Skip to content

Roadmap

Active areas of work and the longer-running pieces that have a design proposal but not yet a shipped implementation. This page is updated as items move from "planned" to "shipping" to "shipped".

If a feature you care about is not on this list, the right place to ask is a GitHub discussion or issue. Roadmap order reflects design readiness and community demand, not strict priority.

Shipped in v0.9.0

The first non-patch release in the v0.8 line. Everything in this section is released and available. Full detail in the v0.9.0 release notes.

Recording and capture

  • Multi-platform system audio capture. Platform detection on click with a per-OS help guide (macOS BlackHole + Multi-Output Device, Windows "Share system audio", Linux pavucontrol + a one-line pactl load-module module-virtual-source). An honest capability matrix flags when full system audio isn't expected to work on the current browser / OS.
  • Multi-input recording. A new Input devices picker lets you choose a primary microphone AND an optional "Also mix in" secondary device; Speakr mixes both streams via Web Audio into one track — the canonical way to record your voice plus a meeting's remote participants where the browser can't capture full system audio natively. Includes a toggle to disable Chrome's echo cancellation / noise suppression / auto-gain and virtual-audio-device discovery.
  • Server-side recording sessions (issue #287 c/d). Long browser recordings stream chunks to the server during recording instead of holding everything in memory. The size-based cap is replaced by a configurable hours-based ceiling, and a page refresh prompts you to resume / finalize whatever was already uploaded. RECORDING_SESSION_MAX_BYTES_PER_USER is a per-user soft limit. Full setup, env-var reference, reverse-proxy guidance, and on-disk layout in Recording Sessions.
  • Failed-upload safety net. When an upload fails, the audio blob is persisted to IndexedDB and offered as a browser download as a defense-in-depth fallback, so the recording never silently disappears (issues #297, #287).

Interface

  • Upload-modal redesign. The upload view is a real modal overlay (not a full-screen takeover) with progressive disclosure of Options behind a chip summary, inline file preview with duration probe, a sticky-footer Upload action, last-used tag / folder / language auto-restore with clearable chips, and a mobile bottom-sheet with drag-to-dismiss.
  • Mobile UI rebuild. The detail view on mobile is now a first-class member of the design system: 56 px bottom navigation, contextual icons in the chevron row, edge-to-edge content, sticky speaker pills, a sticky editor Cancel / Save footer, and audio-player polish.
  • Stats tab. A per-recording tab showing total length, speaker count, turns, and word count as headline cards, plus a per-speaker time / % / turns / words / WPM breakdown and a silence row. Available on desktop and mobile.
  • Inquire ?upload=1 deep-link. The "+ New Recording" button in inquire mode now opens the main app's upload modal directly via a query param instead of dropping you on the empty list.
  • Audio-player position preference. A new Display tab in account settings lets you choose whether the desktop audio player sits at the bottom (default) or top of the recording detail surface.
  • Design-system unification. 22 bespoke modals moved onto shared .modal-* primitives; .btn + .field primitives across the app; native <select> dropdowns themed for dark mode; header consolidation; sidebar redesign; floating dockable chat panel.

Backend and API

  • Webhooks (issue #275). HMAC-SHA256-signed outbound notifications on recording lifecycle events. Each user manages their own webhook endpoints from Account settings → Webhooks (or programmatically via /api/v1/webhooks), with an SSRF guard against private IPs and exponential-backoff retries that auto-pause after repeated failures. Event types, signature-verification examples in Python / Node / bash, retry schedule, and env-var reference in Webhooks.
  • GET /api/v1/users/me (issue #281). Companion apps and automation flows can identify the current user — and their group memberships — without scraping internal endpoints.
  • PWA Web Share Target (issue #285). Install the PWA and pick Speakr from your phone's native share sheet to send a recording straight in.
  • Auto speaker labelling honoring the user threshold. Automatic speaker labelling now respects the user-configured confidence threshold rather than a hard-coded default.

Known follow-up — debounce recording.updated. Rapid edits (notes autosave, retitling, tag changes) currently emit one recording.updated event per mutation. A 30s per-recording debounce window was in the original design and is planned for a later release. Receivers that want to deduplicate today can group on (recording_id, fields_changed) within a short window.

Open ideas (not yet designed)

Feature requests that are on the radar but have not been worked through in detail yet:

  • Watch-folder targeting. Map a watch directory to a specific recording folder so files picked up by that folder are auto-routed (discussion #276).
  • Read-only watch folders. Process audio files in place without moving them, useful for preservation projects and externally-managed storage (discussion #277).
  • Worker priority / fallback routing for hybrid transcription setups. Schema changes required for proper worker identity tracking (issue #255).
  • File attachments on recordings. Attach slides, related docs, or other supporting material to a recording (issue #174).

How to influence priority

  • Open a feature request issue or a discussion describing the use case and the workflow it would enable. The richer the use case, the easier it is to design for.
  • Reactions on existing issues / discussions help signal demand.
  • For larger features, a design proposal in a discussion that you have thought through is the fastest path. Several items now shipped in v0.9.0 — webhooks and server-side recording sessions among them — reached design-and-build because someone cared enough to articulate what they needed.

How releases are versioned

Speakr is in alpha. The version scheme is v0.MINOR.PATCH, with releases through v0.8.x carrying an -alpha suffix:

  • MINOR increments for larger feature batches. v0.9.0 was the first such bump in the v0.8 line — a coherent baseline rolling up the recording, mobile, and design-system work rather than another patch.
  • PATCH increments for bug fixes, security patches, and small targeted features on top of the current minor.
  • Security patches always ship as their own release, separate from feature work, so the security advisory record stays narrowly scoped.