-
What's new in Catalyst QuickML?
-
Last Updated : September 10, 2026
-
10 Views
-
3 Min Read
Catalyst QuickML expands its GenAI capabilities with new interaction modes, flexible deployment options, broader model support, and enhancements that simplify the development of AI-powered applications. This release introduces new ways to build, deploy, and interact with GenAI applications in QuickML.
Here’s a closer look at what’s new.
Expanded LLM support with intelligent and powerful models #
QuickML now supports GLM 4.7 Flash and ** Qwen3.6-35B-A3B - Vision Language Model**, giving developers another high-performance large language model for building AI applications.
You can use GLM 4.7 Flash across both:
-
LLM serving
-
Retrieval-augmented generation (RAG) Like other supported models, GLM 4.7 Flash can be configured using model parameters, tested interactively, and deployed through secure REST endpoints with OAuth authentication.
LLM tool calling #
Large language models become significantly more powerful when they can interact with external systems.
QuickML now supports tool calling in LLM serving. With tool calling, your AI applications can:
- Invoke APIs during conversations
- Execute business functions
- Retrieve real-time information
- Return structured responses
- Combine enterprise data with LLM reasoning
This enables developers to build intelligent assistants that don't just answer questions; they perform actions.
#
Smarter RAG with three task modes
QuickML's RAG now supports three specialized modes.
Response generation
Generate contextual answers grounded in your knowledge base.
Ideal for:
- Customer support
- Internal knowledge assistants
- Documentation chatbots
Document search
Retrieve the most relevant document chunks without generating a final answer.
Ideal for:
- Semantic search
- Downstream AI pipelines
- External processing
Agentic RAG
This is the most advanced addition in this release.
Agentic RAG introduces an intelligent reasoning layer that:
- Breaks complex questions into multiple sub-queries
- Retrieves information iteratively
- Performs multi-step reasoning
- Produces more comprehensive responses
- This enables AI assistants to handle significantly more complex enterprise queries.
Deploy GenAI as endpoints
QuickML now allows you to deploy both LLM serving and RAG configurations as GenAI endpoints.
Every endpoint includes:
- REST API
- OAuth authentication
- System prompt
- Tool definitions
- RAG and LLM serving configuration
- Document storage for RAG settings
Because endpoints are created from saved test configurations, the behavior you validate during testing is exactly what gets deployed to production.
Conversation Mode for LLM serving
LLM serving now includes a conversation mode.
Unlike traditional single-shot prompting, conversation mode preserves context across multiple interactions within a session.
This enables applications such as:
-
AI copilots
-
Customer support assistants
-
Interactive business workflows
-
Multi-turn conversations The result is more natural and context-aware AI interactions.
More powerful ML capabilities, now in QuickML #
Clustering models
Clustering groups similar data points to uncover natural patterns and segments within unlabeled data. With QuickML, businesses can use clustering for use cases such as customer segmentation, product categorization, personalization, and fraud analysis.
Anomaly Detection models
Anomaly Detection identifies data points or patterns that significantly deviate from expected behavior. QuickML can help detect anomalies across use cases such as fraud detection, predictive maintenance, cybersecurity, and operational monitoring.
Bring it all together: Build complete AI applications #
The real power of these capabilities comes from using them together. With QuickML, you can combine RAG, Agentic RAG, conversation mode, tool calling, and GenAI endpoints to build AI applications that can understand context, reason over enterprise knowledge, and take action.
Example: Build an AI-powered IT Support Assistant
Consider an IT Support Assistant that helps employees troubleshoot issues and manage support requests.
- RAG can retrieve relevant information from the company's IT documentation, such as troubleshooting guides, policies, and FAQs.
- Agentic RAG can handle more complex issues by breaking a problem into multiple queries, retrieving information iteratively, and reasoning across different documents.
- Conversation mode can maintain context across the user's messages, allowing the assistant to understand follow-up questions without requiring the user to repeat information.
- Tool calling can connect the assistant to external systems to perform actions, such as checking ticket status, retrieving system information, or creating a support ticket.
- GenAI endpoints can securely expose the configured LLM or RAG workflow to the application through REST APIs and OAuth authentication.
The same approach can power Customer Support Assistants for resolving queries and checking orders, Sales Assistants for product insights and live pricing, and HR Assistants for answering policy questions and initiating HR requests.
These are just a few ways QuickML can help build AI applications tailored to different business needs.
Build smarter AI applications with QuickML #
With expanded model support, Agentic RAG, tool calling, conversation mode, and GenAI endpoints, QuickML gives developers the building blocks to create AI applications that can understand, retrieve, reason, and act.
Explore the latest QuickML features and start building smarter AI applications today!
Leave a Reply
Your email address will not be published. Required fields are marked