Release notes

Release date: 14-10-2026

1. Agent Interoperability (A2A)
   • Publish your agents over the Agent2Agent protocol: other platforms and agents can discover and talk to them
   • Every publication gets a public agent card for discovery, with JSON RPC and HTTP bindings, protected by an API key
   • Nothing is discoverable until you publish, and unpublishing takes the agent offline immediately
2. Governance & Audit
   • Decision log: a timeline per conversation of routing decisions, tool calls, and guardrail verdicts, with export to your SIEM
   • Settings history: every settings change is recorded, with who changed it, what changed, and when
3. Conversations & Voice
   • Switching language no longer restarts the conversation, with configurable preservation
   • Clear error codes per conversation turn, with a bounded turn lifecycle
   • Fewer false microphone errors in the chat widget
   • Fixes for late empty speech results and the Digital Human microphone after a language switch
4. Knowledge Base
   • Vector storage can now run on PostgreSQL for sturdier scaling
5. AI Models
   • New Gemini voice models added, deprecated Gemini models retired with automatic replacement
   • Test suite grading now defaults to GPT 6 Luna
   • Live conversations get priority over test traffic at the model providers
6. Platform
   • Faster responses through improved caching across the platform
   • Faster deployments through smaller images and quicker rollouts
   • Security updates
7. Bug Fixes
   • Fixes for card label translations, right to left landing text in the webview, flow condition labels, knowledge tables with free text, and demo share titles
   • Many smaller fixes across the platform

Release date: 29-09-2026

1. AI Models & Providers
   • Added GPT 6 (Sol and Luna) and GPT 5.4 Pro
   • New speech provider: Gradium, for both text to speech and speech recognition
   • Mistral models now run directly on the Mistral API
   • Google Gemini on Vertex AI hardened with automatic fallback and priority routing for lower latency
   • Faster responses through latency based routing for OpenRouter models
2. File Output
   • Document rendering now runs fully inside the platform
   • Generated files are cleaned up automatically after their retention period
3. Usage & Cost Insight
   • AI usage metering across the platform: interactions are measured per tenant, per bot, and per domain
   • New AI cost calculator and interaction calculator in the platform tools
4. Outbound Campaigns
   • Campaign improvements: switch the bot on a campaign, better linking of bots, and a call now button
5. Digital Human
   • Configurable refresh behaviour and switching to the normal chat view from the 3D experience
   • Response feedback with thumbs now also available in the Digital Human view
6. Webview
   • Mobile previews and demo links
   • Voice input starts recording on the first tap
   • Example prompts can link straight into the webview
7. Platform
   • Workflow runs now stay within the currently linked agents, and tool toggles are saved reliably
   • Model reasoning can no longer leak into customer answers
8. Bug Fixes
   • Fixes for Arabic card layouts, the chat bubble on window resize, lead capture forms, and website crawl defaults
   • Fix for training texts with special characters in NLU providers
   • Many smaller fixes across the platform

Release date: 18-09-2026

1. Webview Channel
   • New Webview channel: embed the assistant as a native in app experience, with its own channel analytics
   • Host context variables: allowlisted parameters from the host app become variables the bot can use
   • Offers with detail pages, maps, and redemption flows inside the conversation
   • Personalized conversations through a new user data provider connector: verified balances, loyalty points, and tiers become variables, badges, and cards
2. Channels
   • Direct integrations for Facebook Messenger, Slack, Telegram, and Email: faster, more reliable, and without third party dependencies
   • Microsoft Teams overhaul: automatic bot provisioning, reclaiming an already provisioned bot, and better error handling with fallbacks
3. Rich Cards
   • Full audit of the card library, now more than 45 card kinds, including new telco and credit card benefit cards
   • Prompt instructions per card and label translations, so cards render correctly in every language
   • Card previews in the platform
   • Many rendering fixes: right to left text in card values, carousels, buttons, and disabled cards
4. Voice & Audio
   • Smoother audio playback in the chat widget
   • Cancelling a voice recording no longer sends the transcript to the bot
   • The conversation language no longer switches when automatic language switching is off
5. Guardrails & Safety
   • Blocked topics are now enforced deterministically instead of judged by the model
   • Content scanner no longer blocks answers over ordinary words
   • Rejected files are kept out of vector storages and public links
   • Fix for prompt leakage
6. AI Models
   • Google Gemini is now served through Vertex AI
   • OpenAI speech recognition migrated to the newest transcription models
7. Platform
   • Deployments without dropped conversations through graceful shutdown
   • Response timeout configurable per bot
   • Deeper platform observability and security updates
8. Bug Fixes
   • Fixes for incomplete Teams setups, video cards, workflow reasoning settings, and the copy link button
   • Many smaller fixes across the platform

Release date: 25-08-2026

1. Blits.ai MCP Server
   • Expose your agent tools and data as a hosted MCP server: one URL per server, secured with an API key
   • External MCP clients (Claude, IDEs, other platforms) use exactly the same tools as your own agents, so internal and external answers never diverge
   • Built on the Streamable HTTP profile, with rate limiting per server
   • MCP usage is tracked per server and visible in the platform
2. Knowledge & Data
   • Smarter answers over SQL data through a semantic layer: value mapping, taxonomy translation, and place name recognition
   • Eligibility rules on data tools: control which rows and records an agent may use, in soft or strict mode
   • Tracing of tool and SQL calls, including origin, arguments, and latency
   • Recency ranking now automatically drops superseded duplicate documents
3. Agentic Workflows
   • Operators can stop a running workflow, and stranded runs are cleaned up automatically
   • The run trail is saved live while a workflow runs, with a bounded retry budget so runs always terminate
   • Stricter run outcomes: a run now fails when the agent was never invoked or returned an empty result, and check rules grade the records a run produced
4. Platform
   • Improved platform monitoring
   • Security and compliance updates
5. Bug Fixes
   • Fix for a click sound at the start of spoken replies
   • Arabic pronunciation of large numbers corrected
   • Content scanner no longer blocks phone country codes
   • Guardrail refusals are sent once per blocked turn instead of repeating
   • Fixes for custom TTS configuration and mobile view layouts
   • Many smaller fixes across the platform

Release date: 13-08-2026

1. Knowledge & Answers
   • Source citations: answers can now cite the documents and pages they are grounded in, with links and source dates
   • Recency ranking: retrieval prefers newer content when sources overlap
   • Retention rules for the knowledge base, with automatic expiry and cleanup of outdated content
2. Voice & Speech
   • Bring your own speech provider: connect any OpenAI compatible TTS or STT endpoint with your own key, including self hosted models
   • Barge in: users can interrupt the bot while it is speaking and the bot stops talking to listen
   • Adaptive voice detection that calibrates to background noise, so soft speakers are no longer cut off
   • Speech recognition settings per bot language, including dialect selection
   • New option to hide voice transcripts in the chat: voice turns show a voice message placeholder with a waveform instead
3. Handover & Leads
   • Live agent handover on the web chat widget, with agent messages clearly marked in the conversation
   • New lead capture tool: agents collect structured lead information during the conversation
4. Platform
   • Qatar region selectable at login
   • Expanded analytics dashboards
   • Security and compliance updates
5. Bug Fixes
   • Fix for speech recognition answering in the wrong language
   • Fixes for CSV imports in the knowledge base, the model picker, and Google Dialogflow projects
   • Many smaller fixes across the platform

Release date: 05-08-2026

1. Outbound Calls
   • New Outbound Calls channels: AI agents place proactive phone calls, through Twilio or your own SIP telephony
   • Updated campaign management: upload a contact list, set the calling window, and let the campaign dial, pace, and retry automatically
   • Compliance built in: spoken AI disclosure, instant opt out with suppression, calling windows, and attempt caps
   • Advance call controls for agents: transfer to a human, schedule a callback, send keypad tones, and capture lead outcomes
   • New tenant permissions for viewing and managing campaigns
2. Voice & Speech
   • Numbers, dates, and amounts are now spoken in the right dialect, based on the language and locale of the bot
   • Cleaner transcripts: speech input is rewritten in the background by a lightweight model, without added latency
   • New microphone recording settings and visuals in the chat widget
3. Conversations
   • Status messages are now shown in the language of the user, with detailed, simple, or off modes
   • Improved idle messages while the agent is working
   • Voice hints are now off by default and can be enabled in the settings
   • Faster responses through caching improvements in the answer pipeline
   • More reliable conversation history under high load
4. Feedback
   • Response feedback on voice conversations now captures the caller's voice and the bot's spoken reply, with playback in the admin feedback view
5. Digital Human
   • Refreshed Digital Human colors and UI
6. Platform
   • Sturdier conversation locking
   • Security updates
7. Bug Fixes
   • Fixes for test chat settings, saving engine settings, vector storage editing, and forced RAG with conversation history
   • Many smaller fixes across the platform

Release date: 28-07-2026

1. Payments
   • Additional Payment provider integrations: Stripe, Adyen, and Mollie
   • PCI DSS compliance across the platform
2. Test & Monitor Suite
   • New test suite: build test sets with an LLM grader and run them on demand or as scheduled monitors
   • Test conversations with multiple turns, editable grader prompts, file import, and reusable templates
   • Live progress, results shown directly in the platform, asynchronous runs and regrades, and extra filters on test conditions
   • Large speed improvements for big test runs
3. Channels & Live Agent Handover
   • Freshdesk integration: live agent handover and Freshdesk available as a tool for your bots
   • Zoho SalesIQ live chat handover
   • WhatsApp improvements: WhatsApp mode for the web widget with better cards and headers
   • Test Chat now has its own channel settings, including response feedback
   • More stable websocket connections: ping and pong messages, connection IDs, and automatic reconnection
   • Chat widget upgrades: improved language switching, font switcher and font size setting, a growing input field, new card types and themes, and an overhaul of status messages with a simplified messaging setting
4. AI Models
   • Added GPT 5.6 (Sol, Terra, and Luna), GPT 5.5 Pro, and GPT 5.4 mini and nano
   • Added Claude Opus 4.8 and Claude Sonnet 5
   • Added Gemini 3.5 Flash, Grok 4.5, DeepSeek V4, GLM 5.2, and Kimi K3
   • Deprecated models (older GPT versions and the Gemini 2.5 family) are greyed out in the picker and mapped to their replacements automatically
   • Removed Amazon Lex, Nuance, and voice tone analysis
5. Voice & Text to Speech
   • New provider: Silma, for high quality Arabic voices
   • New models: Murf Falcon 2 and Gemini 3.1 Flash, plus updated Murf and Cartesia models
   • Many new Amazon Polly voices across languages
   • Emotion engine v2: the LLM adds emotion tags that steer the synthesized voice
   • Cache management for synthesized audio, with cleanup and cache logs
6. Agents, Workflows & Knowledge
   • Workflows can now run as a block inside the dialog manager
   • AI use case creation: describe your use case and the platform builds a first bot version, with example prompts and a getting started guide
   • Response feedback: end users can rate answers with a thumbs up or down and leave a comment, visible in a new Feedback tab
   • RAG improvements: hybrid BM25 and vector retrieval, a selectable retriever, whitelist and blacklist fields, and better Arabic support
   • New business connector: Odoo
   • Notification settings for knowledge base updates
7. Platform & UI
   • Admin app rebuilt with a refreshed design system
   • Analytics 2.0
8. Bug Fixes
   • Fixes for TTS streaming timeouts, avatar uploads, test suite disconnections, and dialog flow saving
   • Many smaller fixes across the platform

Release date: 12-05-2026

1. Guardrails & Anomaly Detection
   • Guardrails v2: new setup with content scanner, regex guardrail, and quality guardrail
   • Lexicon-based guardrails for hate speech, self-harm, and violence detection
   • LLM anomaly detection service with configurable per-tenant settings
   • Anomaly flags and stats storage for review and follow-up
   • Improved guardrails switching on input, including fixes for Twilio streaming
2. Agentic Workflows
   • Email triggers for agentic workflows
   • Scheduled (cron) triggers for agentic workflows, powered by NestJS scheduling
   • Custom function tools support in workflow service
   • Force-download action for agentic-workflow-related documents
   • Improved linked-agent resolution between workflows and agents
3. AI Models & Integrations
   • Added GPT 5.5 and Claude Opus 4.7
   • Deprecated Gemini 2.0
   • Cleaned tool schema for Gemini models to improve compatibility
   • ElevenLabs async processing for larger audio files
4. Frontend & UI
   • New themes available in demo mode
   • Environment selector in the frontend for easier switching
   • Button disabled when vector storage is connected (prevents accidental changes)
   • UI fixes for cards rendering
   • Removed stray markdown artifacts from streaming LLM output
   • Page-max selector improvements and multi-tenant rendering fixes
5. Bug Fixes
   • Fix for Teams integration
   • Fix for Twilio streaming when guardrails switch on input
   • Fix for phone number handling
   • Fix for multi-tenant routing
   • Fix for issues #4913, #4948, and #4610
   • Fix for #4954 with additional test coverage
   • Fix for logs not appearing when restoring shortly after a crash

Release date: 22-04-2026

1. Agentic Workflows
   • Sub-agents: agents can now spawn and orchestrate other agents
   • Agent Tasks v1: first version of agent task management
   • Tool execution policy layer that classifies tools as discovery vs execution and enforces retry logic when plans are ambiguous
   • Improved column mapping with richer functionality and prompt improvements
   • SQL storage builder for agentic workflows
   • Configurable transcription settings within agentic workflows
   • Better agentic column mapping (#4919)
2. Speech & Audio
   • Self-hosted diarization service using WhisperX (medium model) + Pyannote for speaker diarization, segmentation, and configurable batch size
   • Adjustable VAD (Voice Activity Detection) settings for self-hosted diarization
   • ElevenLabs added as a Speech-to-Text provider with full streaming support
   • STT streaming enabled across all providers
   • Default ElevenLabs audio format changed to PCM
   • ElevenLabs voice IDs updated with migration script for female/male-only models
   • Temp files cleaned up automatically after diarization
3. AI Models & Integrations
   • Gemini Live: first working integration
   • Gemini + Google TTS fully working with streaming
   • Gemini voice options built dynamically from language codes and voices
   • OpenAI open source models added as a separate section (Ollama removed)
   • MCP Tools: frontend UI, server management with keep-alive and timeout handling, unique MCP server enforcement
   • Separate LLM per language: bots can now use a different LLM model per configured language
4. Platform & Administration
   • Production flag for bots, agents, and workflows to distinguish configurations that are ready for production
   • Traces download and analyse function available on the platform
   • Subscription tiers: editable tier pricing, import/export of subscriptions, public endpoints for additional usage charges
   • Interaction metering with public meter endpoint and error log metering
   • FAB dashboard widget now reads from SQL DB instead of a static file; direct SQL credentials no longer exposed
   • Handover modal for agent to human handover flows
   • MongoDB backups automated
   • Betterstack logging integration
   • Direct SQL link visible to admin users
   • Show all users under master tenant in tenant selector
5. Frontend & Chat Bubble
   • RAG filters reintroduced, with frontend and backend fully working
   • Chat bubble moved into shadow DOM for better CSS isolation and style scoping
   • Separate events for different activity types in the chat bubble
   • TTS streaming cache added, including cache clear on settings change
   • Adaptive cards: improved styling, fixed filter, fixed empty card handling
   • Pixel streaming: buffering of 10 chunks before sending, mute/unmute only sent on new message parts
6. Bug Fixes
   • Fix for tool leak in agent execution
   • Fix for bot interactions count showing 0; interactions now included in metrics
   • Fix for language switch and LLM switch when inputting a different language
   • Fix for vector storage and language detection logging
   • Fix for auth flow for admin users
   • Fix for carousel rendering
   • Fix for bulk delete
   • Fix for intra users
   • Fix for 429 rate limiting on BCR
   • Fix for doubled sentences in LLM streaming output

Release date: 28-01-2026

Overview

• TTS prompts are now configurable per language, including Gemini-specific prompts, with updated selector behavior and streaming/voice options
• Pitch controls and defaults in the TTS test UI were improved for more reliable voice tuning
• Gemini prompt handling now supports multiple prompts with added tests and fixes
• Chat widget usability improved with fixes for iOS keyboard overlap and dialog button interaction
• Security hardening includes updated middleware, expanded test coverage, and CSP policy updates
• MCP endpoints can be configured from the frontend with improved server-side tool loading
• Conversation language rules now support an explicit condition option and per-language configuration toggles
• Unreal widget can host a browser controlled by LLM events for richer experiences
• ElevenLabs added as a speech-to-text provider, plus expanded audio test coverage
• API now exposes bot info for DH automation pipeline env configuration
• Google project creation supports an optional parent parameter

Release date: 15-01-2026

1. Major Changes
   • Complete removal of LUIS (Microsoft Language Understanding) integration
   • Migration from LUIS to CLU (Conversational Language Understanding)
   • Migration of remaining HTTP requests to Axios for improved reliability
2. Security & Performance
   • Fixed all security vulnerabilities across the platform
   • Added TTS streaming cache for improved performance
   • Implemented performance tests and concurrent platform load testing
3. LLM & AI Improvements
   • Added Gemini preview languages for voice options
   • Added OpenAI OSS models to LLM model types
   • Fixed LLM output language detection (#4723 #4735)
   • Fixed issue with LLM response when streaming is enabled (#4740)
   • Added an LLM test suite
4. Text-to-Speech (TTS) Enhancements
   • Fixed TTS streaming when split sentences is enabled (#4398)
   • Extracted streaming player into dedicated service with mute/unmute functionality
   • Improved TTS streaming for all providers (Google, ElevenLabs, OpenAI, etc.)
   • Added support for custom voices on ElevenLabs
5. Speech-to-Text (STT) Improvements
   • Enhanced speech to text services and added tests for them
   • Fixed speech-to-text misdetection issues (#4499)
   • Improved Arabic speech recognition streaming support
6. User Interface & Experience
   • Fixed reactive state issues with continue container language switching
   • Made tooltips interactive and clickable (#4231)
   • Fixed create entity modal scrollbar issues (#4226)
   • Improved navigation and linking throughout the platform
   • Enhanced password reset and forgot password screens (#4710)
   • Fixed icon alignment and positioning issues (#4707)
7. Bot Management
   • Updated bot versions automatically
   • Added functionality to preserve bot settings during import/export
   • Fixed bot name and icon change functionality (#4642)
   • Improved bot settings management and GDPR settings preservation
8. Admin Panel Improvements
   • Fixed LLM details page 404 errors (#4712)
   • Fixed Slack settings modal not showing (#4588)
   • Moved admin panel navigation component for better organization
   • Enhanced connectors panel positioning (#4659)
9. Database & Storage
   • Reduced database requests for better performance (#4595)
   • Added support for mocked database storage in tests
   • Fixed user storage to preserve password fields properly
10. Testing & Quality Assurance
    • Added unit tests for all major components
    • Implemented performance testing framework
    • Added stress testing capabilities
    • Enhanced test coverage for blocks, services, and storage
11. Digital Humans & Unreal Integration
    • Small UI improvements for digital humans
    • Enhanced AI message display on various color backgrounds
    • Improved idle timer functionality for branches and kiosks
12. WebSocket & Real Time Features
    • Moved all WebSocket types to common library
    • Improved WebSocket request and response handling
    • Enhanced chunk streaming for better real time communication
13. Guardrails & Safety
    • Removed Async guardrails setting
    • Added guardrails overwrite functionality
    • Improved guardrails configuration and management
14. Multi-language Support
    • Auto-translate phrases on bot change for all supported languages
    • Improved language code extraction and handling
    • Enhanced Arabic text alignment and display (#4705)
15. Analytics & Monitoring
    • Enhanced statistics and metrics collection
    • Improved performance analytics and monitoring
16. Infrastructure & DevOps
    • Updated package dependencies and security patches
    • Improved Kubernetes configurations for Unreal integration
    • Enhanced build and deployment processes

Release date: 30-12-2025

1. LLM Models
   • Added Google Gemini 3
   • Removed Sonnet 3.5 and Opus 3
2. Voice & Customization
   • Ability to set custom voices on ElevenLabs
3. Audio Cache
   • Improved Audio Cache for all agents & bots
   • Added button to clear cache in bot settings
4. Digital Humans
   • Small UI improvements
   • Improved AI message on various color backgrounds
   • Idle timer for branches & kiosks

Release date: 18-12-2025

1. LLM Models
   • Added Mistral 3.2
   • Added Grok 3 + 4
   • Added Fanar
   • Added Humain Allam
   • Added Qwen 3
   • Added Llama 4
   • And more
2. LLM Features
   • Add custom LLM endpoints & Support for Open Router
   • Custom exit command for LLMs
   • Set AI disclaimer message
3. Dialog Logic
   • Added Authentication, Pin Change and Complaints block in Dialog Logic
4. Speech Services
   • Added new SST and TTS models
   • Chatterbox
   • Elevenlabs v3
   • Google Gemini TTS/SST
5. API & WebSocket
   • Various API / Websocket changes, for the latest details see the documentation here: https://www.blits.ai/api-docs
6. Digital Humans
   • Renewed UI
   • Better portrait & mobile support
   • Ability to add example phrases on screen
   • Settings debug window
   • Multiple microphone recorder modes
   • Language selector on screen
   • TCP connection switching

Release date: 16-10-2025

1. Testing
   • Improved test-suite with multiple test sections and validators
2. LLM Models
   • Added additional genAI models (i.e. GPT 5, Claude 4+, Gemini 2.5, etc.)
3. AI Features
   • Personas for LLM prompts
   • Custom Tool/Agent instruction Prompts
4. Performance
   • Improved DB connections
   • Better text-to-speech speed and stability
5. User Interface
   • Chat Animations for chatWidget and Digital Human UI
6. Tools
   • Websearch
   • Wikipedia
   • Google Search
   • Crypto
7. Dialog Manager
   • History Block in Dialog Manager
8. Web Widget
   • window.ChatWidget.onBase64audios()
   • window.ChatWidget.onMessage()
   • window.ChatWidget.onRestart()

Release date: 27-07-2025

1. Analytics
   • Metrics: The dashboard can now select periods
2. Storage
   • Improved Vector Storage
3. Speech Services
   • Google SST is updated
4. Web Widget
   • window.ChatWidget.setLlmAdditionalInput('')
   • window.ChatWidget.getConversationId()

Release date: 27-05-2025

1. Analytics
   • Dashboard graphs added for users and interactions for months and days
2. LLM Models
   • New LLM models available: GPT 4.1, GPT 4.1-nano, OpenAI o3
3. NLP Services
   • Amazon NLP is upgraded to Amazon Lex v2
4. Bot Management
   • Ability to create bots without NLP engine
5. Test Widget
   • Audio mute button
   • Audio doesn't automatically start
6. Voice & Audio
   • Additional voices added for ElevenLabs
7. API Features
   • The API can now return emotions from a bot

Release date: 12-05-2025

1. Analytics
   • Widget in the dashboard has been changed
   • More elements are measured now like (Output guardrails time, RAG filter time, LLM section time, etc.)
   • LLM metrics are moved to a separate LLM Statistics tab
2. Language Features
   • Automatic language switching in the platform with LLM's or without
   • Language detection and switching (e.g., if you speak Italian, the default language switches)
   • Switch language via API / Websocket
3. LLM Integration
   • LLM is now an integrated service in the platform, improving latency
4. Testing Tools
   • Test widget improved with new UI
   • Can test Text To Speech engines
   • Switch which channel you want to test (production, test or BCR)
5. Voice & Audio
   • New TTS settings per language (better voice selection)
   • Select preferred dialect per language
   • Additional voices added for ElevenLabs
6. User Management
   • Improved UX for inviting users
7. Bot Features
   • Create bots without NLU engine (LLM only)
   • Split of Text and Voice via API / Websocket
   • Sentence splitting
8. API Improvements
   • Improved handling of Websocket, Directline and Rest API
   • Improved conversational logs
9. Security
   • Improved password requirements
10. Safety Features
    • Async guardrails

Release date: 01-04-2025

1. LLM Models
   • Added foundational model support for Deepseek
   • Added foundational model support for Google Gemini
   • Added foundational model support for Anthropic Claude
   • Added foundational model support for Facebook's Llama
   • Added foundational model support for Alibaba Cloud Qwen
   • Added foundational model support for Mistral
2. LLM Management
   • Ability to switch LLM providers on existing LLM's
3. RAG Features
   • Advanced RAG filtering now possible
4. Speech Services
   • Added OpenAI's new TTS & SST models
   • gpt-4o-transcribe, gpt-4o-mini-transcribe
   • gpt-4o-mini-tts
5. Multi-language
   • Multi-language support on system messages
6. Safety
   • Customized guardrails messages

Release date: 20-02-2025

1. LLM Improvements
   • Improved RAG prompts
   • HTTPS only for LLM services
   • OpenAI defaults to GPT-4o
2. Speech Services
   • ElevenLabs added as Text-To-Speech and Speech-To-Text engine
3. Monitoring
   • Uptime monitoring moved from bots to https://status.blits.ai/

Release date: 13-02-2025

1. LLM Improvements
   • Regional support for LLM's
   • Evaluate settings for LLM (advanced use-case)
   • Improved Guardrails configs for LLM
   • Additional embedding models available
2. Speech Services
   • Deepgram & Speechmatics added as Text-To-Speech and Speech-To-Text engines
3. Dialog Manager
   • Additional options for LLM in dialog manager
4. API Features
   • UserId can be set via REST and Websocket
5. User Experience
   • Improved conversational history on recurring users
   • Auto-enabling engines on save
6. Engine Changes
   • Removed RASA

Release date: 26-01-2025

Release date: 01-09-2024

1. LLM Features
   • Select GPT-4o as a chitchat provider
   • Create bespoke LLMs and use them in your bot (coming soon)
2. Speech Services
   • Added OpenAI Text to Speech and Speech-to-Text services
   • Configure language for each Speech-to-text service
   • Fixed Microsoft Text to Speech service errors
3. Voicebot Features
   • Add random idle phrases to fill silence when building a Voicebot
4. API Improvements
   • Extended Blits API with Websockets support
   • Check the documentation: https://documenter.getpostman.com/view/12778718/2sA3kd8wiS

Release date: 15-02-2024 Microsoft CLU stands for 'Conversational Language Understanding'. It is the latest version of the Microsoft Conversational AI Engines. Microsoft supports multi-language out of the box, in the sense that if you write an English chatbot an end-user asks a question in another language, it will understand the right intent and/or trigger the right entity. As an example. The below bot is written in English. Still it has the capability to understand the German question it was asked, and trigger the appropriate response: The ability to make the answers of your chatbot truly multi-language. We'll be back!

Release date: 15-02-2024 We introduced a new AI Engine provider: NeuralSpace ! Neuralspace supports languages spoken across India, South East Asia, Africa, the Middle East, Scandinavia and Eastern Europe. Aka; low-resource languages. We're really happy with this valuable addition to our stack of AI providers.

Release date: 13-02-2024

1. Service Upgrades
   • Upgraded to version V3 of Azure Cognitive Services Text Analytics
2. User Interface
   • Fixed scrolling issues when dialog content exceeds page height

Release date: 19-08-2023 We introduced the integration of GPT-4 by OpenAI. You can now choose between GPT-4 or GPT-3.5 for your chit-chat module. Note: We retired GT-3.0 as a chitchat engine choice.