Fix Gemini Live 1007 CONTENT_TYPE_AUDIO crash + tune pipeline latency

Fix voice override (Gemini Live vs Bulbul), add OpenAI STT/TTS support, and enable intelligent post-call automation from AI analysis

feat(analytics): make NVIDIA NIM primary and Groq fallback per user config

fix: email only for meeting+materials, no recap always-send

feat: email always-send + Groq-primary analytics + TTS on realtime agents

feat: Live Call Command Center + all Gemini Live models + goodbye-aware hangup + latency tuning

fix: transcript via session.history.items (1.x API), email/calendar automation in async analysis path, sheet CRM blank-clobber + double-count fix, CallLogs detail fetches full record

fix(analytics): implement server-side async analysis endpoint with 5-attempt retry loop to guarantee success

fix: voice agent improvements (greeting priority, gemini 3.1 compatibility, valid Groq analytics model fallback, lead webhook upsert and duplicate log fixes)

feat(ScatterAI): chat-to-build agents — describe what you want, NVIDIA NIM designs the full agent (persona/role/voice/greeting) and one-click creates it

chore: silence Firestore positional-filter UserWarnings — convert all 33 .where() calls to .where(filter=FieldFilter(...))

platform polish: real notification bell (hot leads/angry calls/actions), Actions page, agent Sheet CRM viewer, strong KB-grounded RAG, Gemini Live native voices only, brand logos, fixed sticky scrollable sidebar, Settings integrations shortcut

voice: auto Google Sheet CRM (create+lookup+remember callers), email sends meeting invite+materials not summary, silence busy-guard, dedupe double test-call

gemini live: add /api/gemini/live-models probe + current default model name (retired gemini-2.0-flash-live-001 caused 1008 not-found)

Gemini Live: realtime-correct (no STT/TTS, native voices, generate_reply greet) + UI; add public OAuth /integrations/diagnostic to debug 401 invalid_client

ScatterStudio: Jarvis-parity voice stack + integrations + HF/Vercel deploy + speed fixes

error resolving

fix voice agent full-duplex, call logs, KB CORS, dashboard speed, mobile UI

fix import error

Solved some errors

fix: use full groq model ids and update frontend

fix: update decommissioned LLM model to llama-3.3-70b-versatile

Explicitly dispatch LiveKit agent

Run Render Docker deploy as API only

Split LiveKit agent deployment to Hugging Face

Fix all broken features: Vobiz dialing, RAG, real LLM, KB upload, billing, settings

Use bulbul:v3 with 37 verified compatible voices

Add debug TTS endpoint and better error messages

Fix default voice_id to valid shreya

Add all 44 Sarvam voices

Wire real Sarvam voices, TTS preview, settings profile save, fix CORS

Support FIREBASE_JSON environment variable

ScatterStudio: Initial commit with LiveKit & Firebase integrations