Voicetta
    Voicetta.

    Voicetta Docs

    Voicetta Changelog

    Shipped Voicetta product changes, release history, fixes, and important notes.

    Voicetta Changelog

    This changelog records shipped Voicetta product changes.

    Backlog items are tracked internally until they are ready to ship.

    Latest

    Product Roadmap: Know, Deliver, and Grow available (expanding)

    • Updated the Product Roadmap: Know, Deliver, and Grow are now Available today (expanding) — lasting Guest memory summaries, AI Evaluations, and Optimize.
    • Think remains on the roadmap.

    Optimize

    • Added Optimize: evidence-backed suggestions from recent conversations — prompt improvements, Knowledge Base drafts, and trends.
    • Nothing applies without confirmation; Knowledge drafts still follow the normal review path.
    • Available when enabled for your workspace (contact your Voicetta representative if you do not see it in the menu).

    Phone numbers: buy worldwide, Require SMS, Import Existing Number

    • Buy countries: the buy flow lists countries with live carrier inventory instead of a fixed short allowlist.
    • Require SMS: optional toggle when buying (default off). Turn it on only when the number must support SMS; leave it off to include voice-only inventory (important for markets such as Poland local).
    • Import Existing Number: bring your own Twilio, Vonage, Telnyx, or SIP trunk number without buying from Voicetta inventory. See Import Existing Number and Phone Numbers.
    • Telnyx import: paste your Telnyx API key and number only — Voicetta creates or reuses a LiveKit FQDN connection on your Telnyx account automatically (no connection ID field).
    • Delete imported numbers: removes Voicetta routing only; the DID stays on your carrier.

    Conversation Memory (dedicated page)

    • Moved Guest memory to a dedicated Agent → Memory page (/assistant/memory) — one setting for phone, WhatsApp, and SMS.
    • Removed duplicate memory controls from WhatsApp and SMS channel sheets (those sheets now link to Memory).
    • Added look-up and clear by phone for a Guest’s lasting summary on that agent.
    • Memory settings are stored per agent; lasting summaries are namespaced per agent so agents in the same workspace do not share summaries.

    v3.2

    Added

    • Added the new History experience that groups SMS, WhatsApp, and voice conversations by customer.
    • Added customer timelines with cross-channel events and links to conversation details.
    • Added estimated outbound message costs for SMS and WhatsApp in History.
    • Added configurable cross-channel memory for SMS, WhatsApp, and voice.
    • Added long-term memory summaries with a user-selectable summary LLM.
    • Added inbound WhatsApp call support so customers can call the agent through WhatsApp.
    • Added user-facing communication tools in the prompt editor for SMS confirmations and WhatsApp messages.

    Changed

    • Renamed the customer-facing Call History label to History while keeping the existing route.
    • Linked WhatsApp text conversations to customers by phone number.
    • Added LLM provider, model, temperature, max token, dedicated prompt, and memory controls for SMS and WhatsApp channels.
    • Made agent tools more explicit in the prompt editor so users can see real actions the agent can perform.

    Notes

    • Outbound WhatsApp calling is not exposed as an active agent tool in v3.2.
    • History and memory remain scoped to the current workspace.

    v3.2.1

    Fixed

    • Fixed unresolved prompt variables in SMS and WhatsApp messages. Variables such as {{hotel_name}} and {{agent_name}} are resolved before inbound text replies and voice-tool messages are sent.

    v3.3

    Added

    • Added Release Notes: v3.3.
    • Added an always-current model catalog: model and voice pickers now sync automatically with our AI partners, so new models appear without app updates.
    • Added live voice syncing for ElevenLabs and Cartesia, and live discovery of OpenAI Realtime built-in voices with short descriptions.
    • Added Voicetta-managed voice, STT, and TTS options that can run through the Voicetta subscription with no provider account required.
    • Added billing badges in provider pickers: Your key (your provider bills you) and Voicetta (billed to your wallet).
    • Added Wallet history to Billing / Balance & payments so workspace owners can review top-ups, usage debits, bonuses, refunds, and adjustments.
    • Added Lead Scraper exports for GHL Companies CSV and Excel workbook files.
    • Added GoHighLevel Company export with tagged Contact creation and Contact-to-Company linking.

    Changed

    • Providers without your own API key are no longer disabled — they stay selectable and run through Voicetta where covered.
    • Saving agent settings no longer requires adding provider keys for Voicetta-covered providers.
    • Pricing and billing copy now refers to the built-in voice path as Voicetta voice / Voicetta voice line.
    • Lead Scraper jobs can resume after backend restarts or deploys instead of staying stuck in active progress.

    Notes

    • SMS and WhatsApp text replies still require your own LLM key.
    • Current in-app pricing remains sourced from the backend Pricing endpoint. At release time, standard Voicetta AI is $0.14/min, realtime Voicetta AI is $0.39/min, Speechify hosted agents are $0.12/min, and the Voicetta voice line is $0.012/min where applicable.

    v3.4

    Added

    • Added Release Notes: v3.4 and the AI Evaluations guide.
    • Added AI Evaluations: define quality criteria (title plus expected behavior) and Voicetta grades every new call automatically as pass, fail, or not applicable — with a short rationale per verdict.
    • Added the AI Evaluations page to the main menu for creating, editing, enabling/disabling, and deleting evaluations, with aggregate pass rates per criterion.
    • Added an evaluations section to call details in History showing per-call results and rationales.
    • Evaluations are now shared across all agents in a workspace, with per-agent on/off control in the evaluation settings.
    • Added WhatsApp/SMS thread grading: text conversations are graded as whole threads 30 minutes after the last message, with a Thread Summary per thread.
    • Added optional recovery follow-up per evaluation for text threads, phone calls, and WhatsApp calls. Phone calls prefer SMS; WhatsApp calls prefer WhatsApp text when its 24-hour window is open. Per-thread and cross-channel daily limits are configurable, and opt-outs remain enforced.
    • Added Recovery Journey reporting: initial Fail remains visible when the final outcome becomes Recovered, with manager metrics and filters for awaiting replies and needs-attention items.

    Changed

    • The AI models used for post-call analysis and AI evaluations are now managed centrally by Voicetta, so model upgrades reach all workspaces immediately without app updates.

    Notes

    • Evaluations grade calls that end after the evaluation is created and enabled; existing history is not graded retroactively.
    • Calls without a usable transcript are marked not applicable.