2654400439/Paper2Seminar / skills/paper-ppt-orchestrator
paper-ppt-orchestrator
VerifiedBuild complete fixed-layout academic-paper presentations from a PDF with bundled PPTX and one-page HTML templates. Use for full paper reading, complexity-aware slide budgeting, subsection coverage mapping, figure extraction, TikZ or Matplotlib evidence visuals, deterministic PPTX assembly, and final QA. Trigger for requests to turn a paper PDF into a discussion-group deck, especially 12-16+ page systems, security, networking, or measurement papers.
webapp/streamlit_app.py
View on GitHub"""Paper2Seminar local task console."""
from __future__ import annotations
import hashlib
import json
import sys
from pathlib import Path
import streamlit as st
WEBAPP_DIR = Path(__file__).resolve().parent
SCRIPT_DIR = WEBAPP_DIR.parent / "scripts"
if str(WEBAPP_DIR) not in sys.path:
sys.path.insert(0, str(WEBAPP_DIR))
if str(SCRIPT_DIR) not in sys.path:
sys.path.insert(0, str(SCRIPT_DIR))
from backend import ( # noqa: E402
CodexAdapter,
UIError,
create_job,
inspect_pdf,
list_artifacts,
load_status,
repository_root,
run_preflight,
start_job,
)
from job_request import load_contract # noqa: E402
REPO_ROOT = repository_root()
JOBS_ROOT = REPO_ROOT / "runs"
_JOB_SCHEMA, JOB_DEFAULTS, FEATURE_REGISTRY = load_contract()
DEFAULT_PROFILE_NAME = JOB_DEFAULTS["default_profile"]
DEFAULT_PROFILE = JOB_DEFAULTS["profiles"][DEFAULT_PROFILE_NAME]
def option_index(options: list[str], selected: str) -> int:
return options.index(selected) if selected in options else 0
st.set_page_config(page_title="Paper2Seminar", page_icon="📑", layout="wide")
st.markdown(
"""
<style>
:root { --ink: #17212b; --muted: #66717c; --line: #d8dee5; --blue: #194a96; --teal: #147d78; --amber: #a86508; }
.stApp { background: #f7f9fb; color: var(--ink); }
[data-testid="stHeader"] { background: rgba(247,249,251,.96); }
.block-container { max-width: 1180px; padding-top: 2rem; padding-bottom: 4rem; }
h1, h2, h3 { letter-spacing: 0 !important; color: var(--ink); }
h1 { font-size: 2rem !important; }
h2 { font-size: 1.25rem !important; margin-top: 1.5rem !important; }
div[data-testid="stMetric"] { background: #fff; border: 1px solid var(--line); border-radius: 6px; padding: 12px 14px; }
div[data-testid="stForm"], div[data-testid="stExpander"] { border-color: var(--line); border-radius: 6px; background: #fff; }
.status-line { padding: 12px 14px; border-left: 4px solid var(--blue); background: #fff; border-radius: 4px; }
.muted { color: var(--muted); font-size: .9rem; }
.reserved { color: var(--amber); font-weight: 600; }
</style>
""",
unsafe_allow_html=True,
)
def recent_jobs() -> list[Path]:
if not JOBS_ROOT.is_dir():
return []
return sorted(
(path for path in JOBS_ROOT.glob("ui-*") if (path / "job-request.json").is_file()),
key=lambda path: path.stat().st_mtime,
reverse=True,
)
def render_status(job_dir: Path) -> None:
status = load_status(job_dir)
request = json.loads((job_dir / "job-request.json").read_text(encoding="utf-8"))
notes_enabled = request["features"]["speaker_notes"]["enabled"]
state = status.get("state", "unknown")
labels = {
"created": ("已创建", 5),
"running": ("制作中", 55),
"completed": ("已完成", 100),
"failed": ("失败", 100),
"unknown": ("未知", 0),
}
label, progress = labels.get(state, (state, 0))
st.progress(progress, text=f"{label} · {status.get('message', '')}")
cols = st.columns(4)
expected = [
("论文理解", job_dir / "paper-notes.json"),
("页面规划", job_dir / "deck-plan.json"),
("PPT 组装", job_dir / "build" / "presentation.pptx"),
("后置讲稿", job_dir / "build" / "presentation.notes.json"),
]
for column, (name, path) in zip(cols, expected):
value = "完成" if path.exists() else "等待"
if name == "后置讲稿" and not notes_enabled:
value = "未启用"
column.metric(name, value)
log_path = job_dir / "agent.log"
if log_path.is_file():
lines = log_path.read_text(encoding="utf-8", errors="replace").splitlines()
with st.expander("执行日志", expanded=state == "failed"):
st.code("\n".join(lines[-120:]) or "暂无日志", language="text")
artifacts = list_artifacts(job_dir)
deliverables = [path for path in artifacts if path.suffix.lower() in {".pptx", ".json", ".md"}]
if deliverables:
st.subheader("任务产物")
for path in deliverables:
relative = path.relative_to(job_dir)
col_name, col_size, col_action = st.columns([5, 1, 1])
col_name.write(str(relative))
col_size.caption(f"{path.stat().st_size / 1024:.0f} KB")
col_action.download_button(
"下载",
data=path.read_bytes(),
file_name=path.name,
key=f"download-{hashlib.sha1(str(path).encode()).hexdigest()}",
)
st.title("Paper2Seminar(测试版)")
st.caption("本地论文汇报任务控制台")
with st.sidebar:
st.subheader("运行状态")
codex_authentication = CodexAdapter.authentication()
if codex_authentication["authenticated"]:
codex_label = "已安装并登录"
elif codex_authentication["installed"]:
codex_label = "已安装,未登录"
else:
codex_label = "未检测到"
st.write("Codex CLI", codex_label)
jobs = recent_jobs()
selected_job = st.selectbox(
"最近任务",
options=[""] + [path.name for path in jobs],
format_func=lambda value: "未选择" if not value else value,
)
st.caption(f"任务目录:{JOBS_ROOT}")
create_tab, status_tab, environment_tab = st.tabs(["新建任务", "任务状态", "环境检查"])
with create_tab:
st.subheader("1. 论文与模板")
pdf_file = st.file_uploader("论文 PDF", type=["pdf"])
template_file = st.file_uploader("演示模板(可选)", type=["pptx"])
inspection = None
if pdf_file is not None:
pdf_payload = pdf_file.getvalue()
if len(pdf_payload) > 100 * 1024 * 1024:
st.error("PDF 超过 100 MB,请先压缩后再上传。")
st.stop()
digest = hashlib.sha256(pdf_payload).hexdigest()
if st.session_state.get("inspected_pdf_sha") != digest:
try:
st.session_state["pdf_inspection"] = inspect_pdf(pdf_payload)
st.session_state["inspected_pdf_sha"] = digest
st.session_state["confirmed_title_en"] = st.session_state["pdf_inspection"]["title"]
st.session_state["confirmed_title_cn"] = ""
except UIError as exc:
st.error(str(exc))
inspection = st.session_state.get("pdf_inspection")
if inspection:
metrics = st.columns(3)
metrics[0].metric("页数", inspection["page_count"])
metrics[1].metric("文本解析", "通过" if inspection["text_extractable"] else "失败")
metrics[2].metric("文件指纹", inspection["sha256"][:10])
st.text_input("确认论文原文标题", key="confirmed_title_en")
st.text_input("中文标题(可留空,由 agent 翻译)", key="confirmed_title_cn")
with st.expander("标题识别候选"):
for candidate in inspection["candidates"]:
st.write(candidate)
st.subheader("2. 汇报配置")
identity_left, identity_right = st.columns(2)
presenter = identity_left.text_input("汇报人")
advisor = identity_right.text_input("指导老师")
config_left, config_middle, config_right = st.columns(3)
language_options = ["zh-CN", "bilingual", "en-US"]
language = config_left.selectbox(
"输出语言",
language_options,
index=option_index(language_options, DEFAULT_PROFILE["presentation"]["language"]),
format_func={"zh-CN": "中文", "bilingual": "中英双语", "en-US": "英文"}.get,
)
target_minutes = config_middle.number_input(
"目标时长(分钟)",
min_value=5,
max_value=120,
value=DEFAULT_PROFILE["presentation"]["target_minutes"],
step=5,
)
slide_mode_options = ["auto", "manual"]
slide_count_mode = config_right.selectbox(
"页数策略",
slide_mode_options,
index=option_index(slide_mode_options, DEFAULT_PROFILE["presentation"]["slide_count_mode"]),
format_func={"auto": "按论文复杂度自动", "manual": "手动指定"}.get,
)
target_slide_count = None
if slide_count_mode == "manual":
target_slide_count = st.number_input("目标总页数", min_value=8, max_value=80, value=30)
st.subheader("3. 图表、视觉与质量")
extraction_col, visual_col, qa_col = st.columns(3)
with extraction_col:
extraction_options = ["doclayout", "captioncrop"]
extraction_backend = st.selectbox(
"论文图表裁剪",
extraction_options,
index=option_index(extraction_options, DEFAULT_PROFILE["extraction"]["backend"]),
format_func={"doclayout": "DocLayout-YOLO(默认)", "captioncrop": "CaptionCrop(轻量)"}.get,
)
device_options = ["cpu", "cuda", "mps"]
device = st.selectbox(
"推理设备",
device_options,
index=option_index(device_options, DEFAULT_PROFILE["extraction"]["device"]),
)
confidence = st.slider(
"检测阈值", 0.05, 0.80, float(DEFAULT_PROFILE["extraction"]["confidence"]), 0.01
)
crop_dpi = st.number_input(
"裁剪 DPI",
min_value=150,
max_value=600,
value=DEFAULT_PROFILE["extraction"]["crop_dpi"],
step=25,
)
captioncrop_command = ""
if extraction_backend == "captioncrop":
captioncrop_command = st.text_input("CaptionCrop 命令路径")
with visual_col:
capability_labels = {
"unknown": "由 agent 首次确认",
"available": "当前会话可用",
"unavailable": "当前会话不可用",
}
capability_options = ["unknown", "available", "unavailable"]
imagegen_state = st.selectbox(
"概念图片生成能力",
capability_options,
index=option_index(capability_options, DEFAULT_PROFILE["visuals"]["imagegen"]),
format_func=capability_labels.get,
)
web_search_state = st.selectbox(
"授权图片搜索能力",
capability_options,
index=option_index(capability_options, DEFAULT_PROFILE["visuals"]["web_search"]),
format_func=capability_labels.get,
)
with qa_col:
readability_default = DEFAULT_PROFILE["features"]["readability"]["config"]["mode"]
readability_options = ["overview", "full", "off"]
readability_mode = st.selectbox(
"最终可读性检查",
readability_options,
index=option_index(readability_options, readability_default),
format_func={"overview": "总览检查(推荐)", "full": "逐组精查", "off": "关闭"}.get,
)
st.caption("其余可选功能统一显示在下方功能配置中。")
st.subheader("4. 逐页讲稿")
speaker_default = DEFAULT_PROFILE["features"]["speaker_notes"]
notes_enabled = st.toggle("核心 PPT 完成后生成逐字稿", value=speaker_default["enabled"])
notes_left, notes_middle, notes_right = st.columns(3)
notes_left.text_input("生成阶段", value="页面审批后(post-QA)", disabled=True)
delivery_style = notes_middle.selectbox("讲稿形式", ["verbatim"], format_func=lambda _: "可直接口述的逐字稿", disabled=not notes_enabled)
pace = notes_right.number_input(
"参考语速(中英混合单位/分钟)",
min_value=80,
max_value=400,
value=speaker_default["config"]["pace_units_per_minute"],
step=10,
disabled=not notes_enabled,
)
st.caption("讲稿只读取页面文字、可视化设计元数据、论文笔记和论文文本;不会通过幻灯片截图生成。")
st.subheader("5. 可选功能")
generic_feature_values: dict[str, dict[str, object]] = {}
feature_columns = st.columns(2)
generic_features = [
(feature_id, metadata)
for feature_id, metadata in FEATURE_REGISTRY["features"].items()
if feature_id not in {"speaker_notes", "readability"}
]
for index, (feature_id, metadata) in enumerate(generic_features):
default_value = DEFAULT_PROFILE["features"][feature_id]
reserved = metadata["status"] == "reserved"
label = metadata["label_zh"] + (" · 未完成" if reserved else "")
enabled = feature_columns[index % 2].checkbox(
label,
value=default_value["enabled"],
disabled=reserved,
key=f"feature-{feature_id}",
)
generic_feature_values[feature_id] = {"enabled": enabled, "config": {}}
with st.expander("高级配置"):
st.selectbox("Agent 适配器", ["codex"], format_func=lambda _: "Codex CLI")
st.text_input("OpenCode 适配器 · 未完成", value="", disabled=True)
st.text_input("Claude Code 适配器 · 未完成", value="", disabled=True)
ready = pdf_file is not None and inspection is not None and inspection.get("text_extractable")
if st.button("创建并开始制作", type="primary", disabled=not ready, use_container_width=True):
if not codex_authentication["installed"]:
st.error("未检测到 Codex CLI,当前版本无法启动 agent 任务。")
elif not codex_authentication["authenticated"]:
st.error("Codex CLI 尚未登录。请先在 PowerShell 中运行以下命令,然后刷新页面:")
st.code('& "$env:APPDATA\\npm\\codex.CMD" login', language="powershell")
else:
configuration = {
"title_en": st.session_state.get("confirmed_title_en", ""),
"title_cn": st.session_state.get("confirmed_title_cn", ""),
"presenter": presenter.strip(),
"advisor": advisor.strip(),
"language": language,
"target_minutes": int(target_minutes),
"slide_count_mode": slide_count_mode,
"target_slide_count": int(target_slide_count) if target_slide_count else None,
"extraction": {
"backend": extraction_backend,
"device": device,
"confidence": confidence,
"crop_dpi": int(crop_dpi),
"captioncrop_command": captioncrop_command.strip(),
},
"visual_capabilities": {
"imagegen": imagegen_state,
"web_search": web_search_state,
},
"readability_mode": readability_mode,
"speaker_notes": {
"enabled": notes_enabled,
"generation_stage": "post_qa",
"delivery_style": delivery_style,
"target_minutes": int(target_minutes),
"pace_units_per_minute": int(pace),
},
"features": generic_feature_values,
}
try:
request = create_job(
pdf_file.getvalue(),
pdf_file.name,
configuration,
template_payload=template_file.getvalue() if template_file else None,
)
pid = start_job(request)
st.session_state["active_job"] = request["job"]["id"]
st.success(f"任务已启动,进程号 {pid}")
st.code(request["execution"]["paths"]["job_dir"], language="text")
except UIError as exc:
st.error(str(exc))
with status_tab:
active_name = st.session_state.get("active_job") or selected_job
active_dir = JOBS_ROOT / active_name if active_name else None
if active_dir and active_dir.is_dir():
st.subheader(active_name)
render_status(active_dir)
st.button("刷新状态", use_container_width=True)
else:
st.info("选择或创建任务后,这里会显示进度、日志和下载项。")
with environment_tab:
st.subheader("本机能力")
if codex_authentication["authenticated"]:
st.success("Codex CLI 已安装并登录,可以接收后台任务。")
elif codex_authentication["installed"]:
st.warning("Codex CLI 已安装但尚未登录,完整任务当前无法启动。")
st.code('& "$env:APPDATA\\npm\\codex.CMD" login', language="powershell")
else:
st.error("未检测到 Codex CLI。")
env_left, env_right = st.columns(2)
declared_imagegen = env_left.selectbox("当前 agent 的图片生成能力", ["available", "unavailable", "unknown"])
declared_web = env_right.selectbox("当前 agent 的网络搜索能力", ["available", "unavailable", "unknown"])
if st.button("运行环境检查", use_container_width=True):
try:
report = run_preflight(JOBS_ROOT / "ui-preflight" / "capabilities.json", declared_imagegen, declared_web)
tool_rows = []
for name, record in report["tools"].items():
if name == "browsers":
tool_rows.append({"工具": "browser", "状态": "available" if record else "unavailable", "版本/路径": record[0]["path"] if record else ""})
else:
tool_rows.append({"工具": name, "状态": record.get("status", "unknown"), "版本/路径": record.get("version") or record.get("path") or ""})
st.dataframe(tool_rows, use_container_width=True, hide_index=True)
for warning in report["warnings"]:
st.warning(warning)
except UIError as exc:
st.error(str(exc))
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |