Prompts

Overview

The prompts command group manages prompt templates and their versions. Prompts store LLM message templates with variable substitution, tool configurations, and provider settings.

Command	Description	Client Method
`prompts list`	List all prompts	`get_all_prompts`
`prompts get`	Get a prompt by name	`get_prompt`
`prompts versions`	List all versions of a prompt	`get_all_prompt_versions`
`prompts create`	Create a prompt or new version	`create_prompt`
`prompts update`	Update prompt metadata	`update_prompt`
`prompts delete`	Delete a prompt	`delete_prompt`

`prompts list`

arize_toolkit prompts list

Lists all prompts in the current space.

Example

$ arize_toolkit prompts list
                        Prompts
┌──────────┬───────────────────┬──────────────┬────────────┐
│ id       │ name              │ description  │ createdAt  │
├──────────┼───────────────────┼──────────────┼────────────┤
│ p1       │ qa-system-prompt  │ QA pipeline  │ 2025-01-10 │
│ p2       │ summarizer        │ Summarize... │ 2025-02-01 │
└──────────┴───────────────────┴──────────────┴────────────┘

`prompts get`

arize_toolkit prompts get NAME

Retrieves full details for a prompt including messages, provider settings, and tools.

Arguments

NAME — The prompt name.

Example

arize_toolkit --json prompts get "qa-system-prompt"

`prompts versions`

arize_toolkit prompts versions NAME

Lists all versions of a prompt with their commit messages.

Arguments

NAME — The prompt name.

Example

arize_toolkit prompts versions "qa-system-prompt"

`prompts create`

arize_toolkit prompts create NAME --messages JSON [OPTIONS]

Creates a new prompt or a new version if a prompt with the same name already exists.

Arguments

NAME — The prompt name.

Required Options

--messages — A JSON array of message objects, either inline or from a file using @filepath.

Optional Options

--commit-message — Commit message. Defaults to "created prompt".
--description — Prompt description.
--tag — Tags (repeatable).
--provider — LLM provider: openAI, awsBedrock, azureOpenAI, vertexAI, custom.
--model-name — LLM model name (e.g. gpt-4o).
--input-variable-format — Variable format: f_string, mustache, none.

Message Format

Messages should be a JSON array:

[
  {"role": "system", "content": "You are a helpful assistant."},
  {"role": "user", "content": "{question} {context}"}
]

Examples

# Inline JSON
arize_toolkit prompts create "qa-prompt" \
    --messages '[{"role":"system","content":"Answer questions."},{"role":"user","content":"{question}"}]' \
    --provider openAI \
    --model-name gpt-4o

# From a file
arize_toolkit prompts create "qa-prompt" \
    --messages @messages.json \
    --description "QA system prompt" \
    --tag production --tag v2

`prompts update`

arize_toolkit prompts update NAME [--new-name NAME] [--description TEXT] [--tag TAG]...

Updates a prompt's metadata (name, description, tags). This does not create a new version — use prompts create with the same name for that.

Arguments

NAME — The current prompt name.

Options

--new-name (optional) — Rename the prompt.
--description (optional) — Updated description.
--tag (optional) — Updated tags (repeatable). Replaces existing tags.

Example

arize_toolkit prompts update "qa-prompt" --new-name "qa-prompt-v2" --tag production

`prompts delete`

arize_toolkit prompts delete NAME [--yes]

Deletes a prompt and all its versions. Prompts for confirmation unless --yes is passed.

Arguments

NAME — The prompt name.

Options

--yes — Skip confirmation.

Example

arize_toolkit prompts delete "old-prompt" --yes