fix: raise token limit to 2B and surface server validation errors

- Raise MAX_TOKEN_VALUE from 100M to 2B per entry.
  Large Copilot CLI agent sessions can easily exceed 100M tokens in
  a single day/model/workspace combination. Entries exceeding the old
  limit were silently rejected, causing ~40% of data to be dropped.

- Make SharingServerUploadService read the actual server-side 'uploaded'
  count from the response body instead of assuming all sent entries were
  accepted. Logs a warning when the server rejects entries (validation
  errors), making silent data loss visible.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: rajbos

sharing-server/src/routes/api.ts

@@ -11,7 +11,7 @@ const MAX_STRING_LENGTHS = {
 	datasetId: 128,
 	editor: 100,
 };
-const MAX_TOKEN_VALUE = 100_000_000; // 100M tokens per entry is already absurd
+const MAX_TOKEN_VALUE = 2_000_000_000; // 2B tokens — large agent sessions can exceed 100M in one day
 const MAX_ENTRIES_PER_UPLOAD = 500;
 
 export const api = new Hono<{ Variables: AuthVariables }>();

vscode-extension/src/backend/services/sharingServerUploadService.ts

@@ -50,7 +50,17 @@ export class SharingServerUploadService {
 					warn(`Sharing server upload: ${message}`);
 					return { success: false, entriesUploaded: totalUploaded, message };
 				}
-				totalUploaded += batch.length;
+
+				// Read actual uploaded count — server may reject entries that fail validation
+				let serverUploaded = batch.length;
+				try {
+					const result = await response.json() as { uploaded?: number; errors?: string[] };
+					if (typeof result.uploaded === 'number') { serverUploaded = result.uploaded; }
+					if (result.errors && result.errors.length > 0) {
+						warn(`Sharing server upload: server rejected ${batch.length - serverUploaded} entries: ${result.errors.slice(0, 3).join('; ')}`);
+					}
+				} catch { /* response not JSON, use batch.length */ }
+				totalUploaded += serverUploaded;
 			}
 
 			const message = `Uploaded ${totalUploaded} entries`;