FLORA Builds a Creative Agent Using the Vercel AI Stack

FLORA is a platform that supports visual iteration through a digital canvas. Its recently launched creative agent, FAUNA, acts as a design partner that gathers references based on the user's idea, selects models, and automatically generates various image variations.

Existing node-based workflows required users to manually set up prompts and models, which was cumbersome. FAUNA, on the other hand, removes this setup burden and helps users focus solely on the creative process.

Professional visual workflows are much more demanding than text-based tasks. Image and video generation can take several minutes, and numerous jobs must run simultaneously to explore various angles and styles.

The FLORA team previously used LangChain and Temporal, but ran into the problem of consuming too many resources maintaining and synchronizing the two systems. To solve this, they switched to the Vercel AI Stack, gaining the following benefits:

• AI SDK and agent framework: Provides foundational functionality for model calls, tool calling, and agent logic interfaces.
• DurableAgent from the Workflow SDK: Supports durable orchestration that maintains the state of each step and retries on failure.
• Fluid compute: Cost-effectively handles many long-running image generation jobs executed in parallel.

Through this integrated stack, FLORA has reduced the cost of building infrastructure and can now immediately reflect new image models and features released every week into its service.