Tool for exploring large language model outputs on an infinite canvas. You can run prompts, compare outputs, and chain them together.
Features
Run prompts
Click anywhere on the canvas to place generations of the selected prompt. Click on other prompts to change the current selection.
Compare outputs
Click on any text or image node to see its details. Links can show relationships between chat messages or prompt versions.
Chain prompts together
Continue from any message in a conversation thread. Fork from any prompt to use different text or models. Use outputs as context for new prompts.
Creating Projects
Allows users to create projects by choosing templates and adding content. Projects can be customized based on user needs.
Editing Scenes
Enables users to edit specific scenes within a project, allowing for details to be adjusted and refined.
Explore Projects
Users can explore and access various projects for inspiration or reference, seeing different scenes users have created.
Run Prompts
Provides functionality to run specific prompts, generating text based on scenarios or input data.
Complete Outputs
Generates outputs based on user inputs and prompt settings, which can then be refined further.
Clean Team Workspace
Offers organized workspace views, allowing teams to efficiently manage and collaborate on projects.
Multimodal Playground
Dreamspace acts as a model playground for multi-modal projects, allowing the user to integrate various models such as Claude 3 Quartz and OpenAI's daII-e 3 for diverse tasks like character development and design generation.
Infinite Canvas
The application provides an infinite canvas where nodes can be created and manipulated to organize and experiment with outputs from AI models.
Spacers and Nodes
Users can create nodes for different modalities like text, audio, and images, allowing for flexible experimentation and interlinking of ideas.
Prompt Experimentation
Offers tools to adjust and refine prompts with parameters such as the number of examples, and APIs like daII-e 3 to generate content dynamically.