April 23, 2026
9 min read
A packed release that raises the bar on AI, security, developer experience, and enterprise readiness across both platforms.
It has been an intense development cycle, and the teams have delivered. Jaggle 2026.1 and Jaggle.Conversational AI 2026.1 are now Generally Available, bringing dozens of new capabilities that touch virtually every area of the platform.
Here is a tour of the highlights.
Jaggle 2026.1 - The Headlines
AI That Does More of the Heavy Lifting: This release dramatically expands what the AI layer can do for developers and business users alike:
Page Development Helper: An AI-powered coding companion that assists in React page development with intelligent code suggestions and automated error detection. Developers can now scaffold and refine pages faster than ever.
Auto Documentation: AI-driven documentation generation that covers Forms, Stored Procedures and Functions. The days of manually writing boilerplate docs are behind us.
Literals Translator: Need your application in French, German, or Arabic? The new AI-powered translator leverages LLM providers to deliver context-aware translations of application literals across multiple languages.
SQL NLP Wizard ("Sorcerer"): A natural-language-to-SQL conversion system that lets non-technical users describe their data needs in plain language and get optimized, syntactically correct queries in return - complete with preview, explanation, and iterative refinement.
AI Usage Report: A full reporting tool that tracks LLM provider usage with cost estimates and provider-specific metrics, giving organizations clear visibility into their AI consumption.
AutoML: The machine-learning engine is now live, with ready-to-use classification and analytics capabilities built into the platform.
Anthropic Claude Integration: A new AI adapter brings Claude, Anthropic's state-of-the-art LLM, into the Jaggle ecosystem, expanding the multi-LLM strategy alongside existing providers.
Dynamic LLM Switching: End users can now switch between LLM models on the fly directly from the AI Companion interface - no configuration changes needed.
Workflow & Business Process Enhancements
Swim Lanes: A visual workflow element that clearly separates participants, departments, or roles within a business process, making process ownership instantly visible.
Enhanced Declaration Access: Quick-access buttons for REST Web Service and Stored Procedure declarations streamline day-to-day workflow configuration.
Send Report Action: A new workflow action that enables automated report distribution directly from workflow definitions.
Data Model & Security
Extended Views Enhancement: Filter conditions (SQL WHERE clauses) can now be applied per joined table in extended views, delivering more flexible and performant data retrieval.
Test Data Generation: Automated creation of test datasets accelerates development and QA cycles.
Security Plan Visualization: A new comparison view shows permission differences by role and entity at a glance, making security audits far more intuitive.
Role Clone: Security plans can now be cloned across roles, dramatically reducing setup time for complex RBAC hierarchies.
Page Designer & UX
Enhanced Data Controls: Improved controls for consuming data from REST APIs and Stored Procedures, including data controls without direct data-model binding.
User @Mentions: Application users can now mention colleagues with "@", triggering in-app notifications - including group mentions and rich-text formatting.
SpreadJS Integration: Full spreadsheet capabilities are now embedded in the page designer for advanced data entry and manipulation scenarios.
Enhanced Versioning with Snapshots: Every page save now generates an automated snapshot with retention policies and one-click rollback, providing a comprehensive audit trail of design evolution.
Platform & Infrastructure
.NET 8.0 Upgrade: The platform core, along with Kendo UI and DevExpress components, has been upgraded to the latest framework version for improved performance and long-term support.
WCAG 2.1 & 2.2 Conformance: Jaggle now supports ADA compliance and WCAG 2.1/2.2 standards via integrated accessibility tooling, opening doors for public-sector and regulated-industry deployments.
Enterprise Authentication: Full support for OAuth 2.0, JWT, and OpenID Connect authentication methods, meeting the security requirements of diverse enterprise environments.
Multi-Language Menu Titles: Menu builder titles can now be overridden in multiple languages, supporting true multi-locale application shells.
RAG (Retrieval-Augmented Generation)
Microsoft Teams Integration: The RAG module now connects directly to Microsoft Teams, enabling conversational AI assistants within the collaboration tools teams already use.
REDIS BM25 Scoring: A new caching and scoring layer using REDIS with BM25 ranking improves retrieval accuracy and reduces LLM token consumption.
Qdrant Vector Database Support: Jaggle's RAG engine now integrates with Qdrant, a high-performance, open-source vector database built in Rust. Qdrant brings purpose-built vector similarity search with HNSW indexing, hybrid retrieval (combining dense semantic vectors with sparse keyword-based vectors), and advanced payload filtering - all in a single query. Its binary quantization can reduce memory usage by up to 64√ó while maintaining search quality, and its Rust-native architecture delivers up to 4√ó higher throughput compared to alternatives. For enterprise deployments, Qdrant supports on-premises, hybrid-cloud, and fully managed SaaS models with multi-tenant isolation, making it a natural fit for regulated industries. Combined with Jaggle's existing REDIS BM25 scoring layer, the Qdrant integration gives organizations a best-of-both-worlds retrieval stack: fast lexical matching for exact-term queries and deep semantic search for meaning-based retrieval.
Jaggle.Conversational AI 2026.1 - The Headlines
The Jaggle.Conversational AI platform receives significant upgrades, focused on NLU depth, voice channel readiness, and operational independence.
Voice Channel - Barge-In and Advanced Voice Settings:
One of the most anticipated features in this release is full barge-in support. Users can now interrupt the system mid-utterance - just as they would in a natural phone conversation - and the platform will immediately stop playback, capture the new input, and route the dialogue accordingly. This is a game-changer for voice IVR deployments where callers know what they want and shouldn't have to wait for a menu to finish playing.
Barge-in is fully configurable at multiple levels: a global default can be set per dialogue, and then overridden on a per-response basis within the Dialogue Designer. The same applies to DTMF support, Speech-to-Text language, and Text-to-Speech language - all of which are now first-class settings in the voice tab of every response node. This granular control means a single dialogue can switch languages mid-conversation, enable or disable barge-in contextually (e.g., disable it during a critical confirmation step), and accept keypad input where appropriate.
The underlying integration with Asterisk + UniMRCP is now live, with the UniMRCP plugin supporting dynamic parameterization at node level. Combined with barge-in, platform is now a production-grade voice platform.
LLM-Based NLU - Multilingual Without Training Data
This release introduces a fundamentally new approach to Natural Language Understanding: an LLM-based NLU adapter that replaces traditional intent-classification models with LLM prompts. The practical impact is enormous - organizations can now deploy multilingual dialogues without building and maintaining separate training datasets for each language. The LLM interprets user input in any supported language, maps it to the correct intents and entities, and returns structured results that the dialogue engine processes as usual.
For clients operating across multiple geographies or serving multilingual customer bases, this eliminates one of the biggest bottlenecks in conversational AI deployment: the cost and time of curating training examples in every target language.
The adapter works alongside the existing NLU options (including the new NLP.js adapter for on-premises, cost-effective deployments), so teams can choose the right engine for each use case - LLM-based for rapid multilingual rollout, traditional NLU for high-volume low-latency scenarios, or a hybrid approach.
Conversation Simulation - AI-Powered Dialogue Testing at Scale
Testing conversational flows has always been manual, slow, and prone to blind spots. The new Conversation Simulation feature leverages LLMs to generate realistic test conversations automatically, based on use-case descriptions and scenario parameters. It operates in two modes:
Non-Interactive Mode
The LLM generates complete conversations (both user turns and expected system responses), which are then used to automatically create new intents and training examples. This is a powerful bootstrapping tool for new dialogues - describe the use case, and the system produces dozens of realistic conversation paths that enrich the training data.
Interactive Mode
The LLM plays the role of the end user and interacts in real time with the platform backend. This is true automated regression testing: the simulated user probes the dialogue with varied phrasing, edge cases, and unexpected inputs, and the system records whether the dialogue responded correctly. When a dialogue is updated, teams can re-run simulations to catch regressions before they reach production.
Automated Testing - Record, Replay, Validate
Complementing conversation simulation is a new automated testing utility for deterministic regression testing. Administrators can "record" test cases using the existing Dialogue Tester, capturing input utterances and their expected outputs. When changes are made to a dialogue, the recorded test suite can be executed automatically: the platform backend simulates each test case, compares actual responses against expected ones, and marks each case as passed or failed.
This gives teams a lightweight but rigorous safety net for dialogue changes - no scripting required, no external testing tools, just record and replay within the platform.
Natural Language Understanding - Deeper and Broader
Expanded Entity Support: The Entities Module now supports four entity types - List, Regex, Built-In, and Trainable - each with a dedicated UI tab. Regex entities, previously limited to the Keyword Adapter, now work across all adapters. Trainable entities are a brand-new type for NLU vendors that support them.
Training Example Hygiene: Automatic trimming and lowercase normalisation prevent duplicate training examples and keep NLU models clean.
System Integration
NLP.js Adapter: A new cost-effective NLU adapter allows deployments without third-party NLU costs. NLP.js can run on-premises with no specialized hardware, making it ideal for GDPR-sensitive or budget-conscious installations.
Sensitive Data Masking: New settings on User Action nodes control masking of sensitive input phrases and output variables. When enabled, sensitive data is replaced by asterisks in the transcript - essential for PCI-DSS and GDPR compliance in voice and chat channels.
Reporting
Redesigned Built-In Reports: All three built-in reports have been rebuilt using the Visual Page Designer, enabling synchronized report updates across customer installations without manual intervention.
Looking Ahead in the Jaggle.Conversational AI Pipeline: The roadmap continues at full pace: dialogue transfer with reusable modules, GraphRAG for richer contextual retrieval, end-of-conversation surveys for multi-dimensional feedback, user-question analytics with topic clustering and knowledge-gap detection and research into next-generation voice technologies, including Diffusion LLMs and Gemini native audio models. Stay tuned.
By the Numbers
What This Means for Our Clients
With 2026.1, Jaggle is no longer just a low-code development platform - it is an AI-augmented enterprise application factory. From natural-language SQL queries to automated documentation, from multi-LLM orchestration to WCAG-compliant accessibility, this release positions Jaggle as a platform that enterprises can trust for their most demanding digital transformation programs.
For Jaggle.Conversational AI, the Asterisk integration and NLP.js adapter open up new deployment models - voice IVR without third-party lock-in, on-premises NLU for regulated environments, and a clear path toward LLM-native dialogue management.
Congratulations to every team member who contributed to this milestone. Onwards to the next release!
A packed release that raises the bar on AI, security, developer experience, and enterprise readiness across both platforms.
It has been an intense development cycle, and the teams have delivered. Jaggle 2026.1 and Jaggle.Conversational AI 2026.1 are now Generally Available, bringing dozens of new capabilities that touch virtually every area of the platform.
Here is a tour of the highlights.
Jaggle 2026.1 – The Headlines
AI That Does More of the Heavy Lifting: This release dramatically expands what the AI layer can do for developers and business users alike:
Page Development Helper: An AI-powered coding companion that assists in React page development with intelligent code suggestions and automated error detection. Developers can now scaffold and refine pages faster than ever.
Auto Documentation: AI-driven documentation generation that covers Forms, Stored Procedures and Functions. The days of manually writing boilerplate docs are behind us.
Literals Translator: Need your application in French, German, or Arabic? The new AI-powered translator leverages LLM providers to deliver context-aware translations of application literals across multiple languages.
SQL NLP Wizard (“Sorcerer”): A natural-language-to-SQL conversion system that lets non-technical users describe their data needs in plain language and get optimized, syntactically correct queries in return – complete with preview, explanation, and iterative refinement.
AI Usage Report: A full reporting tool that tracks LLM provider usage with cost estimates and provider-specific metrics, giving organizations clear visibility into their AI consumption.
AutoML: The machine-learning engine is now live, with ready-to-use classification and analytics capabilities built into the platform.
Anthropic Claude Integration: A new AI adapter brings Claude, Anthropic’s state-of-the-art LLM, into the Jaggle ecosystem, expanding the multi-LLM strategy alongside existing providers.
Dynamic LLM Switching: End users can now switch between LLM models on the fly directly from the AI Companion interface – no configuration changes needed.
Workflow & Business Process Enhancements
Swim Lanes: A visual workflow element that clearly separates participants, departments, or roles within a business process, making process ownership instantly visible.
Enhanced Declaration Access: Quick-access buttons for REST Web Service and Stored Procedure declarations streamline day-to-day workflow configuration.
Send Report Action: A new workflow action that enables automated report distribution directly from workflow definitions.
Data Model & Security
Extended Views Enhancement: Filter conditions (SQL WHERE clauses) can now be applied per joined table in extended views, delivering more flexible and performant data retrieval.
Test Data Generation: Automated creation of test datasets accelerates development and QA cycles.
Security Plan Visualization: A new comparison view shows permission differences by role and entity at a glance, making security audits far more intuitive.
Role Clone: Security plans can now be cloned across roles, dramatically reducing setup time for complex RBAC hierarchies.
Page Designer & UX
Enhanced Data Controls: Improved controls for consuming data from REST APIs and Stored Procedures, including data controls without direct data-model binding.
User @Mentions: Application users can now mention colleagues with “@”, triggering in-app notifications – including group mentions and rich-text formatting.
SpreadJS Integration: Full spreadsheet capabilities are now embedded in the page designer for advanced data entry and manipulation scenarios.
Enhanced Versioning with Snapshots: Every page save now generates an automated snapshot with retention policies and one-click rollback, providing a comprehensive audit trail of design evolution.
Platform & Infrastructure
.NET 8.0 Upgrade: The platform core, along with Kendo UI and DevExpress components, has been upgraded to the latest framework version for improved performance and long-term support.
WCAG 2.1 & 2.2 Conformance: Jaggle now supports ADA compliance and WCAG 2.1/2.2 standards via integrated accessibility tooling, opening doors for public-sector and regulated-industry deployments.
Enterprise Authentication: Full support for OAuth 2.0, JWT, and OpenID Connect authentication methods, meeting the security requirements of diverse enterprise environments.
Multi-Language Menu Titles: Menu builder titles can now be overridden in multiple languages, supporting true multi-locale application shells.
RAG (Retrieval-Augmented Generation)
Microsoft Teams Integration: The RAG module now connects directly to Microsoft Teams, enabling conversational AI assistants within the collaboration tools teams already use.
REDIS BM25 Scoring: A new caching and scoring layer using REDIS with BM25 ranking improves retrieval accuracy and reduces LLM token consumption.
Qdrant Vector Database Support: Jaggle’s RAG engine now integrates with Qdrant, a high-performance, open-source vector database built in Rust. Qdrant brings purpose-built vector similarity search with HNSW indexing, hybrid retrieval (combining dense semantic vectors with sparse keyword-based vectors), and advanced payload filtering – all in a single query. Its binary quantization can reduce memory usage by up to 64√ó while maintaining search quality, and its Rust-native architecture delivers up to 4√ó higher throughput compared to alternatives. For enterprise deployments, Qdrant supports on-premises, hybrid-cloud, and fully managed SaaS models with multi-tenant isolation, making it a natural fit for regulated industries. Combined with Jaggle’s existing REDIS BM25 scoring layer, the Qdrant integration gives organizations a best-of-both-worlds retrieval stack: fast lexical matching for exact-term queries and deep semantic search for meaning-based retrieval.
Jaggle.Conversational AI 2026.1 – The Headlines
The Jaggle.Conversational AI platform receives significant upgrades, focused on NLU depth, voice channel readiness, and operational independence.
Voice Channel – Barge-In and Advanced Voice Settings:
One of the most anticipated features in this release is full barge-in support. Users can now interrupt the system mid-utterance – just as they would in a natural phone conversation – and the platform will immediately stop playback, capture the new input, and route the dialogue accordingly. This is a game-changer for voice IVR deployments where callers know what they want and shouldn’t have to wait for a menu to finish playing.
Barge-in is fully configurable at multiple levels: a global default can be set per dialogue, and then overridden on a per-response basis within the Dialogue Designer. The same applies to DTMF support, Speech-to-Text language, and Text-to-Speech language – all of which are now first-class settings in the voice tab of every response node. This granular control means a single dialogue can switch languages mid-conversation, enable or disable barge-in contextually (e.g., disable it during a critical confirmation step), and accept keypad input where appropriate.
The underlying integration with Asterisk + UniMRCP is now live, with the UniMRCP plugin supporting dynamic parameterization at node level. Combined with barge-in, platform is now a production-grade voice platform.
LLM-Based NLU – Multilingual Without Training Data
This release introduces a fundamentally new approach to Natural Language Understanding: an LLM-based NLU adapter that replaces traditional intent-classification models with LLM prompts. The practical impact is enormous – organizations can now deploy multilingual dialogues without building and maintaining separate training datasets for each language. The LLM interprets user input in any supported language, maps it to the correct intents and entities, and returns structured results that the dialogue engine processes as usual.
For clients operating across multiple geographies or serving multilingual customer bases, this eliminates one of the biggest bottlenecks in conversational AI deployment: the cost and time of curating training examples in every target language.
The adapter works alongside the existing NLU options (including the new NLP.js adapter for on-premises, cost-effective deployments), so teams can choose the right engine for each use case – LLM-based for rapid multilingual rollout, traditional NLU for high-volume low-latency scenarios, or a hybrid approach.
Conversation Simulation – AI-Powered Dialogue Testing at Scale
Testing conversational flows has always been manual, slow, and prone to blind spots. The new Conversation Simulation feature leverages LLMs to generate realistic test conversations automatically, based on use-case descriptions and scenario parameters. It operates in two modes:
Non-Interactive Mode
The LLM generates complete conversations (both user turns and expected system responses), which are then used to automatically create new intents and training examples. This is a powerful bootstrapping tool for new dialogues – describe the use case, and the system produces dozens of realistic conversation paths that enrich the training data.
Interactive Mode
The LLM plays the role of the end user and interacts in real time with the platform backend. This is true automated regression testing: the simulated user probes the dialogue with varied phrasing, edge cases, and unexpected inputs, and the system records whether the dialogue responded correctly. When a dialogue is updated, teams can re-run simulations to catch regressions before they reach production.
Automated Testing – Record, Replay, Validate
Complementing conversation simulation is a new automated testing utility for deterministic regression testing. Administrators can “record” test cases using the existing Dialogue Tester, capturing input utterances and their expected outputs. When changes are made to a dialogue, the recorded test suite can be executed automatically: the platform backend simulates each test case, compares actual responses against expected ones, and marks each case as passed or failed.
This gives teams a lightweight but rigorous safety net for dialogue changes – no scripting required, no external testing tools, just record and replay within the platform.
Natural Language Understanding – Deeper and Broader
Expanded Entity Support: The Entities Module now supports four entity types – List, Regex, Built-In, and Trainable – each with a dedicated UI tab. Regex entities, previously limited to the Keyword Adapter, now work across all adapters. Trainable entities are a brand-new type for NLU vendors that support them.
Training Example Hygiene: Automatic trimming and lowercase normalisation prevent duplicate training examples and keep NLU models clean.
System Integration
NLP.js Adapter: A new cost-effective NLU adapter allows deployments without third-party NLU costs. NLP.js can run on-premises with no specialized hardware, making it ideal for GDPR-sensitive or budget-conscious installations.
Sensitive Data Masking: New settings on User Action nodes control masking of sensitive input phrases and output variables. When enabled, sensitive data is replaced by asterisks in the transcript – essential for PCI-DSS and GDPR compliance in voice and chat channels.
Reporting
Redesigned Built-In Reports: All three built-in reports have been rebuilt using the Visual Page Designer, enabling synchronized report updates across customer installations without manual intervention.
Looking Ahead in the Jaggle.Conversational AI Pipeline: The roadmap continues at full pace: dialogue transfer with reusable modules, GraphRAG for richer contextual retrieval, end-of-conversation surveys for multi-dimensional feedback, user-question analytics with topic clustering and knowledge-gap detection and research into next-generation voice technologies, including Diffusion LLMs and Gemini native audio models. Stay tuned.
By the Numbers
What This Means for Our Clients
With 2026.1, Jaggle is no longer just a low-code development platform – it is an AI-augmented enterprise application factory. From natural-language SQL queries to automated documentation, from multi-LLM orchestration to WCAG-compliant accessibility, this release positions Jaggle as a platform that enterprises can trust for their most demanding digital transformation programs.
For Jaggle.Conversational AI, the Asterisk integration and NLP.js adapter open up new deployment models – voice IVR without third-party lock-in, on-premises NLU for regulated environments, and a clear path toward LLM-native dialogue management.
Congratulations to every team member who contributed to this milestone. Onwards to the next release!
