Atoms

Capability to invoke a large language model for text generation. Required by all inference plugins.

Read files from the filesystem without write access. Required by any plugin that loads config or data files.

Standard interface contract for LLM inference plugins in the Olympus runtime. Required methods: invoke, stream, count_tokens, supports_tools.

Fires when a new session or conversation begins. Plugins use this to load state, verify credentials, or inject context.

Read-only access to GitHub repositories, issues, and PRs via the GitHub REST API