Make Space public and speed up initial UI

README.md

@@ -16,9 +16,9 @@ startup_duration_timeout: 30m
 
 Hugging Face Gradio Space，通过 **OpenRouter** 调用 Claude。支持中文多轮聊天、流式输出、联网搜索和文档/图片上传，无需下载模型权重。
 
-## 模型与价格
+## 支持的模型
 
-界面启动和点击「刷新价格」时从 [OpenRouter 模型目录 API](https://openrouter.ai/docs/api/api-reference/models/get-models)读取报价。网络不可用时明确显示参考日期，不将旧价格标为实时。
+为缩短首屏加载时间，界面不再请求 OpenRouter 模型目录，也不显示价格面板。每次调用后仍会显示 OpenRouter 实际返回的 token 用量和费用；未返回费用时提示到 Activity 查看。
 
 2026-10-06 核实的价格（美元 / 每百万 tokens）：
 
@@ -41,10 +41,10 @@ Hugging Face Gradio Space，通过 **OpenRouter** 调用 Claude。支持中文
 | 类别 | 扩展名 | 处理方式 |
 |---|---|---|
 | PDF | `.pdf` | Base64 文件交给 Claude 原生阅读，支持扫描页和图表 |
-| Word | `.doc`, `.docx` | LibreOffice 转 PDF 后交给模型，保留页面、图片和表格 |
-| PowerPoint | `.ppt`, `.pptx` | LibreOffice 转 PDF，按幻灯片页面阅读；不包含讲者备注、动画或隐藏页 |
-| Excel | `.xls`, `.xlsx`, `.csv` | 按工作表、行号读取单元格文本；XLSX 保留公式，XLS 使用已保存的值 |
-| 图片 | `.png`, `.jpg`, `.jpeg`, `.webp`, `.gif`, `.bmp`, `.tif`, `.tiff` | 单帧图片转 JPEG，最长边缩放至 2048px，发送视觉输入 |
+| WORD | `.DOC`, `.DOCX` | LibreOffice 转 PDF 后交给模型，保留页面、图片和表格 |
+| POWERPOINT | `.PPT`, `.PPTX` | LibreOffice 转 PDF，按幻灯片页面阅读；不包含讲者备注、动画或隐藏页 |
+| EXCEL | `.XLS`, `.XLSX`, `.CSV` | 按工作表、行号读取单元格文本；XLSX 保留公式，XLS 使用已保存的值 |
+| IMAGES | `.PNG`, `.JPG`, `.JPEG`, `.WEBP`, `.GIF`, `.BMP`, `.TIF`, `.TIFF` | 单帧图片转 JPEG，最长边缩放至 2048px，发送视觉输入 |
 
 Excel 不执行公式，不分析嵌入图表/图片；需要视觉分析时请另存为 PDF 或上传截图。多帧 GIF/TIFF 明确拒绝，避免只读首帧而遗漏内容。Office 转换可能改变字体、分页；对版式要求严格时建议直接上传原始 PDF。加密文件需先移除密码。
 
@@ -102,7 +102,7 @@ API Key 只在服务器读取，浏览器不会收到。上传文件在 Space
 
 ```text
 app.py           Gradio 界面、会话管理与流式显示
-router.py        模型目录、价格、API 请求、引用和费用
+router.py        模型配置、API 请求、引用和费用
 attachments.py   PDF / Office / Excel / 图片处理
 requirements.txt Python 依赖（Gradio 由平台管理）
 packages.txt     系统依赖（LibreOffice、中文字体）

app.py

@@ -1,4 +1,4 @@
-"""Claude Document Studio — a private, server-side OpenRouter Space."""
+"""Claude Document Studio — a server-side OpenRouter Space."""
 
 import copy
 import json
@@ -7,6 +7,9 @@ import time
 from datetime import datetime, timezone
 from pathlib import Path
 
+# Avoid a nonessential analytics request during initial page load.
+os.environ.setdefault("GRADIO_ANALYTICS_ENABLED", "False")
+
 # Optional ZeroGPU registration for eligible free HF accounts. Never invoked by chat.
 if os.environ.get("USE_ZEROGPU") == "1":
     import spaces
@@ -20,7 +23,7 @@ import gradio as gr
 
 from attachments import ALLOWED, FileError, prepare_files
 from router import (DEFAULT_MODEL, MODELS, RouterError, build_payload, citations_markdown,
-                    default_choices, price_catalog, stream_reply, usage_summary)
+                    stream_reply, usage_summary)
 
 CSS = """
 .gradio-container { max-width: 1240px !important; margin: auto; }
@@ -31,12 +34,7 @@ CSS = """
 """
 MAX_REQUEST_BYTES = 28 * 1024 * 1024
 MAX_TURNS = 40
-
-
-def refresh_prices(model: str) -> tuple:
-    """Refresh the three model prices from OpenRouter's public catalog."""
-    choices, table = price_catalog(force=True)
-    return gr.update(choices=choices, value=model if model in MODELS else DEFAULT_MODEL), table
+MODEL_CHOICES = [(details[0], model_id) for model_id, details in MODELS.items()]
 
 
 def chat(text: str, files: list[str] | None, session: dict, model: str,
@@ -160,19 +158,18 @@ def create_app() -> gr.Blocks:
         session = gr.State({"messages": [], "display": [], "pages": 0}, time_to_live=3600)
         with gr.Row():
             with gr.Column(scale=1, min_width=290):
-                model = gr.Dropdown(default_choices(), value=DEFAULT_MODEL, label="选择模型 · 美元 / 百万 tokens")
+                model = gr.Dropdown(MODEL_CHOICES, value=DEFAULT_MODEL, label="选择模型")
                 search = gr.Checkbox(value=True, label="开启联网搜索", info="模型按需检索；搜索另行计费。")
-                files = gr.File(label="上传文档 / 图片", file_count="multiple", file_types=ALLOWED, type="filepath")
-                gr.Markdown("PDF · Word（doc/docx）· PPT（ppt/pptx）· Excel（xls/xlsx/csv）· 图片\n\n"
+                files = gr.File(label="上传 PDF / WORD / POWERPOINT / EXCEL / IMAGES",
+                                file_count="multiple", file_types=ALLOWED, type="filepath")
+                gr.Markdown("PDF · WORD (DOC/DOCX) · POWERPOINT (PPT/PPTX) · "
+                            "EXCEL (XLS/XLSX/CSV) · IMAGES\n\n"
                             "每次 ≤ 8 个文件，单文件 ≤ 15 MB，总计 ≤ 20 MB；会话 PDF ≤ 100 页。"
-                            "Excel 读取单元格和公式，不读取嵌入图表。")
+                            "EXCEL 读取单元格和公式，不读取嵌入图表。")
                 with gr.Accordion("生成设置", open=False):
                     effort = gr.Dropdown([("低", "low"), ("中", "medium"), ("高", "high"),
                                           ("更高", "xhigh"), ("最高", "max")], value="medium", label="推理强度")
                     max_tokens = gr.Slider(1024, 32768, value=8192, step=1024, label="输出上限（含推理 tokens）")
-                with gr.Accordion("模型价格", open=True):
-                    prices = gr.Markdown("正在加载价格…")
-                    refresh = gr.Button("刷新价格", size="sm")
             with gr.Column(scale=3, min_width=440):
                 chatbot = gr.Chatbot(height=540, label="对话", layout="bubble")
                 prompt = gr.Textbox(label="输入问题", placeholder="输入问题，或上传文件后让我分析…", lines=3)
@@ -192,8 +189,6 @@ def create_app() -> gr.Blocks:
         stop.click(stop_chat, inputs=[session], outputs=[chatbot, prompt, files, status, send, clear],
                    cancels=[event], queue=False, api_name=False)
         clear.click(clear_chat, outputs=outputs, api_name=False)
-        refresh.click(refresh_prices, inputs=[model], outputs=[model, prices], api_name="prices")
-        demo.load(refresh_prices, inputs=[model], outputs=[model, prices], api_name=False)
     return demo
 
 
@@ -206,7 +201,7 @@ if __name__ == "__main__":
     demo.queue(max_size=16).launch(
         server_name="0.0.0.0", server_port=int(os.environ.get("PORT", "7860")),
         theme=gr.themes.Soft(primary_hue="indigo"), css=CSS,
-        auth=auth, max_file_size="15mb", show_error=False, mcp_server=True,
+        auth=auth, max_file_size="15mb", show_error=False,
         run_history=False, state_session_capacity=100,
         allowed_paths=[], blocked_paths=[str(Path(__file__).parent / ".env")],
     )

requirements.txt

@@ -4,4 +4,3 @@ pypdf>=6.0,<7
 openpyxl>=3.1.5,<4
 xlrd>=2.0.2,<3
 defusedxml>=0.7.1,<1
-mcp>=1.0,<2

router.py

@@ -1,12 +1,7 @@
-"""OpenRouter transport and live model pricing. No credentials in UI or logs."""
+"""OpenRouter transport and model configuration. No credentials in UI or logs."""
 
-import copy
 import json
 import os
-import threading
-import time
-from datetime import datetime, timezone
-from decimal import Decimal
 from urllib.parse import urlsplit
 
 import httpx
@@ -18,63 +13,12 @@ MODELS = {
     "anthropic/claude-fable-5.1": ("Claude Fable 5.1", "10", "50"),
 }
 DEFAULT_MODEL = "anthropic/claude-sonnet-5.5"
-_cache = {"data": {}, "time": 0.0, "updated": ""}
-_lock = threading.Lock()
 
 
 class RouterError(Exception):
     """A safe, user-facing error without request headers or document content."""
 
 
-def price_catalog(force: bool = False) -> tuple[list, str]:
-    """Refresh public prices; label dated fallback prices when offline."""
-    error = False
-    with _lock:
-        if force or time.monotonic() - _cache["time"] > 900 or not _cache["time"]:
-            try:
-                response = httpx.get(f"{API_BASE}/models", timeout=15)
-                response.raise_for_status()
-                data = {m["id"]: m for m in response.json()["data"] if m["id"] in MODELS}
-                _cache.update(data=data, time=time.monotonic(),
-                              updated=datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M UTC"))
-            except (httpx.HTTPError, ValueError, KeyError, TypeError):
-                error = True
-        data = copy.deepcopy(_cache["data"])
-        updated = _cache["updated"]
-    rows, choices = [], []
-    for model_id, (name, fallback_in, fallback_out) in MODELS.items():
-        m = data.get(model_id)
-        try:
-            p = m["pricing"]
-            inp = format((Decimal(p["prompt"]) * 1_000_000).normalize(), "f")
-            out = format((Decimal(p["completion"]) * 1_000_000).normalize(), "f")
-            search = f'${Decimal(p["web_search"]):g}/次' if p.get("web_search") else "按搜索引擎计费"
-            source = "目录报价"
-        except (KeyError, TypeError, ArithmeticError):
-            inp, out, search = fallback_in, fallback_out, "$0.01/次（参考）"
-            source = "2026-10-06 参考价"
-        if m is None and updated:
-            source += "；当前目录未列出"
-        choices.append((f"{name} · 输入 ${inp} / 输出 ${out}", model_id))
-        rows.append(f"| [{name}](https://openrouter.ai/{model_id}) | ${inp} | ${out} | {search} | {source} |")
-    note = f"目录更新时间：{updated}" if updated else "暂未取得实时目录，显示 2026-10-06 已核实的参考价格。"
-    if error:
-        note += " · 刷新失败，保留上次价格。"
-    table = "\n".join([
-        "**美元 / 每百万 tokens**，搜索按次另计。", "",
-        "| 模型 | 输入 / 1M | 输出 / 1M | 原生搜索 | 价格来源 |",
-        "|---|---:|---:|---|---|", *rows, "", note,
-        "", "搜索默认使用原生引擎；供应商不支持时 OpenRouter 可回退到 Exa，按该引擎费率结算。"
-        "图片、PDF、搜索结果和历史消息均会消耗输入 tokens；推理 tokens 计入输出。"
-        "实际账单以 OpenRouter usage 和供应商费率为准。",
-    ])
-    return choices, table
-
-
-def default_choices() -> list:
-    return [(f"{name} · 输入 ${i} / 输出 ${o}", model) for model, (name, i, o) in MODELS.items()]
-
-
 def build_payload(messages: list, model: str, search: bool, max_tokens: int, effort: str) -> dict:
     if model not in MODELS:
         raise RouterError("请选择列表中的模型。")

scripts/smoke_remote.py

@@ -26,7 +26,7 @@ def main() -> None:
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument("repo_id")
     parser.add_argument("--fixtures", type=Path)
-    parser.add_argument("--case", choices=["models", "files", "search", "errors", "prices"], required=True)
+    parser.add_argument("--case", choices=["models", "files", "search", "errors"], required=True)
     args = parser.parse_args()
     client = Client(args.repo_id, token=get_token(), httpx_kwargs={"timeout": 360}, verbose=False)
 
@@ -36,11 +36,7 @@ def main() -> None:
         print(json.dumps({"model": model, "answer": answer[:2000], "status": result[-1]}, ensure_ascii=False), flush=True)
         return answer, result
 
-    if args.case == "prices":
-        result = client.predict("anthropic/claude-sonnet-5.5", api_name="/prices")
-        assert all(name in result[-1] for name in ["Opus 5.5", "Sonnet 5.5", "Fable 5.1"])
-        print(result[-1], flush=True)
-    elif args.case == "models":
+    if args.case == "models":
         for slug in ["sonnet-5.5", "opus-5.5", "fable-5.1"]:
             answer, result = chat("This is a connectivity test. Reply with exactly CONNECTED.", model="anthropic/claude-" + slug)
             assert "CONNECTED" in answer, result[-1]