Stream Jupyter workspace file downloads

backend/api/server.py

@@ -83,6 +83,9 @@ FEISHU_SPREADSHEET_MAX_CELL_CHARS = 5000
 FEISHU_SPREADSHEET_WRITE_BATCH_ROWS = 500
 FEISHU_ONLINE_DOCS_FILENAME = 'online-docs.json'
 JUPYTER_WORKSPACE_FILENAME = 'jupyter_workspace.json'
+JUPYTER_STREAM_CHUNK_BYTES = 1024 * 1024
+FEISHU_REMOTE_DOC_MAX_BYTES = 50 * 1024 * 1024
+FEISHU_REMOTE_SHEET_MAX_BYTES = 100 * 1024 * 1024
 
 
 @dataclass
@@ -1096,11 +1099,20 @@ def create_app(state: AgentState) -> FastAPI:
         if runtime is None:
             raise HTTPException(status_code=404, detail='Jupyter runtime is not connected')
         try:
-            data, filename = runtime.read_file_bytes(path)
+            response, filename = runtime.open_file_stream(path)
         except JupyterRuntimeError as exc:
             raise HTTPException(status_code=400, detail=str(exc)) from exc
-        return Response(
-            data,
+
+        def stream_chunks() -> Any:
+            try:
+                for chunk in response.iter_content(chunk_size=JUPYTER_STREAM_CHUNK_BYTES):
+                    if chunk:
+                        yield chunk
+            finally:
+                response.close()
+
+        return StreamingResponse(
+            stream_chunks(),
             media_type=_content_type_for_filename(filename),
             headers={
                 'content-disposition': (
@@ -3050,9 +3062,27 @@ def _materialize_jupyter_file_for_feishu(
     if runtime is None:
         raise HTTPException(status_code=404, detail='Jupyter runtime is not connected')
     try:
-        data, filename = runtime.read_file_bytes(remote_path)
+        info = runtime.file_info(remote_path)
+        filename = _safe_uploaded_filename(str(info.get('name') or 'remote-file'))
+        suffix = Path(filename).suffix.lower()
+        max_bytes = (
+            FEISHU_REMOTE_SHEET_MAX_BYTES
+            if suffix in FEISHU_SPREADSHEET_SUFFIXES
+            else FEISHU_REMOTE_DOC_MAX_BYTES
+        )
+        size = info.get('size')
+        if isinstance(size, int) and size > max_bytes:
+            raise HTTPException(
+                status_code=413,
+                detail=(
+                    f'远端文件过大，当前在线文档转换上限为 '
+                    f'{_format_bytes(max_bytes)}，请先在 Jupyter 侧裁剪或抽样后再转换。'
+                ),
+            )
+        response, stream_filename = runtime.open_file_stream(remote_path)
     except JupyterRuntimeError as exc:
         raise HTTPException(status_code=400, detail=str(exc)) from exc
+    filename = filename or _safe_uploaded_filename(stream_filename)
     target_dir = (
         state.account_paths(account_id)['base']
         / 'integrations'
@@ -3061,8 +3091,27 @@ def _materialize_jupyter_file_for_feishu(
         / session_id
     )
     target_dir.mkdir(parents=True, exist_ok=True)
-    target = target_dir / _safe_uploaded_filename(filename)
-    target.write_bytes(data)
+    target = target_dir / filename
+    written = 0
+    try:
+        with target.open('wb') as handle:
+            for chunk in response.iter_content(chunk_size=JUPYTER_STREAM_CHUNK_BYTES):
+                if not chunk:
+                    continue
+                written += len(chunk)
+                if written > max_bytes:
+                    handle.close()
+                    target.unlink(missing_ok=True)
+                    raise HTTPException(
+                        status_code=413,
+                        detail=(
+                            f'远端文件过大，当前在线文档转换上限为 '
+                            f'{_format_bytes(max_bytes)}，请先在 Jupyter 侧裁剪或抽样后再转换。'
+                        ),
+                    )
+                handle.write(chunk)
+    finally:
+        response.close()
     return target
 
 
@@ -3080,6 +3129,16 @@ def _safe_uploaded_filename(filename: str) -> str:
     return cleaned[:180] or 'remote-file'
 
 
+def _format_bytes(value: int) -> str:
+    if value >= 1024 * 1024 * 1024:
+        return f'{value / (1024 * 1024 * 1024):.1f}GB'
+    if value >= 1024 * 1024:
+        return f'{value / (1024 * 1024):.0f}MB'
+    if value >= 1024:
+        return f'{value / 1024:.0f}KB'
+    return f'{value}B'
+
+
 def _resolve_account_file(state: AgentState, account_id: str | None, raw_path: str) -> Path:
     path = Path(raw_path).expanduser().resolve()
     account_base = state.account_paths(account_id)['base'].resolve()