LLM CLI Reference: Every Command, Flag, and Option

llm [OPTIONS] [PROMPT]
llm prompt [OPTIONS] [PROMPT]

# Basic prompt
llm 'Capital of France?'

# Choose a model
llm 'Capital of France?' -m gpt-4o

# Add a system prompt
llm 'Capital of France?' -s 'answer in Spanish'

# Multi-modal with an image attachment
llm 'Extract text from this image' -a image.jpg
llm 'Describe' -a https://static.simonwillison.net/static/2024/pelicans.jpg

# Pipe an image via stdin with an explicit MIME type
cat image | llm 'describe image' --at - image/jpeg

# Extract the first fenced code block from the response
llm 'JavaScript function for reversing a string' -x

llm chat [OPTIONS]

# Chat with the default model
llm chat

# Chat with a specific model and system prompt
llm chat -m gpt-4o -s 'You are a helpful assistant'

# Continue a previous conversation
llm chat --cid <conversation-id>

# Chat with a tool available
llm chat -T llm_time

llm keys [COMMAND]

llm logs [COMMAND] [OPTIONS]

llm logs list [OPTIONS]

# Show the last 3 logs (default)
llm logs

# Show all logs for a specific model
llm logs -m gpt-4o -n 0

# Full-text search
llm logs -q "pelican"

# Show the most recent conversation
llm logs --cid $(llm logs list -n 1 --json | jq -r '.[0].conversation_id')

# Filter by tool usage
llm logs --tool llm_time

# Extract schema data as JSON
llm logs --data

# Toggle logging
llm logs off
llm logs on
llm logs status

llm models [COMMAND] [OPTIONS]

llm models list [OPTIONS]

# List all default options set for any model
llm models options list

# Show defaults for a specific model
llm models options show gpt-4o

# Set a default option
llm models options set gpt-4o temperature 0.5

# Clear one option
llm models options clear gpt-4o temperature

# Clear all options for a model
llm models options clear gpt-4o

# List all available models
llm models list

# Show which model is the default
llm models default

# Set a new default model
llm models default gpt-4o

# Search for models supporting tools
llm models list --tools

# List models matching a search string
llm models list -q claude

llm templates [COMMAND]

# List all saved templates
llm templates list

# Create or edit a template
llm templates edit summarize

# Inspect a template
llm templates show summarize

# Use a template
cat file.txt | llm -t summarize

# Use a template with parameters
llm -t summarize -p voice GlaDOS

# Use a template from a URL
llm -t https://example.com/my-template.yaml

# Use a template from a file path
llm -t /path/to/template.yaml

llm schemas [COMMAND]

# List stored schemas
llm schemas list

# Show a specific schema by ID
llm schemas show abc123

# Convert a DSL definition to JSON schema
llm schemas dsl 'name, age int, bio: their bio'

# Use a schema in a prompt
llm 'Extract data' --schema 'name, age int, bio'

# Use a schema expecting multiple results
llm 'Extract all people' --schema-multi 'name, age int'

llm tools [COMMAND]

# List all available tools from plugins
llm tools list

# Use a built-in tool
llm -T llm_time 'What time is it?'

# Define and use an inline tool
llm --functions '
def multiply(x: int, y: int) -> int:
    """Multiply two numbers."""
    return x * y
' 'what is 34234 * 213345'

# Use a named plugin tool
llm -T my_tool 'Do something with this tool'

llm aliases [COMMAND]

# Create an alias
llm aliases set mini gpt-4o-mini

# Create alias using query strings
llm aliases set mini -q 4o -q mini

# List all aliases
llm aliases list

# Remove an alias
llm aliases remove mini

# Use an alias
llm -m mini 'Hello!'

llm fragments [COMMAND]

# Use a URL as a fragment in a prompt
llm -f https://llm.datasette.io/robots.txt 'explain this file'

# Set an alias for a fragment
llm fragments set mydocs ./docs.md

# Use an aliased fragment
llm -f mydocs 'summarize the documentation'

# Show the content of a fragment
llm fragments show mydocs

# List all fragments
llm fragments list

# Remove a fragment alias
llm fragments remove mydocs

# List available fragment loader prefixes (from plugins)
llm fragments loaders

llm plugins [OPTIONS]

# List installed third-party plugins
llm plugins

# Include built-in plugins
llm plugins --all

# Filter by hook implementation
llm plugins --hook register_tools

llm install [OPTIONS] [PACKAGES]...

# Install a plugin
llm install llm-anthropic

# Upgrade a plugin
llm install -U llm-anthropic

# Install a plugin in editable mode for development
llm install -e /path/to/my-plugin

# Install a pre-release version
llm install --pre llm-some-plugin

llm uninstall [OPTIONS] PACKAGES...

# Uninstall a plugin (with confirmation prompt)
llm uninstall llm-anthropic

# Uninstall without confirmation
llm uninstall -y llm-anthropic

llm embed [OPTIONS] [COLLECTION] [ID]

# Embed text and print the vector as JSON
llm embed -c "Hello world" -f json

# Embed a file and store in a collection
llm embed my-docs doc1 -i document.txt --store

# Use a specific embedding model
llm embed -m ada-002 -c "Hello world"

llm embed-multi [OPTIONS] COLLECTION [INPUT_PATH]

# Embed from a CSV file
llm embed-multi docs input.csv

# Embed from JSON via stdin
cat data.json | llm embed-multi docs -

# Embed all Markdown files in a directory
llm embed-multi docs --files docs '**/*.md' --store

# Embed from a SQL query
llm embed-multi docs --sql "SELECT id, title, body FROM posts" \
  --attach blog blog.db

llm similar [OPTIONS] COLLECTION [ID]

# Find items similar to a string
llm similar my-collection -c "I like cats"

# Find items similar to a stored embedding
llm similar my-collection 1234

# Return top 10 results in plain text
llm similar my-collection -c "machine learning" -n 10 -p

# Restrict to IDs with a specific prefix
llm similar my-collection -c "query" --prefix "docs/"

llm embed-models [COMMAND]

llm collections [COMMAND]

# List all collections
llm collections list

# Delete a collection
llm collections delete my-collection

# Print the path to the embeddings database
llm collections path

llm openai [COMMAND]

# List available OpenAI models
llm openai models

# Output as JSON
llm openai models --json