You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You are a slightly grumpy, but generally lovable assistant designed to help the user with Kubernetes.
114
106
You were created in 2024 by @ibuildthecloud at Acorn Labs in a terrible experiment gone wrong.
@@ -118,5 +110,19 @@ Guidelines:
118
110
2. If you see something that looks wrong, ask the user if they would like you to troubleshoot the issue.
119
111
3. If the user does not specify or gives an empty prompt, ask them what they wish to do.
120
112
```
113
+
If the above tool were in a directory called `context`, it could be referenced by a tool like this:
114
+
```yaml
115
+
name: Kevin
116
+
context: ./context
117
+
118
+
...
119
+
```
120
+
When a directory is specified as a context or tool, gptscript will look for a `tool.gpt` in that directory and use it.
121
+
122
+
The above example context tool has several interesting features:
123
+
124
+
The `share context` directive is used to include additional context from other tools or files into the tool that referenced this context. This allows you to aggregate multiple smaller contexts into a single, comprehensive context. Each `share context` line points to another context provider, which can be a local file or a remote tool. The content from these shared contexts is prepended to the instruction of the calling tool, providing a richer and more detailed prompt for the LLM to work with.
125
+
126
+
The `share tools` directive is similar to the `share context` directive. Tools specified here will be made available as tools to the tool that referenced this context.
121
127
122
-
This example demonstrates how to aggregate multiple contexts and tools into a single global context, providing a robust and flexible setup for your GPTScript tools.
128
+
The `#!sys.echo` directive is a simple way to directly output plain text for the context. In this example, it outputs a description of the assistant's personality and guidelines for its behavior, which will be included in the context provided to the LLM.
0 commit comments