Usually SuprFlow keeps a usable local result. When Cloud API or Ollama formatting returns an error, the app falls back to the locally cleaned transcript and uses that text for insertion and history. This is not a guarantee that every destination accepted the paste, nor proof that a failed hosted request sent nothing. Confirm the result in the intended field and in the app's Transcriptions view before speaking the paragraph again.
The fallback can look less polished than expected. It may have punctuation and filler cleanup but no requested structure from the optional AI model. SuprFlow 1.0.0 does not show a prominent formatter-failure dialog at that point, so a user may notice only that the inserted text is plainer than usual. Test this in the signed version before relying on silent recovery for important notes.
Find which stage failed
| What you see | Likely stage to check | First action |
|---|---|---|
| No words appear anywhere | Microphone, model, transcription, or insertion may have failed. | Try a short fictional sentence in a blank local note; use the paste troubleshooting guide. |
| Words appear, but headings or lists are missing | Optional formatter may have failed or a guardrail may have rejected its output. | Compare the inserted result with local history and check the selected formatter. |
| Words appear in history but not the intended app | Insertion or focus is the immediate problem. | Check the active window and clipboard behavior; do not re-dictate into an uncertain field. |
| Words appear in the wrong record | Destination routing is the problem, regardless of whether formatting succeeded. | Follow the wrong-window recovery checklist before proceeding. |
SuprFlow also rejects some AI outputs that contain prompt artifacts or are at least 1.3 times the length of its pre-cleaned input. That length guard is not a semantic guarantee: a short rewritten sentence can still change a number or negation. The app may use local cleanup after either an API error or a guardrail rejection. The user should inspect the final words, not infer the outcome from the presence of text alone.
Recover without sending the same sensitive note twice
- Stop and inspect the destination field. If text is present, preserve it in the approved draft location and compare it with what you said. Check names, amounts, dates, negations, and action owners.
- In Transcriptions, look for the recent entry. History can retain raw, cleaned, and formatted versions. Treat this as a recovery aid, not the official client or audit record.
- Check Formatting → Provider. Default uses local cleanup. Cloud API needs the configured provider, model, API key, and reachable endpoint. Local (Ollama) needs the configured server and model; its endpoint may point away from this Mac.
- Test the provider with a fictional sentence before retrying a real note. A missing key or invalid URL can fail before transmission; a non-200 provider response can occur after transcript text reached the server. A local fallback does not distinguish these cases.
- If the failure repeats, report the signed app version, macOS version, provider category, model, endpoint host where appropriate, and a synthetic reproduction. Do not send the API key, client note, raw diagnostic archive, or real transcript to support.
SuprFlow 1.0.0 forwards some formatter error descriptions to release telemetry and prints an error description locally. Avoid using a sensitive paragraph to diagnose an unknown hosted endpoint.
Questions people ask
If I see plain text, did cloud formatting definitely fail?
No. The selected provider may have returned plain text, local cleanup may have been selected, or a fallback may have run. Record the setting and test with fictional content. SuprFlow 1.0.0 does not give an unambiguous per-entry success indicator in this path.
Did a failed Cloud API call still upload my transcript?
It may have. Some failures happen before a request; others happen after the provider receives it. Check the stage, logs or test-account records available to your organization, and the provider's terms.
Should I repeat the dictation immediately?
First check the destination and history. Repeating can create duplicate records or send the same text to a provider twice without solving the formatter problem.
Related guides
- Where does my dictation go?
- Does cloud formatting send audio or text?
- Choose local or cloud formatting
- What SuprFlow history contains
- Report a bug without client data