A single unit of process that takes users input into account and provides any form of output to satisy the user’s request.

There are 4 types of promptables to meet the user’s needs. Higher types increase the chances of fulfilling the user’s query.

​

Language Model v0

The user directly interacts with any language model to get the desired output.

  • Question Answering

  • Creative writing

  • Deep Analysis

  • Summarization

  • Code generation

​

Agent v1

An instruction-following language model combined with an agent algorithm takes the user’s query and strategically interacts with apps and knowledgebases to fulfill the desired outcome.

With reasoning abilities, it can verify the rationality of its thoughts and rethink its decisions.

It can also ask the user for clarification if needed, on the way.

  • Everything in LLM

  • App calls

  • Knowledge retrieval

  • Reasoning

  • Strategic task plan

Instruction-following language models are finetuned to abide by instructions given by the system or user. Agentic algorithms provide complex instructions for the language models to follow.

Therefore, Chat models are not supported for agents.

​

Swarm v2

Orchestrates multiple agents to work together to fulfill the user’s query.

  • Everything in Agent

  • Role playing tasks

  • Co-operative tasks

  • Organizational tasks

Swarms will be available in v2.

​

Workflow v3

A graphical, logical flow of llms, agents, swarms, knowledge and apps to fulfill complex user queries.

  • Business Process Automation

  • Controlled agents

  • Logical Decisionmaking

Workflows will be available in v3.

IntroductionModel