feat(web): info icon on every param with hover/click tooltip; add help text to all params

refactor: remove dead ZeroGPU shim (Docker Spaces don't support ZeroGPU)

fix(progress): run sync adapter.generate in a thread so SSE events stream during generation

feat(progress): wrap /api/generate/dialog in session; emit per-turn events

feat(progress): /api/progress SSE endpoint; wrap /api/generate in session

feat(progress): ProgressBus with sessions, ticks, and turn-complete events

feat(dialog): /api/generate/dialog endpoint + per-turn dispatcher with seed reuse

feat(dialog): parse_dialog with SPEAKER A-D regex and edge-case handling

feat(models): expand chatterbox-mtl params (seed, repetition_penalty, min_p, top_p)

feat(models): expand chatterbox-turbo params (seed, top_k, exaggeration, cfg_weight, etc.)

feat(models): expand chatterbox-en params (seed, repetition_penalty, min_p, top_p)

feat(api): adapter generate returns seed_used; expose X-Seed-Used header

feat(seed): apply_seed helper that returns the seed actually used

feat(schemas): add group field to ParamSpec (basic/advanced)

chore(api,web): include validation message and detail in /api/generate error response

fix: re-encode reference voices as PCM WAV in browser; expand turbo tag list to 9

feat(models): chatterbox-turbo and chatterbox-mtl adapters

feat(api,models): FastAPI app with /api/health, /api/models, activate, /api/generate + chatterbox-en adapter

feat(registry): active-model swap with async lock and SSE event bus

feat(models): ModelAdapter protocol + is_valid_adapter check

feat(zerogpu): decorate() shim for HF ZeroGPU compatibility

feat(schemas): pydantic models for ParamSpec/ModelInfo/Health/Errors

feat(audio): wav validation, write helper, mono/16k normalization

feat(device): auto-detect cuda/mps/cpu with env override

chore: scaffold python project (requirements, gitignore, pytest, package skeleton)