AI 🔮 Command¶

The ai command works with Agents with input from the CLI:

$ enola ai --llm=echo:/ --prompt="hello, world"
hello, world

$ enola ai --llm="google://?model=gemini-2.5-flash-lite" --prompt="hello, world"
Hello, world! How can I help you today?

You, of course, also use the Chat Web UI or the Console Chat to interact with Agents.

All of these 3 commands (ai, chat, and server) support the following CLI flags / options.

Agents¶

--agents loads AI Agents. It is possible to specify multiple agents by repeating the flag.

If a “short name” ([a-zA-Z0-9\-]+) is given, e.g. --agents=weather, then this is implicitly mapped to https://raw.githubusercontent.com/enola-dev/$NAME-agent/refs/heads/main/enola.agent.yaml.

Otherwise, it loads the agent definition from the given local file (e.g. --agents=dir/example.yaml) or fetches a non-file remote URL (e.g. --agents=https://example.com/agent.yaml).

See the tutorial for usage examples.

Default Agent¶

--default-agent specifies which of the loaded AI Agents to use if none is otherwise selected.

LLM¶

--lm needs to be a valid AI LM URI.

It is optional, because Agents can set this via model: as well.

Prompt¶

--prompt is the input prompt to the LLM / Agent.

Attach¶

--attach allows attaching files to the LLM / Agent prompt. It can be repeated to attach multiple files (e.g. --attach=image.png --attach=document.pdf).

The files are referenced by URL, and support relative URLs for local files and fetching remote URLs with various schemes.

Example with image attachment:

$ enola ai --llm="google://?model=gemini-2.5-flash-lite" \
    --prompt="What do you see in this image?" \
    --attach=test/mystery.jpeg

Supported image formats, understanding, token cost, maximum number of images per request and their maximum sizes obviously depend on the LLM used:

Gemini

MCP¶

--mcp enables MCP for the Agent(s).