What's New

October 2026

Send each finished session to your own bucket: transcript, events, a meta.json with a SHA-256 hash for every file, and the call recording for voice sessions
Connect Amazon S3, S3-compatible storage (MinIO, Wasabi and others), Cloudflare R2, Google Cloud Storage, Azure Blob or Supabase Storage under Settings → Storage; test a connection before saving and set the delivery order
Choose what is saved per agent under Advanced → Storage, and see delivery status or press Retry export on any session in Observe
Zero-data-retention workspaces deliver the transcript and events straight to your bucket and keep no copy in Taliiq; calls are not recorded
Rolling out to workspaces gradually
Guide: Object storage

Each squad now has Settings: pick one agent to share its voice pipeline (or its LLM on process squads), one agent whose storage choices apply to the whole call, and one agent whose Server URL receives the call's events
Instructions, tools and knowledge always stay on each agent
A consistency banner flags squads whose agents disagree (different voices, buckets or webhooks) and takes you to the right picker
Shared models and single-receiver webhooks are rolling out to workspaces gradually; storage follows Object storage
Guide: Squad settings

An agent with only the API capability (no chat, no voice) no longer has a Server URL; its runs report through your organization webhooks
The Advanced tab explains this and links to the guide; stored Server URLs are kept
Guide: Webhooks

New Taliiq Private SIP Link integration: a dedicated, encrypted and isolated connection between your phone system and Taliiq, so call signaling and audio never travel the open internet
No public SIP or RTP ports on your side; each link is scoped to your organization and can reach Taliiq SIP only
Create a link, download the device config, and attach it on BYO SIP channels or outbound trunks
Suspend, resume, rotate keys (with a fresh config download), and track usage on Billing → Plan
Guide: Private SIP Link

New built-in Classification tool: agents get typed, confidence-scored answers about a conversation (urgency, intent, qualification, opt-out) from a small classifier model
Pick a classifier model from a dedicated Classifier Providers catalog, using Taliiq-funded usage or Your key
Built-in Test panel to try questions on pasted text or a recent session before attaching the tool
Each check shows inside the conversation, in a new Classification tab in session logs, and as a Classifier member of the session cost (charged with the session on Taliiq-funded usage, $0 on Your key)
Guide: Classification

Paid plans now include usage equal to the monthly fee, spent before your top-ups and expiring at the end of each billing period
One Available balance in the sidebar and Billing page, with included usage and top-ups shown separately
Guide: Billing and plans

Multi-agent workflows: conversation (voice handoffs) and process (API invoke)
Per-agent tools; canvas handoff edges with context and variable extraction
Public API: /api/v1/squads/, graph read, POST /api/v1/invoke/, calls with squad_id
Guides: Squads · API: Squads

Added gpt-live-1, OpenAI's full-duplex realtime model, as a voice provider — select it for an agent alongside its supported voices and languages
Also added gpt-6-astra and gpt-realtime-2.1-mini to the model catalog
Refreshed OpenAI model pricing across the catalog to match current published rates