Specifying a local agent model #116
Replies: 1 comment 1 reply
-
@Andreas3333 Shortly yes, it is possible to specify a local model for the agent - the tools model will be used for all agents. The example, which you've shown will select an env with local model for the agent. More detailed explanation: If you want, you cold create your own env with an agent - llama-vscode menu (Ctrl+Shift+M), select Envs... -> Add Env... , after that select the models and agent you want, and add the the env with button "Save As New Env". ![]() |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
I know agent features are in development. But is it possible to specify a local agent model in Environments? There doesn't seem to be an "agent" key in the current Environment extension configuration schema. The current Agent button on the Llama Agent page only provides 1) "lama-vscode help" 2) "default" or 3) "Last used agent" options. For me selecting "llama-vscode help" displays a "Agent llama-vscode help is selected." VSCode info message but does not start a llama-server process and running a prompt returns "No response from AI" in Llama Agent.
Also, selecting of the default environments, for example "Local, full package - medium, gpt-oss 20B (> 34 GB VRAM | HD: 20 GB)" does not include a Agent model:
I suppose that is expected given there isn't a one specified in these Environments, but it would be great if we could specify one.
Beta Was this translation helpful? Give feedback.
All reactions