What's New in Wikit Semantics Q2 2026
This new release enhances the capabilities of Semantics Chat and workflows: image understanding and generation, analysis of large documents, more direct interactions with the interface, and artifact creation.
1. Images arrive in Semantics Chat
Compatible applications can now analyze images added to a conversation.
Users can:
- attach PNG, JPG, or non-animated GIF images;
- add them from the import menu, by drag-and-drop, or from the clipboard;
- send multiple files at once;
- ask the assistant directly about a screenshot, a diagram, a photograph, or a scanned document.
Images are sent to the model along with the user's message so it can leverage its multimodal capabilities.
What this means
- Immediate analysis of visual content
- A more natural way to share a screenshot
- Support for requests combining text, documents, and images
2. Image generation and editing
Semantics Chat can now generate an image from a simple natural language description.
The user describes the desired result, and the assistant takes care of building the generation request. The image appears directly in the conversation, where it can be copied or downloaded.
It is also possible to iterate naturally on the result:
- "Make this image brighter"
- "Add a character in the foreground"
- "Turn it into an illustration"
- "Create a new version based on this image"
An imported or previously generated image can thus serve as a starting point for a new creation.
This feature is available on compatible private applications, depending on their configuration.
3. Better analysis of personal documents
The assistant can now work more efficiently with personal documents, including large ones.
Instead of loading the entire file into the model's context, it can browse through its content progressively, search for specific information, and consult only the passages relevant to the answer.
The main office formats are supported, including PDF, DOCX, TXT, CSV, and XLSX.
What this means
- Support for much longer documents
- Targeted search within a file
- More reliable answers on large content
- Fewer errors related to model context limits
The search and reading operations performed by the assistant remain visible in the conversation.
4. Actions from the chat
The assistant can also suggest actions directly within its responses. Depending on the context, the user can access relevant functions in one click:
- import a document;
- view their documents;
- create a ticket;
- view their tickets;
- contact an agent.
These shortcuts make it possible to move more naturally from conversation to the expected action, without needing to search for the feature in the interface.
5. Artifacts are now available in workflows
Previously introduced in conversational applications, artifacts can now be generated by applications built with workflows.
A workflow can produce several structured files — for example text, Markdown, HTML, or CSV — and then display them as interactive cards in Semantics Chat.
Use cases
- Generate a CSV table from a conversation
- Produce a summary in Markdown
- Build an HTML page
- Create several deliverables during the same workflow
This evolution makes it possible to combine the orchestration power of workflows with richer output that users can act on directly.
These new features are being progressively rolled out across Wikit Semantics environments. Their availability may depend on your applications' configuration.
We invite you to explore them and to contact your Customer Success Manager with any questions or for help setting them up.