Slashspace (Formerly RabbitHoles AI) ✓

MCP Support (Beta)

You can now add MCP servers and use them with LLMs that support tool calling. Note: Straico doesn't support tool calling.

Fal AI Provider (Beta)

You can now add Fal AI Provider to generate images with Fal image models.

Retry Button Has Moved to Messages

Regenerate / retry button has been removed in favor of regenerate message button that'll be visible as an icon underneath any LLM response inside Chat nodes.

Updated Feedback and Bug Form

The feedback and bug form now automatically gets the license information and the main.log file. You can attach screenshots and screen recordings and just explain the bug or the feedback.

Version and Update UI

Previously updates happened in the background and didn't offer any view into the download process. You can now click on the version label in the app and it'll open a popup to let you download updates if available. This popup will also display the progress of the download.

Misc

• UI fixes
• Performance improvements
• Default Model settings have been moved to settings page. Set your default image, title, and chat models from settings > default models
• Fixed a bug where having custom providers still show "No AI running" popup in the canvas
• Reverted a change in auto save that was causing aggressive saves and making the app unresponsive in some cases

Image generation Nodes

Introduces Image generation nodes. You can now generate images with nodes. You can also connect other images to an image generation node to provide references to the image generation model. Supports Nano banana, Gpt Image 1, and Ideogram, etc

RAG (Retrieval Augmented Generation)

Introduced RAG mode which let’s you deal with large files, and cases when there’s a lot of nodes connected to a chat node reaching the context limits of the selected model

Search for Nodes on Canvas with AI + Keywords

As canvases got bigger with many nodes, it was difficult to find your place or look up a node. Search modal can now help you search for nodes using a local keyword search, or also using Intelligent AI search

Token summary

You can now view the token count estimation of each node, the tokens incoming into a given node as context, and also the token estimation that you’d send to the AI

Voice Input

Use voice transcription to talk to the Nodes, and get your prompts transcribed

Persona Prompts Library

Save and reuse persona prompts (system prompts or like custom GPT) and assign them to your chat nodes so that the models can follow the instructions of the prompt

Refactored Context Engine

A completely refactored context engine. This is a step forward in making the app ready for upcoming updates, like MCP clients and tool use.