Skip to main content

Changelog

All notable changes to the Lira AI platform are documented here. Dates use YYYY-MM-DD format.

Platform focus

Lira is now exclusively a customer support platform. The Meetings, Interviews, and Sales Coaching products that appear in earlier entries below have been retired — they were part of an earlier iteration of Lira and are not available on the current platform. Their historical entries are preserved for context.

2026-07-29 — Escalation routing to Slack, Linear and webhooks​

  • Escalation destinations are now configurable — Settings → Support → Escalation gains a Also send escalations to your tools card: connect Slack over OAuth and pick a channel, connect Linear and pick a team, or POST to your own https:// endpoint signed with HMAC-SHA256.
  • Fixed: the outbox never fired. The Slack, Linear and webhook adapters have existed since Phase 6, but outbox_providers, outbox_webhook_url and outbox_webhook_secret were missing from the config write schema and had no UI, so no organization could ever enable delivery. All three are now settable and the fan-out works end to end for ticket.created, ticket.escalated and ticket.resolved.
  • New docs — Integrations, SLA policies, Routing & assignment, Roles & permissions, Automation & rules, and a public API reference. Each states plainly what is not supported as well as what is.

2026-07-29 — Self-serve signup restored​

  • Anyone can sign up — /signup is open again. Create an account, verify your email, and the onboarding flow stands up your organization on the Free plan, in sandbox. No invite code, no scoping call, no card.
  • Self-serve plan upgrades — Pro and Scale are purchased from Settings → Billing through Paddle checkout. Enterprise remains a conversation with our team.
  • Concierge invites still work — a ?invite=… link pre-fills signup and attaches the new account to the inviting organization, but is no longer required. This supersedes the 2026-05-26 entry below.
  • Docs corrected — Get an Account, Quickstart, Authentication, and Subscription & Billing no longer describe Lira as invite-only.

2026-07-20 — Paddle-backed subscription billing​

  • Subscription & Billing docs refreshed — plan, usage, sandbox/live, plan-change, downgrade, and limit behavior now match the current platform.
  • Paddle billing foundation documented — Pro and Scale checkout, billing status, receipts, invoices, payment-method updates, customer portal access, and verified webhook entitlement rules are now covered.
  • Sandbox-to-live billing language tightened — going live starts billing only for paid plans, and no organization becomes billable silently.
  • Overage wording corrected — paid overage is described as available where overage continuation is enabled, instead of promising automatic continuation before that feature is fully live.

2026-05-26 — Concierge onboarding + per-employee invites​

Superseded

The invite-only changes in this entry were reversed on 2026-07-29 — signup is self-serve again. The per-employee invite system below is still current.

  • Self-serve org creation removed — new organizations are now provisioned by the Lira team after a scoping call. The in-app "New organization" entry opens a contact-team modal instead.
  • Per-employee invite system — org admins generate one-time, expiring invite links scoped to a specific email + role. Invitees set a password (if new) and land directly in the org. Replaces the static LRA-XXXX shared-code join flow.
  • Resend-powered invite emails — invite links are now emailed automatically when an admin generates them.
  • Public sign-up disabled — the marketing site routes prospects through "Speak to an expert" instead of a self-serve form. /login is the only public auth path; /signup?invite=… works only with a valid concierge invite code.
  • Admin: provision org for existing user — the Lira-team admin dashboard can now spin up a new organization under an existing user account.

2026-04-13 — Customer Support Module​

  • Full customer support documentation — 8 new pages covering every aspect of the support module: activation, portal, widget, inbox, actions, proactive outreach, analytics, and settings reference
  • Docs buttons in app — Docs links added to the Support page header and activation wizard, linking directly to the relevant documentation sections
  • Support Portal — Branded self-service page at support.liraintelligence.com/<slug> where customers can submit tickets, track status, and chat with Lira; can be embedded as an iframe
  • Portal save fix — Portal enabled/disabled state, portal slug, and SLA hours now correctly persist after saving (backend Zod schema was silently stripping these fields)
  • Porter reorder in Settings — Support Portal is now listed first in the Channels tab (above Chat Widget, Voice, and Email), with an inline Docs link
  • Voice card update — Voice Support description updated to "Inbound phone support powered by Lira's voice"; removed outdated Twilio reference

2026-04-08 — Customer Support AI (Phase 1–5)​

  • Inbox — Unified conversation view across email, chat widget, voice, and portal channels; status filters (all / open / pending / escalated / resolved); full-text search by subject, customer name, email, and intent; sentiment indicators and CSAT scores per conversation
  • Support Portal — Public self-service page at support.liraintelligence.com/<slug>; customers can submit tickets, track status, and chat; embeddable via iframe
  • Chat Widget — Single <script> tag embed; floating chat button with Lira AI responses; customisable colour and greeting message; live animated preview during activation
  • Activation Wizard — 5-step onboarding: email setup (with custom domain forwarding option), channel selection, integration connections, Knowledge Base seed, and live activation
  • Actions — Approval queue for autonomous actions Lira proposes; approve or reject with full history log
  • Proactive Outreach — Event-triggered automated messaging; configurable event type, message template with {{variables}}, channel (email/voice), cooldown, and daily cap; activity log
  • Analytics — CSAT scores, autonomous vs escalated resolution breakdown, average first response time, monthly usage tracking, weekly report with top intents and KB improvement stats
  • Escalation engine — Confidence threshold slider (0–100%); force-escalate intents list (always escalate regardless of confidence); email alerts; SLA target (hours)
  • Knowledge Base drafts — Lira creates draft KB entries when it encounters questions it can't answer well; admins approve to improve future responses
  • Support Settings — Four-tab settings panel (Widget / Channels / Behaviour / Escalation) accessible from the main Settings page

2026-03-29 — Documentation Site Launch​

  • Launched docs.liraintelligence.com — comprehensive documentation for the Lira AI platform
  • 35+ pages covering API reference, customer support, knowledge base, and getting started guides
  • Full-text search across all documentation
  • Dark mode support

2026-03-28 — Email System & Custom Domains​

  • Email API — send AI-generated emails on behalf of your organization
  • Custom sending domains — configure DKIM, SPF, and DMARC for branded email delivery
  • Inbound email processing — AI reads and replies to customer email threads via SES → SNS pipeline
  • Email thread management UI in the dashboard

2026-03-23 — Interview Platform v2​

  • AI Draft — generate complete interview configurations from a natural language prompt (POST /interviews/draft)
  • Question Generation — AI generates tailored questions based on role, skills, and resume (POST /interviews/generate-questions)
  • Two-phase evaluation — Phase 1 auto-evaluates after interview ends; Phase 2 scoring is on-demand
  • Decision recording — hire, no_hire, next_round, undecided with notes
  • Multi-round support — link related interviews via GET /interviews/:id/related
  • Interview modes: solo, copilot, shadow

2026-03-22 — New Brand Identity​

  • Updated logo to the trefoil mark across all surfaces
  • New lira_black.png and lira_white.png logos for light/dark modes
  • Refreshed favicon site-wide

2026-03-15 — Task Execution Engine​

  • AI task review — Lira autonomously validates and refines extracted tasks (POST /tasks/:id/lira-review)
  • Task execution — execute approved tasks through the Lira runtime or email (POST /tasks/:id/execute)
  • Execution results — track what was created and where (GET /tasks/:id/result)
  • Email notifications on task events

2026-03-10 — Knowledge Base & Semantic Search​

  • Document upload — PDF, DOCX, TXT with automatic parsing and chunking
  • Web crawl — crawl a website with configurable depth and page limits
  • Semantic search — natural language queries matched against document chunks so answers stay grounded in approved knowledge
  • Document reprocessing — re-index existing documents after embedding model upgrades
  • Knowledge base context automatically injected into Nova Sonic system prompts

2026-03-02 — Organization Context System​

  • Organization profiles — company name, industry, products, terminology, and custom AI instructions
  • Culture settings — define company values and communication style
  • AI automatically uses org context for more relevant meeting participation
  • Dynamic context refresh — optionally refresh org context mid-session

2026-02-20 — Wake Word System​

  • 3-layer detection — exact match, Levenshtein fuzzy match, Soundex phonetic match
  • Configurable AI name (default: "Lira")
  • Cooldown system: 45s general, 30s during meetings
  • Confidence levels: high, medium, low
  • Wake word can be disabled per deployment via wake_word_enabled setting

2026-02-15 — Speaker Diarization​

  • Deepgram Nova-2 integration for real-time speaker identification
  • Speaker-attributed transcripts — each line tagged with the speaker's name
  • Participant tracking throughout the meeting
  • Diarization data used for per-person meeting summaries

2026-02-10 — Core Platform Launch​

  • Bot deployment — launch Lira into Google Meet via headless Chromium + Playwright
  • Audio pipeline — bidirectional PCM streaming via WebRTC interception and getUserMedia override
  • Amazon Nova Sonic — speech-to-speech AI (STT + LLM + TTS in a single stream)
  • Echo gate — prevents Lira from hearing her own audio output
  • 4 personality modes — supportive, challenger, facilitator, analyst
  • Meeting summaries — short (4–6 sentences) and long (detailed breakdown) modes
  • Google Drive source — Drive file access for knowledge documents
  • WebSocket API — real-time transcript and audio streaming to the frontend
  • Usage tracking — beta limits per organization
  • JWT authentication — Google Sign-In + email/password, 7-day token expiry