Voice accessibility: wire ibis phone workflows through linnet #3

Open
opened 2026-04-06 10:48:37 -07:00 by pyr0ball · 0 comments
Owner

Dependency note: linnet voice accessibility layer

When ibis reaches active development, consider wiring its phone-based workflows through linnet (Circuit-Forge/linnet) rather than building a standalone voice layer.

linnet is being designed as shared voice accessibility infrastructure:

  • Notation: real-time tone annotation (Elcor-style) — makes implicit affect explicit for ND/autistic users
  • Navigation: IVR proxy — user types intent, linnet navigates and speaks (nonverbal access)
  • Interaction: full bidirectional voice proxy — user types in their language, linnet speaks; caller speaks, linnet transcribes + annotates
  • Translation: DeepL + LLM layer — real-time translation with contextual flagging of legal/bureaucratic subtext that doesn't survive literal translation; mobile-first PWA

For healthcare, the translation milestone is especially relevant — non-English speakers navigating healthcare phone systems in a second language are a core CF audience.

Why linnet, not inline

These capabilities share an engine (cf_voice.context, cf_voice.telephony, cf_voice.io in circuitforge-core#34). Building them inline per-product creates drift and maintenance burden. linnet owns the voice proxy surface; products call its API.

References

## Dependency note: linnet voice accessibility layer When ibis reaches active development, consider wiring its phone-based workflows through **linnet** (`Circuit-Forge/linnet`) rather than building a standalone voice layer. linnet is being designed as shared voice accessibility infrastructure: - **Notation**: real-time tone annotation (Elcor-style) — makes implicit affect explicit for ND/autistic users - **Navigation**: IVR proxy — user types intent, linnet navigates and speaks (nonverbal access) - **Interaction**: full bidirectional voice proxy — user types in their language, linnet speaks; caller speaks, linnet transcribes + annotates - **Translation**: DeepL + LLM layer — real-time translation with contextual flagging of legal/bureaucratic subtext that doesn't survive literal translation; mobile-first PWA For healthcare, the translation milestone is especially relevant — non-English speakers navigating healthcare phone systems in a second language are a core CF audience. ## Why linnet, not inline These capabilities share an engine (`cf_voice.context`, `cf_voice.telephony`, `cf_voice.io` in circuitforge-core#34). Building them inline per-product creates drift and maintenance burden. linnet owns the voice proxy surface; products call its API. ## References - Circuit-Forge/linnet (placeholder) - Circuit-Forge/roadmap#31 — linnet tracking issue - Circuit-Forge/circuitforge-core#34 — cf_voice module - cf-orch#20 — voice context service endpoint
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: Circuit-Forge/ibis#3
No description provided.