Memory Glass
fcs models offline · edge · voice STT

Suite AI plane

Offline first.
Edge when you need bulk.

fcs models unifies Ollama (local), Cloudflare Workers AI via lang.ugrad.ai, and local whisper for fcs mic. One doctor for lab + agents.

qwen3:8b offline m2m100 edge whisper STT P0 NLLB + m2m100 local 
fcs models doctor
fcs models offline
fcs models edge
fcs models pull qwen3:8b
fcs models p0
fcs models matrix
lang models board Offline pack page mt-models.json Open in hub tree

Offline · Ollama

Default qwen3:8b for MT/desk when network is gone or quota-free. fcs models pull … · ollama run qwen3:8b.

Edge · Cloudflare

GET /api/models · POST /api/ai on lang.ugrad.ai → m2m100 bulk NMT + llama instruct.

Voice STT · whisper

Local whisper-cli + ggml under ~/models/audio/whisper/ggml/. Powers fcs mic and fcs compose transcribe.

P0 pack

fcs models p0 pulls NLLB-200 + local m2m100 weights (lab path under qbitOS models). Status: fcs models p0 status.

Pair with

/language multi-lang desks · /mic + /chat voice loop · /compose listen agent STT · Memory Glass language surfaces.