Skip to content

Commit

Permalink
Updated
Browse files Browse the repository at this point in the history
Updated
  • Loading branch information
v-ypalikila committed Jul 4, 2024
1 parent ee333ba commit 7cd6da0
Showing 1 changed file with 5 additions and 16 deletions.
21 changes: 5 additions & 16 deletions docs/custom-and-declarative-copilot.md
Original file line number Diff line number Diff line change
Expand Up @@ -9,30 +9,19 @@ ms.date: 06/27/2024

# Custom and declarative copilots

Microsoft Copilot is designed to be a versatile platform that integrates with Microsoft 365 and offers a range of functionalities for developers to build upon. It includes features like Graph Connectors, API Plugins, and the ability to create Copilot Extensions.
Custom and declarative copilots are specialized tools within the Copilot for Microsoft 365 framework, designed to streamline and personalize your workflow.

Copilots are natural language assistants that can help with creative tasks, generate insights, and execute automated workflows. Copilots are composed Of workflows, actions, knowledge, and triggers, powered by one or more foundation models and an orchestrator that oversees and synchronizes operations Of the copilot, Copilots can power gem-Al capabilities in apps, web services and can be published as Copilot extensions to extend and customize Microsoft Copilot.
Custom copilots are for those who need a tailored AI experience. They use a unique Large Language Model or Orchestrator to fit specific requirements, perfect for teams with coding skills. With the Copilot SDK, you can create detailed instructions, override skills, and even develop custom plugins.

Let's understand more about each item that makes a copilot:
Declarative copilots are user-friendly and require less coding knowledge. They work by combining Copilot with custom instructions to focus on particular areas or roles within your organization. They're built using Copilot Studio, allowing you to easily extend Copilot's capabilities with a few simple commands.

* **Workflows**: Workflows guide copilot's behavior, scope, and functionalities.
* **Triggers**: Triggers prompt copilot to take proactive actions and automate tasks.
* **Actions**: Actions are copilot can perform on behalf Of users external systems.
* **Knowledge**: Knowledge and RAG techniques are used by copilot to provide relevant, contextual responses.
* **Orchestrator**: Orchestrator synchronizes processing of user prompts across foundation models, knowledge, and actions.
* **Foundation models**: AI models that can perform a wide range of tasks.

Microsoft 365 Copilot offers two distinct pathways for enhancing productivity and streamlining workflows: Declarative Copilots and Custom Copilots. Understanding the differences between these two can empower developers and organizations to make informed decisions tailored to their unique needs.

Declarative Copilots are a type of extension that combines Microsoft Copilot with custom instructions. They allow for a more personalized experience by adding domain knowledge and focusing on specific data sources or user roles within an organization. The Low-Code Approach Declarative Copilots are designed for ease of use, allowing users to converse with, build, or extend functionalities using low to pro code models. These copilots are crafted declaratively with Copilot Studio, utilizing custom instructions and, optionally, training on organizational data.

Custom Engine Copilots are another extension type that allows for the use of a custom Large Language Model (LLM) or Orchestrator, which is useful for scenarios requiring specific AI stacks or for publishing copilots to external users not on Microsoft. The Pro-Code Solution Custom Copilots cater to organizations with their own Copilot-like capabilities. Custom Engine Copilots are built using the Copilot SDK, which enables coding of instructions and action calls. Custom Copilots offer a higher degree of flexibility, allowing for the development of overrides to existing horizontal skills and the packaging of custom plugins, including message extensions and adaptive cards.
Both types are designed to enhance your productivity by fitting into the Copilot for Microsoft 365 ecosystem, giving you the power to choose the right tool for your needs.

## When to Choose Which?

The decision to use a declarative or custom copilot hinges on several factors:

|Declarative Copilots |Custom Engine Copilots |
|Declarative copilots |Custom Engine copilots |
|---------|---------|
|Use Copilot's orcestrator and model | Use custom orchestrator and model |
|For scenarios, which require a high level of focus or specialization.| For scenarios demanding extensive customization and control over the AI’s behavior|
Expand Down

0 comments on commit 7cd6da0

Please sign in to comment.