Changelog v1.3.0

Any LiteLLM-compatible AI provider (api_base / api_key_env), /model profile switcher, and loki chat as default screen.

v1.3.0Sep 30, 2026Syncing from GitHub…

Added

Support for any LiteLLM-compatible AI provider via api_base/api_key_env, not just OpenAI, Anthropic, and Gemini.
Scaffolded local-model examples (Ollama, etc.) in the generated config.yaml.
loki chat is now the default screen when running loki with no subcommand.
/model command to switch or add AI models live, with a persisted model registry.
Boxed chat input with live command autocomplete (prompt_toolkit), replacing the plain loki > prompt.

Chat input now renders as a real bordered box instead of two floating print lines.
Autocomplete dropdown now shows and correctly matches slash commands.
/model config leak: an explicit --model no longer inherits an unrelated api_base/api_key_env from config.yaml.
/model reserved names (add, remove, reset, default) are now rejected instead of becoming unreachable profiles.
/model profile lookups are now case-insensitive, matching the completer's suggestions.

bash

uv tool upgrade loki-chaos-agent
# or directly from LOKI:
loki update