LangChain4j GitHub Bot

A Quarkus-powered GitHub App to simplify issues and pull requests management in the LangChain4j project.

Introduction

This GitHub App is based on the Quarkus GitHub App framework.

It can be run as a native executable.

Current Actions

Triage issues

Based on the .github/langchain4j-github-bot.yml file, this rule affects labels to issues and also pings the appropriate people.

Syntax of the .github/langchain4j-github-bot.yml file is as follows:

triage:
  rules:
    - labels: [Azure]
      title: "azure"
      notify: [developer1,developer2]
      files:
        - langchain4j-azure-ai-search
        - langchain4j-azure-cosmos-mongo-vcore
        - langchain4j-azure-cosmos-nosql
        - langchain4j-azure-open-ai
    - labels: [OpenAI]
      title: "openai"
      notify: [developer3]
      files:
        - langchain4j-open-ai

For issues, each rule can be triggered by:

title - if the title matches this regular expression (case insensitively), trigger the rule
body - if the body (i.e. description) matches this regular expression (case insensitively), trigger the rule
titleBody - if either the title or the body (i.e. description) match this regular expression (case insensitively), trigger the rule
expression - allows to write a Jakarta EL expression testing title, body or titleBody. Be careful when writing expressions, better ping @gsmet in the pull request when creating/updating an expression.

Tip

When writing expressions, you can use the matches(String pattern, String string) function that behaves as follows:

public static boolean matches(String pattern, String string) {
    return Pattern.compile(".*" + pattern + ".*", Pattern.DOTALL | Pattern.CASE_INSENSITIVE).matcher(string)
            .matches();
}

A rule using an expression based on matches() would look like:

    - labels: [OpenAI]
      expression: |
              matches("openai", title)
              && matches("open.ai", title)
      notify: [developer1,developer2]

If the rule is triggered, the following actions will be executed:

notify - will create a comment pinging the users listed in the array
labels - will add the labels to the issue

Triage pull requests

The pull requests triage action uses the same configuration file as the issues triage action.

There are a few differences though as it doesn’t behave in the exact same way.

For pull requests, each rule can be triggered by:

files - if any file in the commits of the pull requests match, trigger the rule. This is not a regexp (it uses startsWith) but glob type expression are supported too *-azure-*/** (make sure you enclose globs into ' in YAML).

If no rule is triggered based on files, or if rules are triggered but they all specify allowSecondPass: true, a second pass will be executed; in that second pass, rules can be triggered by:

title - if the title matches this regular expression (case insensitively), trigger the rule
body - if the body (i.e. description) matches this regular expression (case insensitively), trigger the rule
titleBody - if either the title or the body (i.e. description) match this regular expression (case insensitively), trigger the rule
expression - allows to write a Jakarta EL expression testing title, body or titleBody. Be careful when writing expressions, better ping @gsmet in the pull request when creating/updating an expression.

If the rule is triggered, the following action will be executed:

labels - will add the labels to the issue
notify - will create a comment pinging the users listed in the array only if notifyInPullRequest is true

notifyInPullRequest should be used as follows:

triage:
  rules:
    - labels: [Azure]
      title: "azure"
      notify: [developer1,developer2]
      notifyInPullRequest: true
      files:
        - langchain4j-azure-ai-search
        - langchain4j-azure-cosmos-mongo-vcore
        - langchain4j-azure-cosmos-nosql
        - langchain4j-azure-open-ai

Triage discussions

The rules applied for issues and pull requests are also applied to discussions, as long as the category is monitored.

Monitoring a category is enabled with:

triage:
  discussions:
    monitoredCategories: [33575230]

The number is the numeric id as present in the JSON event payload.

Enforce color for specific labels

The bot enforces a specific color for any label created that starts with area/ so that all these labels are consistent.

Contributing

To participate to the development of this GitHub App, create a playground project in your own org and follow the steps outlined in the Quarkus GitHub App documentation.

GitHub permissions required:

Contents - Read only
Discussions - Read & Write
Issues - Read & Write
Pull Requests - Read & Write

Events to subscribe to:

Discussions
Issues
Label
Pull Request

By default, in dev mode, the Bot runs in dry-run so it’s logging its actions but do not perform them. You can override this behavior by adding _DEV_LANGCHAING4J_GITHUB_BOT_DRY_RUN=false to your .env file.

Deployment

Once logged in to the OpenShift cluster (using oc login…), just run:

$ ./deploy-to-openshift.sh

License

This project is licensed under the Apache License Version 2.0.

langchain4j / langchain4j-github-bot