otty / apps

All apps/LLM

LLM

llm

A command-line utility and Python library for interacting with large language models.

Runs as llm|macOS, Linux, Windows|llm.datasette.io

Open form in Otty otty.toml

Sync rules

What travels, what stays, what never leaves

LLM in your home directory Every rule on this page, by file

· logs.db Exclude

· keys.json Never synced

· logs.db Exclude

· keys.json Never synced

Only what the catalog names. A directory that syncs brings everything inside it except what is listed as excluded here.

apps/llm/otty.tomlThe file this page is generated from

id = "llm"

name = "LLM"

homepage = "https://llm.datasette.io/"

description = "A command-line utility and Python library for interacting with large language models."

executables = ["llm"]

[dotfiles]

# https://llm.datasette.io/en/latest/setup.html — LLM stores keys, aliases,

# model options, default models, templates and logs below its per-user

# application directory: ~/Library/Application Support/io.datasette.llm/ on

# macOS and ~/.config/io.datasette.llm/ on Linux. LLM_USER_PATH can override

# these defaults and is intentionally not a fixed path here.

paths = [

"[appData]/io.datasette.llm/",

"[configHome]/io.datasette.llm/",

]

# https://llm.datasette.io/en/latest/logging.html and the official source

# (https://github.com/simonw/llm/blob/main/llm/cli.py) identify logs.db as a

# SQLite database containing prompts and responses; it is private history,

# not portable configuration.

exclude = [

"[appData]/io.datasette.llm/logs.db",

"[configHome]/io.datasette.llm/logs.db",

]

# The same official setup page documents keys.json as the stored API-key file.

secret = [

"[appData]/io.datasette.llm/keys.json",

"[configHome]/io.datasette.llm/keys.json",

]

side_effect = false

[commands]

format = 1

label = "Run a prompt"

description = "Send a prompt to a language model with llm. Tools, templates, fragments, schemas and logging options are left out."

quality = "ai-draft"

parser = { end_of_options = "--", interleave = "permute", long_equals = true, short_attached = true, short_bundles = false }

[[commands.source]]

id = "help"

kind = "help"

url = "https://raw.githubusercontent.com/simonw/llm/main/docs/help.md"

revision = "2026-10-27"

evidence = """

Usage: llm prompt [OPTIONS] [PROMPT]

-s, --system TEXT System prompt to use

-m, --model TEXT Model to use

-a, --attachment ATTACHMENT Attachment path or URL or -

-o, --option <TEXT TEXT>... key/value options for the model

--no-stream Do not stream output

-n, --no-log Don't log to database

-c, --continue Continue the most recent conversation.

--key TEXT API key to use

-u, --usage Show token usage

-x, --extract Extract first fenced code block

"""

[[commands.field]]

id = "prompt"

label = "Prompt"

type = "string"

position = 0

sources = ["help"]

[[commands.field]]

id = "model"

label = "Model"

type = "string"

option = ["--model", "-m"]

sources = ["help"]

[[commands.field]]

id = "system"

label = "System prompt"

type = "string"

option = ["--system", "-s"]

sources = ["help"]

[[commands.field]]

id = "attachment"

label = "Attachment"

description = "Path or URL."

type = "string"

option = ["--attachment", "-a"]

multiple = true

sources = ["help"]

[[commands.field]]

id = "continue"

label = "Continue last conversation"

type = "boolean"

option = ["--continue", "-c"]

sources = ["help"]

[[commands.field]]

id = "no_stream"

label = "Do not stream"

type = "boolean"

option = ["--no-stream"]

sources = ["help"]

[[commands.field]]

id = "no_log"

label = "Do not log"

type = "boolean"

option = ["--no-log", "-n"]

sources = ["help"]

[[commands.field]]

id = "usage"

label = "Show token usage"

type = "boolean"

option = ["--usage", "-u"]

sources = ["help"]

[[commands.field]]

id = "extract"

label = "Extract first code block"

type = "boolean"

option = ["--extract", "-x"]

sources = ["help"]

[[commands.task]]

id = "ask"

label = "Ask a question"

fields = ["prompt", "model", "system"]

required = ["prompt"]

[[commands.task]]

id = "code"

label = "Get a code block"

fields = ["prompt", "model", "extract"]

required = ["prompt"]

set = { extract = [true] }

[[commands.group]]

id = "basic"

label = "Basic"

fields = ["prompt", "model", "system", "attachment", "continue"]

[[commands.group]]

id = "advanced"

label = "Advanced"

fields = ["no_stream", "no_log", "usage", "extract"]

Use it with your dotfile manager

Same data, in the shape your tool already reads

Opens Otty and adds this tool to what it syncs

Opens otty://sync/add/llm. Otty ships with this catalog, so the link only has to name the tool — nothing is downloaded.

Every one of these keeps all credential files out.