Product: split translation into in-group vs Language Threads

[daopunk]

Goal

Treat translation as two products, not one:

Product	Audience	Behavior
In-group translation	Smaller / V1 use case	Translate (or dual-post) inside a single Signal group — the original product shape (!translate-on / per-user-in-group, etc.)
Language Threads	High-functioning large mutual-aid groups	Parallel sidecar groups per language, bridged to a bilingual main chat

This branch / PR implements Language Threads. In-group translation should be a separate product that can ship and deploy on its own timeline.

Why

• Language Threads optimize for scale and monolingual lanes (many members, many langs, less dual-posting in main).
• In-group translation optimizes for one room, simpler mental model, V1 behavior organizers already know.
• Shipping both in one bot binary/CVM blurs product boundaries and ops (memory, feature flags, support).

Scope

• Document and name the two products clearly (user-facing + ops).
• Package / configure in-group translation so it can run without Language Thread sidecar APIs.
• Package / configure Language Threads so it can run without requiring in-group !translate-on as the primary path.
• Separate deploy stories (compose/env/CVM) where it helps; shared Signal CLI still allowed if contracts stay clean.
• Re-enable or restore in-group commands in the in-group product (currently unregistered on the Language Threads alpha surface).

Interoperability

• Transcription (separate product / CVM) should work with either translation product — see the transcription issue.
• The two translation products do not need to be mutually composable in one chat (pick one translation mode per deployment / group). They should both compose with transcription.

Related

• Language Threads implementation + docs: feat: Language Threads — mutual-aid language sidecars #7
• Overview: docs/language-threads.md on that branch
• Transcription extraction issue: (link after creation)

Acceptance hints

• Clear README / docs per product
• An operator can deploy Language Threads without enabling in-group auto-translate, and vice versa
• Transcription CVM can be pointed at either product via documented config

[daopunk]

Transcription product (separate CVM): #8