diff --git a/docs/2-Concepts/0-Core Concepts.md b/docs/2-Concepts/0-Core Concepts.md index b0d0df8f451b..9adf0509e4bb 100644 --- a/docs/2-Concepts/0-Core Concepts.md +++ b/docs/2-Concepts/0-Core Concepts.md @@ -3,7 +3,7 @@ AGiXT supports different interaction modes with the agent. These features are li ## Instruction Think of an instruction as giving a simple, specific command to your pet. For instance, you might tell your dog to "fetch the ball". The dog knows what each of these words means, and it can perform this single task for you. In the world of AI, an instruction could be something like "Translate this sentence into Spanish". It's a single, straightforward command that the AI can execute immediately. -Instruction uses commands. To know more about the commands refer to [Extension Commands](https://josh-xt.github.io/AGiXT/2-Concepts/4-Extension%20Commands.html). +Instruction uses commands. To know more about the commands refer to [Extension Commands](https://josh-xt.github.io/AGiXT/2-Concepts/05-Extension%20Commands.html). ## Smart Instruct "Smart Instruct" is an advanced feature of the AGiXT software that integrates artificial intelligence designed to comprehend, plan, and execute tasks efficiently. It starts with an instruction from the user, which the system breaks down and analyzes. Leveraging the power of web search, the system seeks out relevant information to understand the task at hand thoroughly. It then formulates multiple strategies, evaluates them for any potential issues, and fine-tunes the best possible solution. This solution is then executed via a series of commands, with each output monitored for accuracy. The process concludes with a detailed report on the task's execution, providing a comprehensive overview of the journey from instruction to completion. "Smart Instruct" truly embodies the synergy of artificial intelligence and the internet, offering an intelligent, adaptive, and reliable approach to task completion. diff --git a/docs/2-Concepts/01-Processes and Frameworks.md b/docs/2-Concepts/01-Processes and Frameworks.md index 4ed3f8b95ae0..074ff76353ed 100644 --- a/docs/2-Concepts/01-Processes and Frameworks.md +++ b/docs/2-Concepts/01-Processes and Frameworks.md @@ -1,25 +1,25 @@ Since AGiXT is designed to build customizable and extensible AI agents, there are different services to satisfy the goal. In the web UI of AGiXT, you will see the services that include Agent Management, Agent Training, Agent Interactions, Memory Management, Prompt Management, and Chains Management. ## Agent Management -This service is dedicated to managing the agent such as agent settings that you can find more details in [Agents](https://josh-xt.github.io/AGiXT/2-Concepts/3-Agents.html) section. The agent settings part modifies the settings in text generation models based on the available [Providers](https://josh-xt.github.io/AGiXT/2-Concepts/2-Providers.html). Moreover, you can find the file system settings that enable the agent to read through different files or connect to different services such as GitHub, Whisper, Dalle, Discord, etc. Applying these settings to the agent makes the agent more aligned with the user's expectations. +This service is dedicated to managing the agent such as agent settings that you can find more details in [Agents](https://josh-xt.github.io/AGiXT/2-Concepts/03-Agents.html) section. The agent settings part modifies the settings in text generation models based on the available [Providers](https://josh-xt.github.io/AGiXT/2-Concepts/02-Providers.html). Moreover, you can find the file system settings that enable the agent to read through different files or connect to different services such as GitHub, Whisper, Dalle, Discord, etc. Applying these settings to the agent makes the agent more aligned with the user's expectations. ## Agent Training -The agent can train based on the additional resources provided to it. These resources include websites, different file formats, text (i.e., a pair of question and answer that the user provides), and GitHub repositories. You can find more details about these training modes in [Agent Training](https://josh-xt.github.io/AGiXT/2-Concepts/8-Agent%20Training.html) section. +The agent can train based on the additional resources provided to it. These resources include websites, different file formats, text (i.e., a pair of question and answer that the user provides), and GitHub repositories. You can find more details about these training modes in [Agent Training](https://josh-xt.github.io/AGiXT/2-Concepts/09-Agent%20Training.html) section. In addition to the fundamental settings for agent training, the user can provide more advanced options such as predefined memory collections that can be used by developers. These options include storing the web search results, positive/negative feedback, etc. ## Agent Interactions -There are different modes to interact with the agent including Chat, Chains, Prompt, and Instruct. You can find more information about each of these modes and how they differ from each other in [Core Concepts](https://josh-xt.github.io/AGiXT/2-Concepts/0-Core%20Concepts.html). Moreover, you can find more details and advanced options for these modes in the [Agent Interactions](https://josh-xt.github.io/AGiXT/2-Concepts/9-Agent%20Interactions.html) section. +There are different modes to interact with the agent including Chat, Chains, Prompt, and Instruct. You can find more information about each of these modes and how they differ from each other in [Core Concepts](https://josh-xt.github.io/AGiXT/2-Concepts/0-Core%20Concepts.html). Moreover, you can find more details and advanced options for these modes in the [Agent Interactions](https://josh-xt.github.io/AGiXT/2-Concepts/10-Agent%20Interactions.html) section. ## Memory Management -The agent can search for a query based on the memory and the documents it was trained on, in Agent Training. You can find more details in the Memory Management section of [Agent Training](https://josh-xt.github.io/AGiXT/2-Concepts/8-Agent%20Training.html). +The agent can search for a query based on the memory and the documents it was trained on, in Agent Training. You can find more details in the Memory Management section of [Agent Training](https://josh-xt.github.io/AGiXT/2-Concepts/09-Agent%20Training.html). Advanced options for memory management include predefined memory collections, number of the output memories, and the minimum relevance score for a memory to be returned. ## Prompt Management -This feature helps in managing the prompts when working with the agents. Prompt management assists in creating, editing, and deleting prompts, or even changing them to conversations. You can learn more about it in [Prompts](https://josh-xt.github.io/AGiXT/2-Concepts/5-Prompts.html) section. There are also more details about the conversations in [Conversations](https://josh-xt.github.io/AGiXT/2-Concepts/7-Conversations.html). +This feature helps in managing the prompts when working with the agents. Prompt management assists in creating, editing, and deleting prompts, or even changing them to conversations. You can learn more about it in [Prompts](https://josh-xt.github.io/AGiXT/2-Concepts/06-Prompts.html) section. There are also more details about the conversations in [Conversations](https://josh-xt.github.io/AGiXT/2-Concepts/07-Conversations.html). ## Chains Management -This service emphasizes the chain feature of the agents. You can learn more about what they are and how you can use predefined variables in AGiXT in [Chains](https://josh-xt.github.io/AGiXT/2-Concepts/6-Chains.html). +This service emphasizes the chain feature of the agents. You can learn more about what they are and how you can use predefined variables in AGiXT in [Chains](https://josh-xt.github.io/AGiXT/2-Concepts/07-Chains.html). In addition to the details listed, there are some advanced options including but not limited to the number of generated results, the number of attempts the agent makes, and the predefined memory collections.