from __future__ import annotations import re from dataclasses import dataclass from hydradeck.core.types import ExtractedFact, Source _LATEX_SPECIALS: dict[str, str] = { "\\": r"\textbackslash{}", "{": r"\{", "}": r"\}", "#": r"\#", "$": r"\$", "%": r"\%", "&": r"\&", "_": r"\_", "^": r"\textasciicircum{}", "~": r"\textasciitilde{}", } def latex_escape(s: str) -> str: return "".join(_LATEX_SPECIALS.get(ch, ch) for ch in s) def _bib_key(i: int) -> str: return f"src{i}" def _bib_escape(s: str) -> str: return s.replace("\\", "\\\\").replace("{", "\\{").replace("}", "\\}") def render_bibtex(sources: list[Source]) -> str: lines: list[str] = [] for i, s in enumerate(sources, start=1): key = _bib_key(i) lines.append(f"@misc{{{key},") lines.append(f" title = {{{_bib_escape(s.title)}}},") lines.append(f" howpublished = {{\\url{{{_bib_escape(s.url)}}}}},") lines.append(" note = {Accessed: 2026-03-04},") lines.append("}") lines.append("") return "\n".join(lines).strip() + "\n" def _replace_numeric_citations(text: str, max_n: int) -> str: def repl(m: re.Match[str]) -> str: num = int(m.group(1)) if 1 <= num <= max_n: return f"\\cite{{{_bib_key(num)}}}" return m.group(0) return re.sub(r"\[(\d{1,3})\]", repl, text) def _markdown_to_latex_paragraphs(md: str, max_n: int) -> str: text = md.strip() text = re.sub(r"```[\s\S]*?```", "", text) text = re.sub(r"^\s*[-*+]\s+", "", text, flags=re.MULTILINE) text = re.sub(r"^\s*#+\s*", "", text, flags=re.MULTILINE) text = re.sub(r"`([^`]+)`", r"\1", text) text = re.sub(r"\*\*(.*?)\*\*", r"\1", text) text = re.sub(r"\*(.*?)\*", r"\1", text) text = re.sub(r"\[(.*?)\]\((.*?)\)", r"\1", text) text = _replace_numeric_citations(text, max_n=max_n) text = latex_escape(text) text = re.sub(r"\\textbackslash\{\}cite\\\{(src\d+)\\\}", r"\\cite{\1}", text) text = text.replace("\n\n", "\n\\par\n") return text def render_paper( topic: str, outline: list[str], body: str, facts: list[ExtractedFact], sources: list[Source], ) -> str: topic_e = latex_escape(topic) url_to_key = {s.url: _bib_key(i) for i, s in enumerate(sources, start=1)} outline_items = "\n".join([f"\\item {latex_escape(x)}" for x in outline[:10]]) fact_sentences: list[str] = [] for f in facts[:18]: key = url_to_key.get(f.url) cite = f"\\cite{{{key}}}" if key else "" sentence = latex_escape(f.claim.strip()) if sentence and sentence[-1] not in ".!?": sentence += "." fact_sentences.append(sentence + (cite if cite else "")) facts_paragraph = ( " ".join(fact_sentences) if fact_sentences else "No extracted facts available." ) body_latex = _markdown_to_latex_paragraphs(body, max_n=len(sources)) return ( "\\documentclass[11pt]{article}\n" "\\usepackage{geometry}\n" "\\usepackage{hyperref}\n" "\\usepackage{url}\n" "\\usepackage{booktabs}\n" "\\usepackage{longtable}\n" "\\geometry{margin=1in}\n" "\\hypersetup{colorlinks=true,linkcolor=black,citecolor=blue,urlcolor=blue}\n" f"\\title{{{topic_e}}}\n" "\\author{hydradeck}\n" "\\date{\\today}\n" "\\begin{document}\n" "\\maketitle\n" "\\begin{abstract}\n" "This report presents a structured analysis with explicit traceability to sources.\n" "\\end{abstract}\n\n" "\\section*{1. Introduction and Background}\n" + facts_paragraph + "\n\n" "\\section*{2. Logical Outline}\n" "\\begin{itemize}\n" + outline_items + "\n\\end{itemize}\n\n" "\\section*{3. Evidence and Key Findings}\n" + body_latex + "\n\n" "\\section*{4. Limitations and Discussion}\n" "The analysis is bounded by available public evidence and may evolve as sources update.\n\n" "\\section*{5. Conclusion}\n" "Conclusions are presented in a source-traceable form and should be interpreted with the\n" "reported assumptions and constraints.\n\n" "\\bibliographystyle{plain}\n" "\\bibliography{refs}\n" "\\end{document}\n" ) def render_report_structured( topic: str, section_blocks: list[dict[str, str]], language: str = "en", ) -> str: lang = language.lower() topic_e = latex_escape(topic) if lang == "zh": preamble = ( "\\documentclass[11pt]{ctexart}\n" "\\usepackage[a4paper,margin=1in]{geometry}\n" "\\usepackage{hyperref}\n" "\\usepackage{url}\n" "\\usepackage{booktabs}\n" "\\usepackage{longtable}\n" "\\hypersetup{colorlinks=true,linkcolor=black,citecolor=blue,urlcolor=blue}\n" f"\\title{{{topic_e}}}\n" "\\author{hydradeck}\n" "\\date{\\today}\n" "\\begin{document}\n" "\\maketitle\n" ) else: preamble = ( "\\documentclass[11pt]{article}\n" "\\usepackage{geometry}\n" "\\usepackage{hyperref}\n" "\\usepackage{url}\n" "\\usepackage{booktabs}\n" "\\usepackage{longtable}\n" "\\geometry{margin=1in}\n" "\\hypersetup{colorlinks=true,linkcolor=black,citecolor=blue,urlcolor=blue}\n" f"\\title{{{topic_e}}}\n" "\\author{hydradeck}\n" "\\date{\\today}\n" "\\begin{document}\n" "\\maketitle\n" ) content_parts: list[str] = [] for block in section_blocks[:10]: title = latex_escape(str(block.get("name", "Section")).strip() or "Section") latex_body = str(block.get("latex", "")).strip() latex_body = re.sub(r"\\section\*?\{[^}]*\}", "", latex_body) latex_body = re.sub(r"\\subsection\*?\{[^}]*\}", "", latex_body) latex_body = re.sub(r"\\cite\{[^}]*\}", "", latex_body) latex_body = re.sub(r"\[(\d{1,3})\]", "", latex_body) if not latex_body: continue content_parts.append(f"\\section*{{{title}}}\n{latex_body}\n") return preamble + "\n".join(content_parts) + "\n\\end{document}\n" @dataclass class SlideFrame: title: str bullets: list[str] note: str = "" def render_beamer(topic: str, outline: list[str], bullets: list[str]) -> str: section_blocks = [{"name": t, "latex": b} for t, b in zip(outline, bullets)] if not section_blocks: section_blocks = [{"name": "Summary", "latex": "Key findings and implications."}] frames = build_slide_frames_from_sections(section_blocks, language="en") frames = enforce_slide_density(frames, language="en") return render_beamer_frames(topic, frames, language="en") def render_beamer_from_report(topic: str, report_tex: str) -> str: frames = build_slide_frames_from_report(report_tex, language="en") frames = enforce_slide_density(frames, language="en") return render_beamer_frames(topic, frames, language="en") def _split_paragraph_to_bullets(text: str, language: str) -> list[str]: lang = language.lower() if lang == "zh": parts = [x.strip() for x in re.split(r"[。！？]\s*", text) if x.strip()] out: list[str] = [] for p in parts: if len(p) < 6: continue out.append(_trim_chars(_clean_text_for_slide(p), 28)) return out parts = [x.strip() for x in re.split(r"[.!?]\s+", text) if x.strip()] out2: list[str] = [] for p in parts: clean = _clean_text_for_slide(p) if len(clean) < 14: continue out2.append(_trim_words(clean, 14)) return out2 def build_slide_frames_from_sections( section_blocks: list[dict[str, str]], language: str = "en", ) -> list[SlideFrame]: lang = language.lower() frames: list[SlideFrame] = [] for block in section_blocks[:8]: title = str(block.get("name", "Section")).strip() or ("章节" if lang == "zh" else "Section") body = str(block.get("latex", "")) body = re.sub(r"\\section\*?\{[^}]*\}", "", body) body = re.sub(r"\\subsection\*?\{[^}]*\}", "", body) body = re.sub(r"\\cite\{[^}]*\}", "", body) body = re.sub(r"\[(\d{1,3})\]", "", body) bullets = _split_paragraph_to_bullets(body, lang) if not bullets: continue chunk = 4 if lang == "zh" else 4 for i in range(0, len(bullets), chunk): part = bullets[i : i + chunk] if not part: continue if i == 0: frame_title = title else: frame_title = f"{title}（续）" if lang == "zh" else f"{title} (cont.)" frames.append(SlideFrame(title=frame_title, bullets=part)) if not frames: raise RuntimeError("insufficient readable section content for slides") return frames def enforce_slide_density( frames: list[SlideFrame], language: str = "en", max_bullets_per_frame: int = 4, max_chars_per_bullet_zh: int = 28, max_words_per_bullet_en: int = 14, ) -> list[SlideFrame]: lang = language.lower() out: list[SlideFrame] = [] for fr in frames: normalized: list[str] = [] for b in fr.bullets: clean = _clean_text_for_slide(b) if not clean: continue if lang == "zh": clean = _trim_chars(clean, max_chars_per_bullet_zh) else: clean = _trim_words(clean, max_words_per_bullet_en) if clean: normalized.append(clean) if not normalized: continue for i in range(0, len(normalized), max_bullets_per_frame): chunk = normalized[i : i + max_bullets_per_frame] if not chunk: continue if i == 0: title = fr.title else: title = f"{fr.title}（续）" if lang == "zh" else f"{fr.title} (cont.)" out.append(SlideFrame(title=title, bullets=chunk, note=fr.note)) if not out: raise RuntimeError("slide density guard removed all frames") return out def _trim_words(text: str, max_words: int) -> str: words = text.split() if len(words) <= max_words: return text return " ".join(words[:max_words]).rstrip(" ,.;") + "..." def _trim_chars(text: str, max_chars: int) -> str: t = text.strip() if len(t) <= max_chars: return t return t[: max_chars - 1].rstrip("，。,. ") + "…" def _clean_text_for_slide(text: str) -> str: t = text.strip() t = re.sub(r"\s+", " ", t) t = re.sub(r"`([^`]+)`", r"\1", t) t = re.sub(r"\*\*(.*?)\*\*", r"\1", t) t = re.sub(r"\*(.*?)\*", r"\1", t) return t def build_slide_frames_from_report(report_tex: str, language: str = "en") -> list[SlideFrame]: lang = language.lower() sections = re.split(r"\\section\*\{([^}]+)\}", report_tex) parsed: list[tuple[str, str]] = [] if len(sections) >= 3: for i in range(1, len(sections), 2): title = sections[i].strip() body = sections[i + 1] if i + 1 < len(sections) else "" parsed.append((title, body)) if not parsed: raise RuntimeError("cannot derive slide frames from report structure") frames: list[SlideFrame] = [] for title, body in parsed[:8]: plain = re.sub(r"\\[a-zA-Z]+\*?(\[[^\]]*\])?(\{[^}]*\})?", " ", body) chunks = [x.strip() for x in re.split(r"[。.!?]\s+", plain) if x.strip()] bullets: list[str] = [] for c in chunks: clean = _clean_text_for_slide(c) if not clean: continue if lang == "zh": if len(clean) < 8: continue bullets.append(_trim_chars(clean, 30)) else: if len(clean) < 12: continue bullets.append(_trim_words(clean, 16)) if len(bullets) >= 5: break if not bullets: raise RuntimeError(f"insufficient bullet content for slide '{title}'") frames.append(SlideFrame(title=title, bullets=bullets)) return frames def render_beamer_frames(topic: str, frames: list[SlideFrame], language: str = "en") -> str: lang = language.lower() topic_e = latex_escape(topic) agenda_label = "目录" if lang == "zh" else "Agenda" summary_title = "总结" if lang == "zh" else "Summary" agenda_items = "\n".join([f"\\item {latex_escape(f.title)}" for f in frames[:8]]) frame_blocks: list[str] = [] for fr in frames[:10]: b = "\n".join([f"\\item {latex_escape(x)}" for x in fr.bullets[:5]]) frame_blocks.append( "\\begin{frame}[t]{" + latex_escape(fr.title) + "}\n" + "\\begin{itemize}\n" + b + "\n\\end{itemize}\n" + (f"\\vspace{{0.6em}}\\footnotesize {latex_escape(fr.note)}\n" if fr.note else "") + "\\end{frame}\n" ) summary_bullets: list[str] = [] for fr in frames[:5]: if fr.bullets: summary_bullets.append(fr.bullets[0]) if not summary_bullets: summary_bullets = ["关键要点见前页。" if lang == "zh" else "Key points are summarized in previous slides."] summary_items = "\n".join([f"\\item {latex_escape(x)}" for x in summary_bullets]) if lang == "zh": return ( "\\documentclass[aspectratio=169]{ctexbeamer}\n" "\\usetheme{Madrid}\n" "\\usefonttheme{professionalfonts}\n" "\\setbeamertemplate{navigation symbols}{}\n" "\\usepackage{hyperref}\n" "\\usepackage{booktabs}\n" "\\definecolor{AccentBlue}{HTML}{1F4E79}\n" "\\setbeamercolor{title}{fg=AccentBlue}\n" "\\setbeamercolor{frametitle}{fg=AccentBlue}\n" "\\setbeamerfont{title}{series=\\bfseries,size=\\Large}\n" "\\setbeamerfont{frametitle}{series=\\bfseries,size=\\large}\n" f"\\title{{{topic_e}}}\n" "\\author{hydradeck}\n" "\\date{\\today}\n" "\\begin{document}\n" "\\frame{\\titlepage}\n" "\\begin{frame}{" + latex_escape(agenda_label) + "}\n" "\\begin{itemize}\n" + agenda_items + "\n\\end{itemize}\n" "\\end{frame}\n" + "".join(frame_blocks) + "\\begin{frame}{" + latex_escape(summary_title) + "}\n" + "\\begin{itemize}\n" + summary_items + "\n\\end{itemize}\n" + "\\end{frame}\n" + "\\end{document}\n" ) return ( "\\documentclass[aspectratio=169]{beamer}\n" "\\usetheme{metropolis}\n" "\\usefonttheme{professionalfonts}\n" "\\setbeamertemplate{navigation symbols}{}\n" "\\usepackage{hyperref}\n" "\\usepackage{booktabs}\n" "\\definecolor{AccentBlue}{HTML}{1F4E79}\n" "\\setbeamercolor{title}{fg=AccentBlue}\n" "\\setbeamercolor{frametitle}{fg=AccentBlue}\n" "\\setbeamerfont{title}{series=\\bfseries,size=\\Large}\n" "\\setbeamerfont{frametitle}{series=\\bfseries,size=\\large}\n" f"\\title{{{topic_e}}}\n" "\\author{hydradeck}\n" "\\date{\\today}\n" "\\begin{document}\n" "\\frame{\\titlepage}\n" "\\begin{frame}{" + latex_escape(agenda_label) + "}\n" "\\begin{itemize}\n" + agenda_items + "\n\\end{itemize}\n" "\\end{frame}\n" + "".join(frame_blocks) + "\\begin{frame}{" + latex_escape(summary_title) + "}\n" + "\\begin{itemize}\n" + summary_items + "\n\\end{itemize}\n" + "\\end{frame}\n" + "\\end{document}\n" )