Added the most basic tool support. You can now select that tools should not only be evaluated but actually be executed. The tools result is then output in the list like the results of other objectives.
It's a start.
To test you can check both checkboxes for the wikipedia tool and set the tool description to "A wrapper around Wikipedia. Useful for when you need to answer general questions about people, places, companies, facts, historical events, or other subjects. Input should be a search query. This tool will fulfill the objective "Search wikipedia for AI models". Say yes to use it."
Then have the agent work on the goal: "Search Wikipedia for AI models. If you see Tool name: Wikpedia, say yes"
The test works one time out of five, or something in that chance. More often the model will just say "Yes" and call it a day. :-)
Added the most basic tool support. You can now select that tools should not only be evaluated but actually be executed. The tools result is then output in the list like the results of other objectives.
It's a start.
To test you can check both checkboxes for the wikipedia tool and set the tool description to "A wrapper around Wikipedia. Useful for when you need to answer general questions about people, places, companies, facts, historical events, or other subjects. Input should be a search query. This tool will fulfill the objective "Search wikipedia for AI models". Say yes to use it." Then have the agent work on the goal: "Search Wikipedia for AI models. If you see Tool name: Wikpedia, say yes"
The test works one time out of five, or something in that chance. More often the model will just say "Yes" and call it a day. :-)