tests/testthat/_snaps/btw_client.md

In btw: A Toolkit for Connecting R and Large Language Models

btw_client() chat client: works with btw.client option

Code
  print(chat)
Output
  <Chat Anthropic/claude-sonnet-4-20250514 turns=1 tokens=0/0 $0.00>
  -- system [0] ------------------------------------------------------------------
  # System and Session Context

  Please account for the following R session and system settings in all responses.

  <system_info>
  R_VERSION: R VERSION
  OS: OPERATING SYSTEM
  SYSTEM: SYSTEM VERSION
  UI: RStudio
  LANGUAGE: es
  LOCALE: C
  ENCODING: LC_CTYPE
  TIMEZONE: Europe/Madrid
  DATE: DAY OF WEEK, MONTH DAY, YEAR (YYYY-MM-DD)
  </system_info>


  # Tools

  You have access to tools that help you interact with the user's R session and workspace. Use these tools when they are helpful and appropriate to complete the user's request. These tools are available to augment your ability to help the user, but you are smart and capable and can answer many things on your own. It is okay to answer the user without relying on these tools.

  ---

  I like to have my own system prompt.

btw_client() adds btw.md context file to system prompt

Code
  print(chat)
Output
  <Chat Anthropic/claude-sonnet-4-20250514 turns=1 tokens=0/0 $0.00>
  -- system [0] ------------------------------------------------------------------
  # System and Session Context

  Please account for the following R session and system settings in all responses.

  <system_info>
  R_VERSION: R VERSION
  OS: OPERATING SYSTEM
  SYSTEM: SYSTEM VERSION
  UI: RStudio
  LANGUAGE: es
  LOCALE: C
  ENCODING: LC_CTYPE
  TIMEZONE: Europe/Madrid
  DATE: DAY OF WEEK, MONTH DAY, YEAR (YYYY-MM-DD)
  </system_info>


  # Tools

  You have access to tools that help you interact with the user's R session and workspace. Use these tools when they are helpful and appropriate to complete the user's request. These tools are available to augment your ability to help the user, but you are smart and capable and can answer many things on your own. It is okay to answer the user without relying on these tools.

  # Project Context

  * Prefer solutions that use {tidyverse}
  * Always use `=` for assignment
  * Always use the native base-R pipe `|>` for piped expressions

  ---

  I like to have my own system prompt.

btw_client() with context files: finds btw.md in parent directories

Code
  print(chat)
Output
  <Chat OpenAI/gpt-4o turns=1 tokens=0/0 $0.00>
  -- system [0] ------------------------------------------------------------------
  # System and Session Context

  Please account for the following R session and system settings in all responses.

  <system_info>
  R_VERSION: R VERSION
  OS: OPERATING SYSTEM
  SYSTEM: SYSTEM VERSION
  UI: RStudio
  LANGUAGE: es
  LOCALE: C
  ENCODING: LC_CTYPE
  TIMEZONE: Europe/Madrid
  DATE: DAY OF WEEK, MONTH DAY, YEAR (YYYY-MM-DD)
  </system_info>


  # Tools

  You have access to tools that help you interact with the user's R session and workspace. Use these tools when they are helpful and appropriate to complete the user's request. These tools are available to augment your ability to help the user, but you are smart and capable and can answer many things on your own. It is okay to answer the user without relying on these tools.

  # Project Context

  * Prefer solutions that use {tidyverse}
  * Always use `=` for assignment
  * Always use the native base-R pipe `|>` for piped expressions

  ---

  I like to have my own system prompt

btw_client() with context files: uses llms.txt in wd and btw.md from parent

Code
  print(chat_parent_llms)
Output
  <Chat OpenAI/gpt-4o turns=1 tokens=0/0 $0.00>
  -- system [0] ------------------------------------------------------------------
  # System and Session Context

  Please account for the following R session and system settings in all responses.

  <system_info>
  R_VERSION: R VERSION
  OS: OPERATING SYSTEM
  SYSTEM: SYSTEM VERSION
  UI: RStudio
  LANGUAGE: es
  LOCALE: C
  ENCODING: LC_CTYPE
  TIMEZONE: Europe/Madrid
  DATE: DAY OF WEEK, MONTH DAY, YEAR (YYYY-MM-DD)
  </system_info>


  # Tools

  You have access to tools that help you interact with the user's R session and workspace. Use these tools when they are helpful and appropriate to complete the user's request. These tools are available to augment your ability to help the user, but you are smart and capable and can answer many things on your own. It is okay to answer the user without relying on these tools.

  # Project Context

  EXTRA CONTEXT FROM llms.txt

  * Prefer solutions that use {tidyverse}
  * Always use `=` for assignment
  * Always use the native base-R pipe `|>` for piped expressions

  ---

  I like to have my own system prompt

btw_client() project vs user settings: falls through to use client settings from user-level btw.md

Code
  print(chat)
Output
  <Chat Anthropic/claude-3-5-sonnet-20241022 turns=1 tokens=0/0 $0.00>
  -- system [0] ------------------------------------------------------------------
  # System and Session Context

  Please account for the following R session and system settings in all responses.

  <system_info>
  R_VERSION: R VERSION
  OS: OPERATING SYSTEM
  SYSTEM: SYSTEM VERSION
  UI: RStudio
  LANGUAGE: es
  LOCALE: C
  ENCODING: LC_CTYPE
  TIMEZONE: Europe/Madrid
  DATE: DAY OF WEEK, MONTH DAY, YEAR (YYYY-MM-DD)
  </system_info>


  # Tools

  You have access to tools that help you interact with the user's R session and workspace. Use these tools when they are helpful and appropriate to complete the user's request. These tools are available to augment your ability to help the user, but you are smart and capable and can answer many things on your own. It is okay to answer the user without relying on these tools.

  # Project Context

  User level context

  ---

  Project level context

btw_client() project vs user settings: falls back to user client settings when project has no client

Code
  print(chat)
Output
  <Chat OpenAI/gpt-4o turns=1 tokens=0/0 $0.00>
  -- system [0] ------------------------------------------------------------------
  # System and Session Context

  Please account for the following R session and system settings in all responses.

  <system_info>
  R_VERSION: R VERSION
  OS: OPERATING SYSTEM
  SYSTEM: SYSTEM VERSION
  UI: RStudio
  LANGUAGE: es
  LOCALE: C
  ENCODING: LC_CTYPE
  TIMEZONE: Europe/Madrid
  DATE: DAY OF WEEK, MONTH DAY, YEAR (YYYY-MM-DD)
  </system_info>


  # Tools

  You have access to tools that help you interact with the user's R session and workspace. Use these tools when they are helpful and appropriate to complete the user's request. These tools are available to augment your ability to help the user, but you are smart and capable and can answer many things on your own. It is okay to answer the user without relying on these tools.

  # Project Context

  User level context

  ---

  Project level context only

btw documentation built on Nov. 5, 2025, 7:45 p.m.