QUICKSILVER PRO / Setup
LiteLLM
Connect LiteLLM to QuickSilver Pro: endpoint, API key, trial model and setup notes.
Create an API key in the dashboard, then use the configuration below. The starter model is available on the API trial.
Set this in the terminal that starts the tool:
export QUICKSILVER_API_KEY='YOUR_QUICKSILVER_API_KEY'model_list:
- model_name: quicksilver-haiku
litellm_params:
model: openai/claude-haiku-5-5
api_base: https://api.quicksilverpro.io/v1
api_key: os.environ/QUICKSILVER_API_KEY
max_tokens: 4096Start with Claude Haiku 5.5 on the trial. Add credit to use Sonnet or Opus; those models are not included in the trial.
Choose a model
Haiku is the low-cost starting point. Sonnet and Opus support reasoning as well as tool calling. Choose a model for your task and check its catalog page for limits.
USD per million tokens · input / output · standard context rates
- Claude Haiku 5.5$0.025 ($0.125 >200K) / $0.125 ($0.625 >200K)
- Claude Sonnet 5.5$0.50 / $2.50
- Claude Opus 5.5$1.00 / $5.00
Before you start
Save config.yaml, then run litellm --config config.yaml. Call the local proxy using model quicksilver-haiku. The openai/ prefix selects Chat Completions; it is not part of the QuickSilver Pro model ID.
Keep API keys in your environment or private user settings. Never commit them. Agent loops and long prompts consume credit; start with a small task and an output limit.