Blocks
The fundamental building blocks of AI workflows in Hunch
Overview
Blocks are the core components of any canvas in Hunch. They represent discrete tasks, inputs, or outputs in your AI workflow. Each block serves a specific purpose, whether it’s providing content, processing data, or generating results.
Types of blocks
There are several types of blocks in Hunch, each designed for different functions:
Content blocks
These blocks hold various types of data such as text, images, files, or documents. They can serve as inputs for AI blocks or aggregate content from other blocks.
AI blocks
AI blocks are where the magic happens. They can access a variety of AI models to perform tasks like text generation, image creation, transcription, text-to-speech conversion, and PDF comprehension.
Batch blocks
Batch blocks allow you to run enclosed blocks multiple times, using a list as input. This is particularly useful for processing large amounts of data or performing repetitive tasks efficiently.
How blocks work
Each block in Hunch is self-contained, with its own inputs, processing capabilities, and outputs. Blocks can be connected to form more complex workflows, with the output of one block serving as the input for another.
When you add a block to your canvas, you can configure it by selecting an AI model (if applicable), providing instructions or prompts, and specifying any necessary parameters. Once configured, blocks can be executed individually or as part of a larger workflow.
The power of blocks
The block-based approach in Hunch offers several advantages:
- Modularity: Complex tasks can be broken down into smaller, manageable steps.
- Flexibility: Blocks can be easily rearranged, added, or removed to refine your workflow.
- Clarity: The visual representation of blocks helps in understanding and explaining AI processes.
- Reusability: Well-designed blocks can be saved and reused across different projects.
By combining different types of blocks, you can create sophisticated AI workflows tailored to your specific needs, all without writing a single line of code.