nia-todoDocs
Menu

nia-todo documentation

BrainDump

Turn voice notes into reviewed todo candidates using your own speech-to-text and language-model providers.

BrainDump is an optional workflow: record or enter a thought, transcribe it, extract structured todo candidates and review them before anything is created. It is disabled by default.

What you need

BrainDump uses two provider stages:

  1. Speech to text (STT): A local or remote whisper.cpp-style transcription endpoint.
  2. Language model (LLM): An OpenAI-compatible endpoint or Ollama.

Supported LLM setups include local services such as LM Studio or Ollama, Ollama Cloud and OpenAI-compatible gateways. Provider compatibility depends on the endpoint implementing the APIs expected by the selected mode.

Configure BrainDump

In /admin:

  1. Configure the STT provider URL, optional token, language and timeout.
  2. Configure the LLM provider, base URL, model, optional API key and timeout.
  3. Use the built-in connection tests.
  4. Enable BrainDump globally.
  5. Grant access only to the users who need it.

The default system prompt is the safest starting point. A custom prompt can change extraction behavior and should be tested with ambiguous dates, multiple todos and notes that should not become todos.

Network placement

Provider URLs are requested by the nia-todo server, not necessarily by the user’s browser. A local provider may therefore use an internal hostname reachable from the container or service host.

For Docker, remember that 127.0.0.1 refers to the nia-todo container itself. Use a shared Docker network, a resolvable service name or an explicitly routed host address.

Privacy and limits

Audio and extracted text are sent to the configured providers. Choose local providers if the content must stay on your own infrastructure. If you use cloud services, review their retention and billing policies.

Set realistic timeouts: speech recognition and larger models can need longer than ordinary API requests. Test with the same audio duration and language your users will actually submit.

BrainDump is available where the client platform grants microphone access, including supported native apps and the desktop web app.