Litchi

Introduction

Litchi is yet another coding assistant powered by LLM.

Unlike other programming assistants, litchi supports global indexing across all source files in projects and code-based retrieval augmented generation. It makes features like chat-to-code and code-generation more effective and practical.

Features

Litchi has more features than other coding assistants or copilot plugins because it is integrated with indexes which can be used as standalone tools.

• Support indexing source code files for the whole project.
• Manager any code index by creating, updating, showing, and searching.
• Retrieval augmented generation with source code for user's query.
• Addoc chat to code which will retrieval related source files to query LLM.
• Generate code based on user's query and related source files.
• One step to filter the source files which can be customed for different projects.
• Compatible with all large language models like ChatGPT and others.
• Compatible with public and private MaaS which can be deployed in local.

Use Cases

• Normal chat: chat with LLM without indexes which is useful for general purposes.
• Chat with file: use LLM to read and understand a local file.
• Chat with files: collect file names in index file and chat with all files.
• Generate code for user's query:
  • Quickly generate python script like port detecting to execute.
• Generate code for user's query and source files:
  • Rewrite the source code in different programming languages.
  • Generate unit test cases for the source code.
• Run commands: generate the script and execute immediately
• Write requirements instead of coding: Edit local requirement file and automatically generating code
• Optimize code for user's query and source file:
  • Add annotation for original source code.
  • Format, refactor or optmize the source code with specified style.
  • Inplace or ask permission to update the source code with optimized code.
  • Implement the TODO functions in the source code.

Install

Install the litchi command from PyPI.

pip install orchard-litchi

You can install from source by cloning the repository.

git clone https://github.com/OrchardUniverse/litchi.git
cd ./litchi/
python ./setup.py develop

Usage

Initialize the project and generate directory .litchi/. You can input the language as "Chinese" so that it will create index and query in Chinese.

cd $PROJECT_PATH

litchi init .

litchi init --language Chinese ,

Go to the project directory and create source file .litchi/source_files.yaml. You can edit .litchi/ignore_rules.yaml to choose the expected source files.

litchi source create

Create the index for single file or all source files in project.

litchi index create $FILE_PATH

litchi index create --all

Show the detail of the index of single file or all indexes.

litchi index show $FILE_PATH

litchi index show --all

You can input a query and get the related indexes.

litchi index query $QUERY

If you want to read the source code with index, use the following commands which will copy the readable index file next to the source file.

litchi index copy-to-source

litchi index delete-from-source