User LLM providers
This file is the catalog of LLM providers configured by the user: it defines how to connect to each provider and which models it offers.
Authentication creates or updates this file at ~/.skiller/settings/providers.json. Use SKILLER_AGENT_PROVIDERS_FILE to apply another override file:
SKILLER_AGENT_PROVIDERS_FILE=~/custom-providers.json skiller run @flowsThe file must exist. Its values override the included catalog and the user configuration; omitted fields keep their previously configured values.
Built-in providers
Section titled “Built-in providers”Skiller includes these provider definitions. Configure their credentials to use them:
| Provider | Adapter |
|---|---|
MiniMax (minimax) |
openai |
Moonshot (moonshot) |
openai |
DeepSeek (deepseek) |
openai |
OpenRouter (openrouter) |
openai |
LM Studio (lmstudio) |
openai |
OpenAI Codex (codex) |
codex |
Amazon Bedrock (bedrock) |
bedrock |
See Supported LLM APIs for the API used by each adapter.
Example providers.json
Section titled “Example providers.json”This example shows the catalog of a user who has configured authentication for several providers and added local models through LM Studio:
{ "providers": { "bedrock": { "profile": "work", "timeout_seconds": 120 }, "codex": { "credentials_file": "~/.skiller/secrets/openai-codex.json", "timeout_seconds": 120 }, "deepseek": { "api_key_file": "~/.skiller/secrets/deepseek_api_key", "timeout_seconds": 60 }, "lmstudio": { "base_url": "http://127.0.0.1:1234/v1", "models": [ { "context_window_tokens": 262144, "model": "qwen3.8-27b" }, { "context_window_tokens": 262144, "model": "ornith-1.0-9b" }, { "context_window_tokens": 262144, "model": "mistralai/ministral-3-14b-reasoning" }, { "context_window_tokens": 262144, "model": "google/gemma-4-12b-qat" } ], "timeout_seconds": 120 }, "minimax": { "api_key_file": "~/.skiller/secrets/minimax_api_key", "timeout_seconds": 60 }, "moonshot": { "api_key_file": "~/.skiller/secrets/moonshot_api_key", "timeout_seconds": 60 } }}See OpenAI adapter for its provider fields, credentials, models, and options.