How to use Ni lo escucho to turn voice notes into n8n drafts
Ni lo escucho uses local Whisper transcription, Gemma through Ollama, and a deterministic compiler to produce n8n workflow drafts and Spanish client questions from informal voice notes.
dev.toSuggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Convert [client voice-note transcript] into a constrained automation specification using [supported trigger and action catalog]. Separate unconditional actions from conditional branches. List missing details and short Spanish follow-up questions. Treat unsupported actions as questions, never invent identifiers, and do not ask clients to send passwords or tokens.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- gartox
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The app produced an importable n8n workflow draft and Spanish follow-up questions. The author reports that compiled flows executed the appropriate branches for 60-person and 20-person inputs after correcting compiler defects.
- Human involvement
- The author executed generated workflows in n8n, fixed observed defects, and retained missing client information as questions in the delivered draft.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Supplied a client's voice note describing a form, sheet storage, thank-you emails, and a conditional WhatsApp alert.
- Transcribed the voice note locally using Whisper.
- Used Gemma through Ollama to fill a constrained workflow specification and identify missing information.
An automation builder serving small businesses in Tabasco, Mexico, built Ni lo escucho to turn rambling WhatsApp voice notes into n8n workflow drafts. Clients describe requests with corrections and missing details, leaving the builder to identify triggers, actions, and conditional behavior.
The app runs on the author's laptop. Whisper transcribes the audio locally, and Gemma 4 E4B through Ollama fills a constrained specification. That specification contains a trigger, catalog-supported steps, conditions, missing information, and questions. A deterministic compiler converts it into importable n8n JSON using tested node templates. The app also returns a Spanish reply containing questions for the client.
The demonstrated request saves submissions to a sheet and sends thank-you emails to everyone, while routing orders above 50 people to a WhatsApp alert. Missing phone and spreadsheet details became questions. The author executed compiled flows with 60 and 20 people and confirmed different branches. Real execution also exposed incorrect condition wiring and an invalid operator, which the author fixed. Additional changes sanitized condition field names and filtered requests for secrets from client replies.
The delivered output remains a draft for import and follow-up questions. Who it's for: Automation builders who receive informal client requests and need a structured starting point.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: dev.to · gartox
Frequently asked questions
Does the language model generate n8n workflow JSON directly?
No. Gemma fills a constrained specification, and a deterministic compiler generates n8n JSON from tested node templates. Requests outside the supported catalog become questions.
How did the author check conditional workflows?
The author executed a compiled flow in n8n with 60 and 20 people. The 60-person input reached the big-order branch, while the 20-person input reached the other branch.
What happens when client details or credentials are missing?
Missing phone and spreadsheet details become questions. The author also changed the reply to filter secret-like requests and tell clients to connect accounts and passwords together in n8n rather than send them through chat.
More details
Remaining steps
- Compiled the specification into importable n8n JSON and produced Spanish questions for the client.
- Executed compiled flows with 60 and 20 people to verify that the condition selected the appropriate branch.
- Fixed condition wiring and operators, sanitized field names, and filtered secret requests from client replies.
About this account
After a button press, the local pipeline transcribes the request, structures it, compiles a workflow draft, and prepares client questions.
Platform: dev.to. Attribution: https://dev.to/gartox/my-clients-explain-automations-in-rambling-whatsapp-voice-notes-now-my-laptop-answers-with-an-n8n-4dj0. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: gartox on dev.to
Source published
Added to Agents In Use
Source checked
Tools and topics
Ni lo escucho · Whisper · Gemma 4 E4B · Ollama · n8n
Voice notes · Workflow drafting · Local AI · n8n · Client requirements
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.