DEV Community: Andres Sarrazola

I built Dictámelo: local voice typing with Rust, Tauri and optional cloud models

Andres Sarrazola — Tue, 08 Sep 2026 18:08:56 +0000

I developed Dictámelo, an open-source dictation app for macOS and Windows

It follows a small loop: hold a shortcut, speak, release, and the transcript appears at your cursor

The application is MIT licensed, including the Rust backend and Tauri interface

Choosing where transcription runs

The local catalog includes Whisper Tiny, Base, Small and Large v3, Parakeet v3 and Canary 180M Flash

Download a model once and local transcription works without an account, API key or internet connection, with no hosted time allowance

Cloud is a separate choice, either your own provider key or Dictámelo Cloud

That distinction also applies to cleanup: local speech recognition does not make every optional feature local, and AI text cleanup sends the transcript to a cloud cleaner only with consent

The desktop details

The recording shortcut is configurable, Escape cancels a recording, and the app keeps a local history so a failed paste does not mean losing the transcript

It can transcribe audio files and use a personal vocabulary for names and terms

The native code handles recording, model execution and platform integration, while Tauri hosts the settings and onboarding interface

For contributors, the useful starting points are src-tauri/src/pipeline.rs, the transcription adapters, and docs/LOCAL_MODELS.md

Trying it

Download the installer for Apple Silicon macOS or Windows x64/ARM64
Choose local models in setup
Download a model and select Use model
Hold Alt/Option + Shift + Space, speak, and release in a text field

Local use is free, personal cloud providers have their own rates and limits, and the hosted service has optional Free and Pro plans

I see it as an open-source alternative to Wispr Flow for people who want a simple dictation workflow and control over where speech is processed

Model speed and language coverage depend on the model and hardware, so I would especially value reports about first-run setup and everyday dictation in different languages

Source code and downloads

The app is MIT licensed, while third-party model weights retain their own licenses, documented in the repository

OpenLivery: agentes de WhatsApp para varios clientes en una instalación propia

Andres Sarrazola — Mon, 07 Sep 2026 18:34:53 +0000

Soy Andrés, el desarrollador de OpenLivery. Es una plataforma de código abierto, con licencia MIT, para agencias que gestionan agentes de WhatsApp para varios clientes.

La unidad de trabajo es el cliente: cada uno tiene su espacio, conversaciones, base de conocimiento y portal con su marca. La agencia administra esos espacios desde una instalación. También se puede pasar una conversación a una persona.

Qué incluye el repositorio

El frontend está en Next.js y el backend en FastAPI, con PostgreSQL, migraciones y Docker Compose. Se pueden conectar proveedores de modelos propios y ampliar los agentes con herramientas HTTP o servidores MCP.

Código: https://github.com/sarrazola/openlivery

Puesta en marcha

Con Git, Docker Compose y make instalados:

git clone https://github.com/sarrazola/openlivery.git
cd openlivery
./scripts/generate-docker-env.sh
make up

El generador de entorno guía la configuración inicial. Sigue el README actualizado para conectar el modelo y el canal de mensajería. En local, la interfaz utiliza el puerto 3000 y la API el 8000. El código es libre; el servidor y el consumo de los proveedores tienen sus propios costes.

Cómo probar el flujo de una agencia

Crea dos espacios de prueba con contenidos distintos. Conversa con cada agente y revisa las respuestas, el historial y el portal de cada cliente. Prueba también el traspaso a una persona. Así puedes evaluar el funcionamiento con varios clientes antes de incorporar datos reales.

Si necesitas integrar otra aplicación, empieza con una operación HTTP concreta o un servidor MCP y comprueba qué ocurre tanto cuando responde como cuando falla.

La documentación incluye español. Los problemas de instalación y las sugerencias se pueden reportar en el repositorio.

Web: https://www.openlivery.com/

Esta es una adaptación en español de mi presentación de OpenLivery en DEV. El artículo fue redactado con IA.

OpenLivery: self-hosting multi-tenant WhatsApp AI agents for agencies

Andres Sarrazola — Mon, 07 Sep 2026 17:48:54 +0000

I’m the developer of OpenLivery, an MIT-licensed platform for agencies managing WhatsApp AI agents across multiple clients.

The core workflow is one workspace per client, managed from one installation. That matters when each client needs their own knowledge base, conversations and branded portal.

What is included

The repository contains the application, rather than just an agent prompt or workflow template:

A Next.js frontend and a FastAPI backend.
PostgreSQL, database migrations and Docker Compose deployment.
Client workspaces, knowledge bases, branded portals and human handoff.
WhatsApp Cloud API, a QR-based WhatsApp bridge and a web chat widget.
OpenAI, Anthropic and OpenAI-compatible model providers.
Per-agent HTTP tools and MCP servers for extending what agents can do.

Start a local installation

With Git, Docker Compose and make available, the repository’s quick start is:

git clone https://github.com/sarrazola/openlivery.git
cd openlivery
./scripts/generate-docker-env.sh
make up

Follow the prompts from the environment generator and the current README for configuration. The local web interface runs on port 3000 and the API on port 8000.

Starting the containers is the first step. You still need to configure your model provider and connect the messaging channel you intend to use. Self-hosting does not remove infrastructure, model or messaging-provider costs.

Try the agency workflow

A useful evaluation is to create two test client workspaces, give each different knowledge content, and run separate conversations. Then try human takeover and the client portal for each workspace. This makes the client-management workflow easier to assess than a single chatbot demo. Use test data while evaluating the setup.

For tools, start with one narrow operation that the client actually needs. Configure its HTTP integration or MCP server for that agent, then test success and failure responses before adding more capabilities.

Why publish the full stack?

Agencies should be able to inspect and adapt the platform they run for clients. The MIT license allows them to self-host and modify the code, including their branding and integrations. English and Spanish documentation is included in the repository.

This is an early project, and feedback on installation and client onboarding is especially useful. If you run agents for several clients, which part is hardest today: setup, knowledge maintenance, integrations or handing conversations to a person?

Source code and installation instructions · Website

Disclosure: AI tools supported development and helped draft this article.