Prompted for entering API keys when using local models which require no keys. #509
Description
gptscript version v0.8.2+1b5e068f
Steps to reproduce the problem:
-
Host a LLM model locally
-
Execute gptscript using this model.
-
User is prompted to enter API keys for the model which is incorrect.
gptscript --disable-cache --default-model 'Llama-3-8b-function-calling-alpha-v1.gguf from http://localhost:1234/v1' examples/bob.gpt
12:15:55 WARNING: Changing the default model can have unknown behavior for existing tools. Use the model field per tool instead.
12:15:55 started [main]
? Please provide your API key for http://localhost:1234/v1
12:15:58 Saved API key as credential http://localhost:1234/v1
12:15:58 sent [main]
12:16:00 ended [main] [output=starttoolcall{\"name\": \"bob\", \"arguments\": \"{\\\"question\\\":\\\"How are you doing?\\\"}\"}endtoolcall]
12:16:00 usage [total=220] [prompt=193] [completion=27]
OUTPUT:
starttoolcall{"name": "bob", "arguments": "{\"question\":\"How are you doing?\"}"}endtoolcall
- This behavior is also seen when listing models
% gptscript --list-models http://localhost:1234/v1
? Please provide your API key for http://localhost:1234/v1
11:40:22 Saved API key as credential http://localhost:1234/v1
- Not entering anything when prompted for credentials (from step 3 and step 4), results in a credential entry being created after which I am not able to list credentials.
% gptscript credentials
2024/06/18 11:41:58 failed to list credentials: unexpected end of JSON input
Deleting the credential created allows for existing credentials to be listed.
%gptscript credential delete http://localhost:1234/v1
% gptscript credential
braveSearchAPIKey
sys.openai