Skip to main content

​
What is Daft Cloud?

  • Serverless platform for experimenting, deploying, and operating AI pipelines with zero infrastructure headaches.
  • Managed LLMs/VLMs and compute workers that scale automatically as your data grows.
  • Production-grade auth, observability, retries, and versioning for minimal overhead.

​
Key features

  • Serverless Execution - autoscales to your workloads with minimal engineering operations
  • Managed Model Inference - use LLMs/VLMs without wrangling rate limits or GPUs
  • Managed Data Connectors - high throughput I/O to read and write from your cloud storage
  • Built-in Versioning - runs and pipelines versioned with git
  • Graceful Degradation - handles errors gracefully without crashing the entire pipeline

​
Getting started

  1. Create a Project: you may simply name your project “hello world”
  2. Create your first Daft Run: enter the “hello world” project and create a new Run!
We are going to create a run off of this function in our public examples repository. This is a simple run which defines a prompt and embedding generation over 3 small rows of text.
  • Source: use the Public URL to the repo at https://github.com/Eventual-Inc/daft-examples
  • Entrypoint: choose Function and point it at hello_world.py for Module and example for function
  • Arguments: leave this blank for now because our hello_world.py:example function doesn’t take any arguments!
  1. Watch it go zoom: watch as Daft Cloud blazes through your workload, running inference/prompts on every row as configured in your example function. Your end results are now downloadable as a JSON file!

​
Next Steps

  • Runs - Execute and monitor your Daft workflows
  • Python SDK - Programmatically create and manage runs
  • Secrets - Securely store credentials and sensitive configuration
  • Data Sources - Connect to AWS S3, Supabase Storage, and other data systems
  • Catalogs - Connect to Unity Catalog, Supabase Database, and other catalogs
  • API Keys - Access Daft Cloud models from your local machine
  • Daft Documentation - Learn more about the Daft library
  • [COMING SOON] Check out our other examples which run on public datasets: image embeddings, document extraction, post summarization/titling and more
  • [COMING SOON] Configure your runs to run automatically with triggers
  • [COMING SOON] Configure a live HTTP endpoint as the source of data for triggering runs