Self-hosted voice-memo pipeline: local ASR, SQLite storage, and configurable LLM presets that summarize and extract structured JSON — with tool-calling planned next.
View on GitHubManually wake the horizontal pipeline. Processes pending, in-progress, and due retry files through ASR and dynamic LLM stages for the current frozen run.
Drag and drop files or folders here, or:
No entries yet.
ASR backend, device, VAD/denoise, and queue drain. Changes apply on the next transcription (device/compute type reload the model).
One OpenAI-compatible server and model for every LLM step. Changes apply to future runs only.
One active preset defines the ordered LLM chain after ASR. Membership edits publish a new
immutable revision. Stable occurrence identity (slot_key) is shown only under
diagnostics.
Builtin presets are read-only. Clone to create an editable custom copy.
Add, remove, and reorder steps. Bind each occurrence to transcription, the previous step, or an earlier slot.
slot-summarizesummarize_sourceslot-keypointskeypoints_jsonShared definitions used by preset membership. Editing a custom step publishes a new revision for every preset that contains it; those updates apply to future runs only. Builtin steps are read-only — use Clone to copy them into an editable custom step.