> Dict[str, Any]:
return {
"theme_name": str(theme.get("theme_name") or "deck_theme").strip(),
"visual_mood": str(theme.get("visual_mood") or "").strip(),
+ "style_family": str(theme.get("style_family") or "modern").strip(),
"palette_anchor": {
"bg": str(palette.get("bg") or "").strip(),
"primary": str(palette.get("primary") or "").strip(),
@@ -2146,20 +2498,21 @@ def _load_reference_slides(
return [self._summarize_reference_slide(slide) for slide in references]
def _summarize_reference_slide(self, slide: Dict[str, Any]) -> Dict[str, Any]:
- html_template = str(slide.get("html_template") or "")
- css_code = str(slide.get("css_code") or "")
editable_fields = slide.get("editable_fields") or []
return {
"page_num": int(slide.get("page_num") or 0),
"title": str(slide.get("title") or "").strip(),
+ "layout_type": str(slide.get("layout_type") or slide.get("layoutType") or "").strip(),
"field_keys": [
str(field.get("key") or "").strip()
for field in editable_fields
if isinstance(field, dict) and str(field.get("key") or "").strip()
][:10],
- "html_outline": self._extract_html_outline(html_template),
- "component_classes": self._extract_component_classes(html_template, css_code),
- "css_selectors": self._extract_css_selectors(css_code),
+ "visual_asset_keys": [
+ str(asset.get("key") or "").strip()
+ for asset in (slide.get("visual_assets") or [])
+ if isinstance(asset, dict) and str(asset.get("key") or "").strip()
+ ][:4],
}
def _extract_html_outline(self, html_template: str, limit: int = 12) -> List[str]:
@@ -2216,6 +2569,42 @@ def _extract_css_selectors(self, css_code: str, limit: int = 8) -> List[str]:
break
return cleaned
+ def _choose_fallback_layout_type(
+ self,
+ *,
+ outline_item: Dict[str, Any],
+ slide_index: int,
+ theme: Dict[str, Any],
+ visual_assets: Sequence[Dict[str, Any]],
+ ) -> str:
+ style_family = str(theme.get("style_family") or "modern").strip().lower()
+ layout_hint = str(outline_item.get("layout_description") or "").lower()
+ title = str(outline_item.get("title") or "").lower()
+ key_points = self._normalize_outline_points(outline_item.get("key_points"), limit=6, item_limit=120)
+ bullet_count = len(key_points)
+
+ if slide_index == 0:
+ return "cover"
+ if any(keyword in title for keyword in ("overview", "agenda", "outline", "introduction", "background", "summary")):
+ return "section"
+ if any(keyword in layout_hint for keyword in ("compare", "contrast", "trade-off", "versus", "vs", "对比", "比较")):
+ return "comparison"
+ if any(keyword in layout_hint for keyword in ("timeline", "process", "workflow", "loop", "pipeline", "流程", "时间线")):
+ return "timeline"
+ if any(keyword in layout_hint for keyword in ("card", "grid", "domain", "application", "industry", "module", "模块", "领域")):
+ return "cards_2x2"
+ if bullet_count >= 5 and style_family in {"business", "modern"}:
+ return "two_column"
+ if visual_assets and style_family in {"creative", "modern"} and slide_index % 3 == 1:
+ return "image_focus"
+ if bullet_count >= 4 and style_family == "academic":
+ return "bullets"
+ if style_family == "business":
+ return "two_column"
+ if style_family == "creative":
+ return "cards_2x2"
+ return "bullets"
+
def _build_fallback_slide(
self,
*,
@@ -2225,315 +2614,46 @@ def _build_fallback_slide(
theme: Dict[str, Any],
visual_assets: Optional[List[Dict[str, Any]]] = None,
) -> Dict[str, Any]:
- palette = theme.get("palette") or self._build_fallback_theme(language="zh", style="")["palette"]
- typography = theme.get("typography") or {}
visual_assets = (visual_assets or [])[:_MAX_INLINE_VISUAL_ASSETS]
- has_visual = bool(visual_assets)
- has_multi_visual = len(visual_assets) > 1
- key_points = [
- str(item).strip()
- for item in (outline_item.get("key_points") or [])
- if str(item).strip()
- ][:4]
- summary = key_points[0] if key_points else str(outline_item.get("layout_description") or "").strip()
+ key_points = self._normalize_outline_points(outline_item.get("key_points"), limit=4, item_limit=120)
+ summary = key_points[0] if key_points else self._clean_text_content(
+ outline_item.get("layout_description"),
+ "",
+ 280,
+ )
takeaway = key_points[-1] if key_points else "Refine the narrative in the editor"
section_template = str(theme.get("section_label_template") or "Slide {page_num:02d}/{slide_count:02d}")
try:
eyebrow = section_template.format(page_num=slide_index + 1, slide_count=slide_count)
except Exception: # noqa: BLE001
eyebrow = f"Slide {slide_index + 1:02d}/{slide_count:02d}"
-
- visual_markup = ""
- if has_visual:
- visual_markup = "\n".join(
- f' {{{{image:{asset.get("key") or self._build_visual_asset_key(asset_index)}}}}}
'
- for asset_index, asset in enumerate(visual_assets)
- )
-
- html_template = """
-
-
-
-
-
-
{{field:eyebrow}}
-
{{field:title}}
-
{{field:summary}}
- """ + (
- """
-
-"""
- if has_visual
- else ""
- ) + """
-
- """ + (
- """
-
-""" + visual_markup + """
-
-"""
- if has_visual
- else """
-
-
{{field:points_label}}
-
-
-"""
- ) + """
-
-
-
-
Free 模式下由后端统一选择 DrawIO 转换使用的视觉模型。
+ )}
diff --git a/frontend-workflow/src/components/Image2PptPage.tsx b/frontend-workflow/src/components/Image2PptPage.tsx
index 1230b844..47762405 100644
--- a/frontend-workflow/src/components/Image2PptPage.tsx
+++ b/frontend-workflow/src/components/Image2PptPage.tsx
@@ -221,7 +221,9 @@ const Image2PptPage = () => {
try {
setIsValidating(true);
setError(null);
- await verifyLlmConnection(llmApiUrl, apiKey, import.meta.env.VITE_DEFAULT_LLM_MODEL || 'deepseek-v3.2');
+ if (userApiConfigRequired) {
+ await verifyLlmConnection(llmApiUrl, apiKey, import.meta.env.VITE_DEFAULT_LLM_MODEL || 'deepseek-v3.2');
+ }
setIsValidating(false);
} catch (err) {
setIsValidating(false);
@@ -268,8 +270,8 @@ const Image2PptPage = () => {
if (userApiConfigRequired) {
formData.append('chat_api_url', llmApiUrl.trim());
formData.append('api_key', apiKey.trim());
+ formData.append('gen_fig_model', genFigModel);
}
- formData.append('gen_fig_model', genFigModel);
} else {
formData.append('use_ai_edit', 'false');
}
@@ -553,7 +555,8 @@ const Image2PptPage = () => {
+ {!userApiConfigRequired ? (
+ Free 模式下由后端统一选择思维导图模型。
+ ) : null}
diff --git a/frontend-workflow/src/components/Pdf2PptPage.tsx b/frontend-workflow/src/components/Pdf2PptPage.tsx
index fe819f0a..abec48b5 100644
--- a/frontend-workflow/src/components/Pdf2PptPage.tsx
+++ b/frontend-workflow/src/components/Pdf2PptPage.tsx
@@ -222,7 +222,9 @@ const Pdf2PptPage = () => {
try {
setIsValidating(true);
setError(null);
- await verifyLlmConnection(llmApiUrl, apiKey, import.meta.env.VITE_DEFAULT_LLM_MODEL || 'deepseek-v3.2');
+ if (userApiConfigRequired) {
+ await verifyLlmConnection(llmApiUrl, apiKey, import.meta.env.VITE_DEFAULT_LLM_MODEL || 'deepseek-v3.2');
+ }
setIsValidating(false);
} catch (err) {
setIsValidating(false);
@@ -266,11 +268,11 @@ const Pdf2PptPage = () => {
if (useAiEdit) {
formData.append('use_ai_edit', 'true');
- if (userApiConfigRequired) {
- formData.append('chat_api_url', llmApiUrl.trim());
- formData.append('api_key', apiKey.trim());
- }
- formData.append('gen_fig_model', genFigModel);
+ if (userApiConfigRequired) {
+ formData.append('chat_api_url', llmApiUrl.trim());
+ formData.append('api_key', apiKey.trim());
+ formData.append('gen_fig_model', genFigModel);
+ }
} else {
formData.append('use_ai_edit', 'false');
}
@@ -579,7 +581,8 @@ const Pdf2PptPage = () => {
Free 模式下由后端统一选择文本模型。
+ )}
@@ -1706,7 +1696,7 @@ const Ppt2PolishPage = () => {
-
-
-
-
+ {userApiConfigRequired ? (
+ <>
+
+
+
+
-
-
- setApiKey(e.target.value)}
- placeholder="sk-..."
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-emerald-500 font-mono"
- />
-
+
+
+ setApiKey(e.target.value)}
+ placeholder="sk-..."
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-emerald-500 font-mono"
+ />
+
-
-
- setModel(e.target.value)}
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-emerald-500"
- />
-
+
+
+ setModel(e.target.value)}
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-emerald-500"
+ />
+
+
+ ) : (
+
+ Free 模式下由后端统一选择深度研究模型、搜索凭证与接口配置。
+
+ )}
-
-
- setSearchApiKey(e.target.value)}
- placeholder="search_api_key"
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-emerald-500 font-mono"
- />
-
- {searchProvider === 'google_cse' && (
+ {userApiConfigRequired && (
+
+
+ setSearchApiKey(e.target.value)}
+ placeholder="search_api_key"
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-emerald-500 font-mono"
+ />
+
+ )}
+ {userApiConfigRequired && searchProvider === 'google_cse' && (
{
const { user } = useAuthStore();
+ const { userApiConfigRequired } = useRuntimeBilling();
const [mindmapGenerating, setMindmapGenerating] = useState(false);
const [generatedMermaidCode, setGeneratedMermaidCode] = useState('');
const [showPreview, setShowPreview] = useState(false);
@@ -49,7 +51,7 @@ export const MindMapTool = ({ files = [], selectedIds, onGenerateSuccess }: Mind
return;
}
- if (!mindmapParams.api_key) {
+ if (userApiConfigRequired && !mindmapParams.api_key) {
alert('请输入 API Key');
return;
}
@@ -75,12 +77,16 @@ export const MindMapTool = ({ files = [], selectedIds, onGenerateSuccess }: Mind
file_paths: filePaths,
user_id: user.id,
email: user.email,
- api_url: mindmapParams.api_url,
- api_key: mindmapParams.api_key,
- model: mindmapParams.model,
mindmap_style: mindmapParams.mindmap_style,
max_depth: mindmapParams.max_depth,
- language: mindmapParams.language
+ language: mindmapParams.language,
+ ...(userApiConfigRequired
+ ? {
+ api_url: mindmapParams.api_url,
+ api_key: mindmapParams.api_key,
+ model: mindmapParams.model,
+ }
+ : {})
})
});
@@ -148,29 +154,37 @@ export const MindMapTool = ({ files = [], selectedIds, onGenerateSuccess }: Mind
{/* Configuration */}
-
-
- setMindmapParams({...mindmapParams, api_key: e.target.value})}
- placeholder="sk-..."
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-cyan-500 font-mono"
- />
-
+ {userApiConfigRequired ? (
+ <>
+
+
+ setMindmapParams({...mindmapParams, api_key: e.target.value})}
+ placeholder="sk-..."
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-cyan-500 font-mono"
+ />
+
-
-
-
-
+
+
+
+
+
+ ) : (
+
+ Free 模式下由后端统一选择思维导图模型与接口配置。
+
+ )}
@@ -178,7 +192,8 @@ export const MindMapTool = ({ files = [], selectedIds, onGenerateSuccess }: Mind
diff --git a/frontend-workflow/src/components/knowledge-base/tools/PodcastTool.tsx b/frontend-workflow/src/components/knowledge-base/tools/PodcastTool.tsx
index dfdb9f18..718d153c 100644
--- a/frontend-workflow/src/components/knowledge-base/tools/PodcastTool.tsx
+++ b/frontend-workflow/src/components/knowledge-base/tools/PodcastTool.tsx
@@ -5,6 +5,7 @@ import { KnowledgeFile } from '../types';
import { getApiSettings } from '../../../services/apiSettingsService';
import { backendFetch } from '../../../services/backendClient';
import { useAuthStore } from '../../../stores/authStore';
+import { useRuntimeBilling } from '../../../hooks/useRuntimeBilling';
const COSYVOICE_TTS_MODELS = ['cosyvoice-v3-flash', 'cosyvoice-v3-plus', 'cosyvoice-v2'];
const COSYVOICE_VOICE_LABEL = '默认';
@@ -75,6 +76,7 @@ interface PodcastToolProps {
export const PodcastTool = ({ files = [], selectedIds, onGenerateSuccess }: PodcastToolProps) => {
const { user } = useAuthStore();
+ const { userApiConfigRequired } = useRuntimeBilling();
const [podcastGenerating, setPodcastGenerating] = useState(false);
const [podcastParams, setPodcastParams] = useState({
api_key: '',
@@ -126,7 +128,7 @@ export const PodcastTool = ({ files = [], selectedIds, onGenerateSuccess }: Podc
return;
}
- if (!podcastParams.api_key) {
+ if (userApiConfigRequired && !podcastParams.api_key) {
alert('请输入 API Key');
return;
}
@@ -151,15 +153,19 @@ export const PodcastTool = ({ files = [], selectedIds, onGenerateSuccess }: Podc
file_paths: filePaths,
user_id: user.id,
email: user.email,
- api_url: podcastParams.api_url,
- api_key: podcastParams.api_key,
- model: podcastParams.model,
- tts_model: podcastParams.tts_model,
voice_name: podcastParams.voice_name,
voice_name_b: podcastParams.voice_name_b,
podcast_mode: podcastParams.podcast_mode,
podcast_length: podcastParams.podcast_length,
- language: podcastParams.language
+ language: podcastParams.language,
+ ...(userApiConfigRequired
+ ? {
+ api_url: podcastParams.api_url,
+ api_key: podcastParams.api_key,
+ model: podcastParams.model,
+ tts_model: podcastParams.tts_model,
+ }
+ : {})
})
});
@@ -223,29 +229,37 @@ export const PodcastTool = ({ files = [], selectedIds, onGenerateSuccess }: Podc
{/* Configuration */}
-
-
- setPodcastParams({...podcastParams, api_key: e.target.value})}
- placeholder="sk-..."
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-green-500 font-mono"
- />
-
-
-
-
-
-
+ {userApiConfigRequired ? (
+ <>
+
+
+ setPodcastParams({...podcastParams, api_key: e.target.value})}
+ placeholder="sk-..."
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-green-500 font-mono"
+ />
+
+
+
+
+
+
+
+ ) : (
+
+ Free 模式下由后端统一选择播客脚本模型、TTS 模型和接口配置。
+
+ )}
@@ -253,7 +267,8 @@ export const PodcastTool = ({ files = [], selectedIds, onGenerateSuccess }: Podc
@@ -285,7 +301,8 @@ export const PodcastTool = ({ files = [], selectedIds, onGenerateSuccess }: Podc
voice_name_b: normalizeVoice(prev.voice_name_b, nextOptions)
}));
}}
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-green-500"
+ disabled={!userApiConfigRequired}
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-green-500 disabled:opacity-50"
>
{COSYVOICE_TTS_MODELS.map(m => (
diff --git a/frontend-workflow/src/components/knowledge-base/tools/PptTool.tsx b/frontend-workflow/src/components/knowledge-base/tools/PptTool.tsx
index 2a3d64af..3df4bd81 100644
--- a/frontend-workflow/src/components/knowledge-base/tools/PptTool.tsx
+++ b/frontend-workflow/src/components/knowledge-base/tools/PptTool.tsx
@@ -5,6 +5,7 @@ import { KnowledgeFile } from '../types';
import { getApiSettings } from '../../../services/apiSettingsService';
import { backendFetch } from '../../../services/backendClient';
import { useAuthStore } from '../../../stores/authStore';
+import { useRuntimeBilling } from '../../../hooks/useRuntimeBilling';
interface PptToolProps {
files: KnowledgeFile[];
@@ -14,6 +15,7 @@ interface PptToolProps {
export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps) => {
const { user } = useAuthStore();
+ const { userApiConfigRequired } = useRuntimeBilling();
const [pptGenerating, setPptGenerating] = useState(false);
const [pptParams, setPptParams] = useState({
api_key: '',
@@ -62,7 +64,7 @@ export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps)
return;
}
- if (!pptParams.api_key) {
+ if (userApiConfigRequired && !pptParams.api_key) {
alert('请输入 API Key');
return;
}
@@ -102,13 +104,17 @@ export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps)
need_embedding: needEmbedding,
user_id: user.id,
email: user.email,
- api_url: pptParams.api_url,
- api_key: pptParams.api_key,
style: getStyleDescription(pptParams.style_preset),
language: pptParams.language,
page_count: pptParams.page_count,
- model: pptParams.model,
- gen_fig_model: pptParams.gen_fig_model
+ ...(userApiConfigRequired
+ ? {
+ api_url: pptParams.api_url,
+ api_key: pptParams.api_key,
+ model: pptParams.model,
+ gen_fig_model: pptParams.gen_fig_model,
+ }
+ : {})
})
});
@@ -166,16 +172,43 @@ export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps)
{/* Configuration */}
-
-
- setPptParams({...pptParams, api_key: e.target.value})}
- placeholder="sk-..."
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-purple-500 font-mono"
- />
-
+ {userApiConfigRequired ? (
+ <>
+
+
+ setPptParams({...pptParams, api_key: e.target.value})}
+ placeholder="sk-..."
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-purple-500 font-mono"
+ />
+
+
+
+
+
+
+ ) : (
+
+ Free 模式下由后端统一选择 PPT 生成使用的文本与生图模型。
+
+ )}
@@ -198,27 +231,6 @@ export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps)
需要向量入库并基于检索生成大纲
-
-
-
-
-
@@ -226,7 +238,8 @@ export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps)
@@ -246,7 +260,7 @@ export const PptTool = ({ files, selectedIds, onGenerateSuccess }: PptToolProps)
-
-
-
-
+ {userApiConfigRequired ? (
+ <>
+
+
+
+
-
-
- setApiKey(e.target.value)}
- placeholder="sk-..."
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-fuchsia-500 font-mono"
- />
-
+
+
+ setApiKey(e.target.value)}
+ placeholder="sk-..."
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-fuchsia-500 font-mono"
+ />
+
-
-
- setModel(e.target.value)}
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-fuchsia-500"
- />
-
+
+
+ setModel(e.target.value)}
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-fuchsia-500"
+ />
+
+
+ ) : (
+
+ Free 模式下由后端统一选择报告生成模型与接口配置。
+
+ )}
@@ -236,29 +243,37 @@ export const SearchTool = ({ files = [], selectedIds = new Set(), knowledgeBases
-
-
-
-
-
-
-
- setApiKey(e.target.value)}
- placeholder="sk-..."
- className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-blue-500 font-mono"
- />
-
+ {userApiConfigRequired ? (
+ <>
+
+
+
+
+
+
+
+ setApiKey(e.target.value)}
+ placeholder="sk-..."
+ className="w-full bg-black/40 border border-white/10 rounded-lg px-3 py-2.5 text-sm text-gray-200 outline-none focus:border-blue-500 font-mono"
+ />
+
+
+ ) : (
+
+ Free 模式下由后端统一选择向量检索使用的嵌入模型与接口配置。
+
+ )}
diff --git a/frontend-workflow/src/components/paper2drawio/index.tsx b/frontend-workflow/src/components/paper2drawio/index.tsx
index 0d20914f..ef3d75f0 100644
--- a/frontend-workflow/src/components/paper2drawio/index.tsx
+++ b/frontend-workflow/src/components/paper2drawio/index.tsx
@@ -18,6 +18,7 @@ import Banner from './Banner';
import QRCodeTooltip from '../QRCodeTooltip';
import ManagedApiNotice from '../ManagedApiNotice';
import { useRuntimeBilling } from '../../hooks/useRuntimeBilling';
+import { appendManagedApiConfig, appendManagedModel } from '../../utils/runtimeBillingForm';
const DRAWIO_ORIGINS = new Set(['https://embed.diagrams.net', 'https://app.diagrams.net']);
const STORAGE_KEY = 'paper2drawio_settings';
@@ -297,17 +298,19 @@ export default function Paper2DrawioPage({
const handleGenerate = useCallback(async () => {
if (!textContent && !file) return;
- // Step 0: Verify LLM Connection first
- try {
- setIsValidating(true);
- setError(null);
- await verifyLlmConnection(apiUrl, apiKey, model);
- setIsValidating(false);
- } catch (err) {
- setIsValidating(false);
- const errorMsg = err instanceof Error ? err.message : '验证 LLM 连接失败';
- setError(errorMsg);
- return;
+ if (userApiConfigRequired) {
+ // Step 0: Verify LLM Connection first
+ try {
+ setIsValidating(true);
+ setError(null);
+ await verifyLlmConnection(apiUrl, apiKey, model);
+ setIsValidating(false);
+ } catch (err) {
+ setIsValidating(false);
+ const errorMsg = err instanceof Error ? err.message : '验证 LLM 连接失败';
+ setError(errorMsg);
+ return;
+ }
}
setIsLoading(true);
@@ -315,11 +318,8 @@ export default function Paper2DrawioPage({
try {
if (generationMode === 'paper2drawio') {
const formData = new FormData();
- formData.append('img_gen_model_name', p2dImageModel);
- if (userApiConfigRequired) {
- formData.append('chat_api_url', apiUrl);
- formData.append('api_key', apiKey);
- }
+ appendManagedModel(formData, userApiConfigRequired, 'img_gen_model_name', p2dImageModel);
+ appendManagedApiConfig(formData, userApiConfigRequired, apiUrl, apiKey);
formData.append('input_type', uploadMode);
formData.append('graph_type', 'model_arch');
formData.append('style', p2dStyle);
@@ -368,12 +368,9 @@ export default function Paper2DrawioPage({
}
const formData = new FormData();
- if (userApiConfigRequired) {
- formData.append('chat_api_url', apiUrl);
- formData.append('api_key', apiKey);
- }
+ appendManagedApiConfig(formData, userApiConfigRequired, apiUrl, apiKey);
const modelToSend = enableModelRace ? withModelOptions(PAPER2DRAWIO_MODELS, model).join(',') : model;
- formData.append('model', modelToSend);
+ appendManagedModel(formData, userApiConfigRequired, 'model', modelToSend);
formData.append('input_type', uploadMode === 'file' ? 'PDF' : 'TEXT');
formData.append('diagram_type', diagramType);
formData.append('diagram_style', diagramStyle);
@@ -927,7 +924,8 @@ export default function Paper2DrawioPage({
- {modelOptions.length > 1 && (
+ {!userApiConfigRequired && (
+
Free 模式下由后端统一选择 DrawIO 生成模型。
+ )}
+ {userApiConfigRequired && modelOptions.length > 1 && (