Skip to content

Latest commit

 

History

History
72 lines (49 loc) · 4.56 KB

commands.md

File metadata and controls

72 lines (49 loc) · 4.56 KB
Raw

Command Reference for Flame AI

Flame AI is a CLI built using Gluegun, and has a number of commands for interacting with OpenAI and doing codemods and codegen. It includes an interactive mode and several other specialized commands.

Upgrade: React Native

Flame AI has a command for upgrading React Native apps. This is the most mature command in Flame AI and is used in production by Infinite Red.

Flame AI React Native upgrade

Run this in the root of your React Native app (where the package.json and app.json live):

# one-shot upgrade
npx flame upgrade react-native

# recommended interactive mode (much more interesting):
npx flame upgrade react-native --interactive

Here's the command with all available options:

npx flame upgrade react-native --interactive --from=0.72.2 --to=auto --list --only=somefile.mm

Options:

  • --interactive: Run in interactive mode. This is the most useful mode.
  • --from=[auto|<version>]: The version to upgrade from. Defaults to the version specified in package.json.
  • --to=[auto|<version>]: The version to upgrade to. Defaults to the latest version of React Native on NPM.
  • --list: List the files that will be upgraded and exit.
  • --only=filename: Only upgrade the specified file (or any filename that includes this string).
  • --cacheFile=./path/to/cache.json: Specify a cache file to use, which stores OpenAI requests and responses. This is mainly useful for demos and testing when you don't want to hit the OpenAI API every time. Not recommended for production use.
  • --debug: Print verbose debug information as the upgrade runs.

Hints:

  • Interactive mode is the most useful. It will ask you questions and guide you through the upgrade process.
  • If running in interactive mode, Flame will ask you for a general prompt that will be used for every file to be upgraded. This is a flexible tool; for example, you can mention that you already attempted the upgrade and some files may have been upgraded already, and instruct it to tell you if it finds any issues.
  • If a file has moved, we are not currently able to find it nor move it for you. You'll need to move it manually or upgrade moved files manually. In the future, we will add ways to find those files either manually or automatically.
  • If you've customized some files extensively, you may need to manually apply the changes. Over time, Flame AI will get better at handling these cases.

EXPERIMENTAL: Interactive Code Editing and Authoring Mode

The interactive code editing mode is one of Flame's most exciting features, but it's still very much a work in progress. It allows you to chat with the AI, load files, list files, and make modifications to code using plain English.

To start the interactive mode, run flame experimental interactive. The current directory will be where you will be working. You can also specify a directory to work in by running flame experimental interactive <directory>.

In the interactive mode, you can use the following special commands:

  • /exit: Exit the interactive mode.
  • /load <filename>: Load a file's contents. Useful for evaluating and changing code.
  • /log: Print the current chat log for debugging purposes.
  • /context: Print the current context.
  • /clear: Clear the chat log.
  • /clearlast: Clear the last message from the chat log.
  • /ls <path>: Load a list files in the specified path. Helps the AI understand what files are available.
  • /logcompress <number>: Compress the chat log to a specified character count to avoid token limits. It'll automatically compress the log when it gets too long, but you can also do it manually. e.g. logcompress 5000. (Note these are characters and not tokens. Tokens are roughly 3 characters each.)
  • /debug: Print a full list of interactions, mostly for debugging issues.

EXPERIMENTAL: Flame AI Convert

Flame AI Convert is a command that allows you to convert code from X to Y using a recipe. It's still very much a work in progress, but it's already useful for converting between the few experimental recipes we have, such as converting from React Native AsyncStorage to react-native-mmkv.

To use it, run flame experimental convert <recipe> <input> <output>. For example:

npx flame experimental convert AsyncStorage mmkv ./my-file.ts

Recipes are stored in src/recipes. They are written in plain English and often are a list of examples. To see what a recipe looks like, check out the AsyncStorage-to-mmkv recipe.