aljaz/goose
1
0
Fork 0
mirror of https://github.com/aljazceru/goose.git synced 2025-12-18 06:34:26 +01:00
Code Issues Packages Projects Releases Wiki Activity

docs: updated spelling in documentation (#1028)

Browse Source
This commit is contained in:
Vincent H.
2025-02-02 11:00:38 -08:00
committed by GitHub
parent 0ca4cf01c7
commit 7679bc36c5
4 changed files with 11 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
documentation/docs/goose-architecture/goose-architecture.md
View File

@@ -5,7 +5,7 @@ sidebar_position: 1
# Goose Architecture
Goose, an open source AI Agent, builds upon the basic interaction framework of Large Language Models (LLMs), which primarily functions as a text-based conversational interface. It processes text input and generates text output. This "text in, text out" approach is enhanced with tool integrations, which allows the AI agent to complete task, creating Goose.
Goose, an open source AI Agent, builds upon the basic interaction framework of Large Language Models (LLMs), which primarily functions as a text-based conversational interface. It processes text input and generates text output. This "text in, text out" approach is enhanced with tool integrations, which allows the AI agent to complete tasks, creating Goose.
## Goose Components
Goose operates using three main components, the **interface**, the **agent**, and the **connected [extensions](/docs/getting-started/using-extensions)**.
@@ -20,7 +20,7 @@ In a typical session, the interface spins up an instance of the agent, which the
[Model Context Protocol (MCP)](https://modelcontextprotocol.io/) is an open standard that allows for interoperability between data sources and AI agents. Goose utilizes MCP to connect to [MCP systems/servers](https://github.com/modelcontextprotocol/servers?tab=readme-ov-file#model-context-protocol-servers). In Goose, these systems/servers are referred to as extensions.
Extensions expose their functionality to Goose through tools. Tools are the functions that allow extensions to perfrom specific actions, such as running commands, or performing file operations. For example, the Google Drive extension includes a tool for searching documents. That tool is what gives Goose the ability to perform that action.
Extensions expose their functionality to Goose through tools. Tools are the functions that allow extensions to perform specific actions, such as running commands, or performing file operations. For example, the Google Drive extension includes a tool for searching documents. That tool is what gives Goose the ability to perform that action.
Goose comes with a set of [built-in extensions](/docs/getting-started/using-extensions#built-in-extensions), each designed to enhance your interaction. These include tools for development, web scraping, automation, memory, and integrations with JetBrains and Google Drive. Goose also supports [connecting to external extensions](/docs/getting-started/using-extensions#adding-extensions) or [creating custom extensions](/docs/tutorials/custom-extensions) as MCP servers.
@@ -30,16 +30,16 @@ To learn more about the design and implementation of extensions and tools, refer
## Interactive Loop
![interactive loop](../assets/guides/interactive-loop.png)
Lets take a closer look at the interactive loop shown above.
Let's take a closer look at the interactive loop shown above.
1. **Human Request**
The process begins and ends with you. Once you give Goose a request, question, command, or problem to solve, the flow begins.
2. **Provider Chat**
Goose sends your request along with a list of available tools to the [LLM provider](/docs/getting-started/providers) you've connected. The provider processes it, and if necesarry, creates a tool call as part of its response.
Goose sends your request along with a list of available tools to the [LLM provider](/docs/getting-started/providers) you've connected. The provider processes it, and if necessary, creates a tool call as part of its response.
3. **Model Extension Call**
The LLM is capable of creating a tool call request but not able to execute it, that's when Goose steps in. Goose takes the tool call which is formattted in JSON, runs it, and gathers the results.
The LLM is capable of creating a tool call request but not able to execute it, that's when Goose steps in. Goose takes the tool call which is formatted in JSON, runs it, and gathers the results.
4. **Response to Model**
After executing the tool call, Goose sends the results back to the model. If more extensions are needed, those steps will repeat.
@@ -62,7 +62,7 @@ For more details on how Goose handles errors, refer to the [Error Handling in Go
While Goose is free and open source, there is typically a cost associated with LLM token usage. Everything competes for token usage including messages, tool requests, resources, file content, instructions, etc. This is where Content Revision comes into play to help reduce some of those costs. There are a few things that are done to assist with this:
* Goose summarizes with faster and smaller LLMs
* Goose includes everything versus a semantic search
* Goose uses algorithims to delete old or irrelevant content
* Goose uses algorithms to delete old or irrelevant content
* Goose will use find and replace instead of rewriting large files, use ripgrep to skip system files, and summarize verbose command outputs