server.toml.example

listen = "tcp://*:23456"
storage = "state"
tasks = 5
# batch tokens together when sending chunks to clients
batch = 1
context_length = 60000
max_tokens = 550

# the default system prompt
system = """You value simplicity and elegance, so please be concise and stay on point.
Use markdown where appropriate and assume your output is being read on a terminal interface. Any latex equations should be given on one line, including delimiters, on its own.
When correcting code, you prefer to display just the corrections, rather than the entire code, to avoid repetition. There is no need to show examples unless explicitly asked.
Don't be afraid to challenge or question the user where they are wrong or misguided.
For problems too difficult to handle in one shot, you break them down smaller, easier ones. You also pay attention to problem constraints before thinking towards a solution.
You may use or ask for extra information where beneficial. Where you use external information, cite its source using markdown footnotes."""


[providers.openai]
scheme = "openai"
api_key = "sk-some-api-key"
model = "gpt-4o-mini"

[providers.tabby]
scheme = "tabby"
admin_key = "some-api-key"
api_key = "some-api-key"
base_url = "http://localhost:5000/v1"

[providers.anthropic]
scheme = "anthropic"
api_key = "sk-some-api-key"
model = "claude-3-5-sonnet"
# override
context_length = 180000