This week: Perplexity and DeepSeek models, workflow redesign based on user feedback, Playgrounds overhaul, and new Fireflies and Mixpanel integrations.
New Models: Perplexity, DeepSeek R1, and More
We've added support for Perplexity models, including Perplexity Pro Search and Sonar models. Build components that search the web and cite sources, then deploy them across chat, workflows, and APIs.
Also new:
- DeepSeek R1
- Latest Grok releases
- New Mistral models
Your prompts work across all of them. Test in Playgrounds, deploy wherever you need.
UX Overhauls: Workflows and Playgrounds
Workflows have been redesigned based on user feedback. The graph is cleaner. Nodes are easier to read. Everything takes less space while showing the same detail.
You can now test any individual node directly in the sidebar. No need to run the entire workflow to debug one step. Build faster, test faster, ship faster.
Playgrounds: Your Testing Environment
This release makes Playgrounds significantly easier to use. Playgrounds let you test different versions of a prompt, compare models and settings side-by-side, and see how components perform before you deploy them.
Save responses. Share them with your team. Track how a prompt works over time. If you're building components that need to work reliably, Playgrounds are where you prove it.
New Integrations: Fireflies + Mixpanel
Two new workflow integrations:
- Fireflies: Pull meeting transcripts directly into workflows
- Mixpanel: Connect product analytics to your AI workflows
Both available now in the workflow builder.
New Resources
We published a complete guide on building parameterized, reusable prompts. If you're still treating prompts like one-off conversations instead of functions, this will help.