needle-cloud 3.7.0-rc.1794373394.9a2fd12 → 3.7.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -159,6 +159,31 @@ npx needle-cloud settings remote-ai-tools false
159
159
  See [MCP tools and setup](https://engine.needle.tools/docs/ai/needle-mcp-server)
160
160
  and [local tool controls](https://engine.needle.tools/docs/cloud/cli#local-tools-in-needle-cloud-ai).
161
161
 
162
+ ## Automatic error reports
163
+
164
+ Automatic diagnostic sharing is **off by default**. To opt in on this machine:
165
+
166
+ ```bash
167
+ npx needle-cloud settings automatic-diagnostics true
168
+ ```
169
+
170
+ Recovered CLI errors are queued for background delivery after about five seconds.
171
+ Fatal errors are saved synchronously and sent on a later CLI start. Reports include
172
+ the redacted error/stack, version, OS, uptime and up to 20 recent CLI log lines;
173
+ they do not upload complete CLI/editor logs. Delivery uses an existing production
174
+ login and never opens a sign-in prompt. Development-auth sessions are not sent to
175
+ production Haystack.
176
+
177
+ The queue holds at most 8 reports, 128 KB each, and uses at most 1 MB including
178
+ atomic replacement files. Duplicates increment a counter. Reports expire after
179
+ 7 days on the next queue check. At most 10 upload attempts are made per rolling
180
+ 24 hours, including failed attempts; failures back off and successful duplicates
181
+ are suppressed for an hour. Reporting failures never create more reports.
182
+
183
+ Disable sharing with `npx needle-cloud settings automatic-diagnostics false`.
184
+ Pending payloads are removed on the next queue check. `send-logs` remains the
185
+ explicit command for sharing a full, redacted support bundle.
186
+
162
187
  ## Browser API
163
188
 
164
189
  ### Import a model URL
@@ -188,3 +213,36 @@ processing. `reused: true` means an identical import is already queued or runnin
188
213
  - [CLI reference](https://engine.needle.tools/docs/cloud/cli): command options, output formats, versions, labels, local tools, and browser SDK notes.
189
214
  - [Cloud documentation](https://engine.needle.tools/docs/cloud/): hosting, asset optimization, teams, and workflows.
190
215
  - [Changelog](./CHANGELOG.md) ? [Pricing](https://needle.tools/pricing)
216
+
217
+ ## Upload workflow selection (development/staging preview)
218
+
219
+ Uploads use the default workflow when `--workflows` is omitted. To skip automatic
220
+ processing, use `--no-workflows`. Select a saved workflow by ID or an exact,
221
+ unambiguous team workflow name:
222
+
223
+ ```sh
224
+ needle-cloud upload ./model.glb --workflows '{"id":"workflow-id"}'
225
+ needle-cloud upload https://example.com/model.glb --workflows '{"name":"Product"}'
226
+ needle-cloud upload ./model.glb --workflows '[{"id":"one"},{"name":"World"}]'
227
+ needle-cloud upload ./model.glb --no-workflows
228
+ ```
229
+
230
+ The API resolves each reference to its ID and saved revision when accepting the
231
+ upload. Later edits do not change that selection. `default` is an ordinary name.
232
+ Several selected workflows run independently on the source version. The same
233
+ `workflows` JSON value is accepted by the SDK upload options and URL import request.
234
+ `workflows: false` skips processing; the existing `optimizations: false` remains
235
+ supported. Conflicting choices are rejected.
236
+
237
+ Explicit selection is available only in development/staging. `--wait` (also used
238
+ by `optimize`) waits for the accepted workflow outputs without requesting another
239
+ optimization. One job-status request per poll covers the source import/upload and
240
+ all outputs. Failures, cancellation and stalled processing return an error; a
241
+ 30-minute timeout stops waiting while processing may continue on Cloud.
242
+
243
+ With `--outdir`, one output downloads directly into that directory; multiple outputs
244
+ use separate `output-<job-id>` subdirectories. JSON receipts include every output's
245
+ workflow ID, revision, output ID, job ID and URL. Servers without workflow processing
246
+ retain the existing single-optimization wait behavior. `--progressive` and `--usecase`
247
+ apply only to that legacy path. Saved-workflow waits reject those overrides; edit
248
+ the saved workflow to change its optimization settings.