Class Settings

java.lang.Object
com.example.adkprtriaging.Settings
public final class Settings extends Object
Configuration read from environment variables. Mirrors settings.py in the Python ADK PR triaging agent, and matches the lazy-accessor style of the sibling ADK Issue Triaging Agent sample.

Values are exposed as accessor methods (read lazily on each call) rather than static final fields. This keeps the class loadable in unit tests and adk web agent loaders without a GITHUB_TOKEN present — only githubToken() throws when the token is actually required (i.e. right before a network call).

Required variables:

  • GITHUB_TOKEN — GitHub Personal Access Token with pull_requests:write permission. Required for both interactive and workflow modes.
  • GOOGLE_API_KEY — Gemini API key. Required for both modes (or set up Vertex AI credentials and GOOGLE_GENAI_USE_VERTEXAI=TRUE).

Optional variables:

  • OWNER — defaults to google.
  • REPO — defaults to adk-java.
  • MODEL — Gemini model used for triaging. Defaults to gemini-pro-latest; a Pro model favors classification quality over latency, which suits this low-volume, accuracy-sensitive task. Overridable without a code change.
  • INTERACTIVE — 1/true for interactive mode (asks for confirmation before labeling/commenting), 0/false for unattended workflow mode. Defaults to interactive when unset.
  • DRY_RUN — 1/true to log intended label/comment changes without calling the GitHub mutation endpoints. Lets you verify the full pipeline (incl. Gemini) without modifying any real pull request. Defaults to off.
  • PULL_REQUEST_NUMBER — the pull request to triage in workflow mode. Populated by the GitHub Actions workflow from the triggering PR.
  • EVENT_NAME — the GitHub event that triggered the workflow ( pull_request_target, workflow_dispatch, ...). Logged for diagnostics.
  • CONTRIBUTING_MD_PATH — path to the repository's CONTRIBUTING.md (default CONTRIBUTING.md, resolved against the working directory, which is the repo root in the workflow). Its content is embedded in the agent instruction so the guideline check stays in sync with the repo. Missing file is tolerated (empty content).

  • Method Details

    • githubToken

      public static String githubToken()
      Returns the GitHub token, throwing a clear error if it is not configured.

    • hasGithubToken

      public static boolean hasGithubToken()
      Returns true if a GITHUB_TOKEN is configured, without throwing.

    • owner

      public static String owner()

    • repo

      public static String repo()

    • model

      public static String model()
      Returns the Gemini model used for triaging. Defaults to gemini-pro-latest (a Pro model favors classification quality over latency for this low-volume, accuracy-sensitive task) and is overridable via the MODEL environment variable, so it can be changed without editing source.

    • eventName

      public static @Nullable String eventName()

    • pullRequestNumber

      public static @Nullable String pullRequestNumber()

    • isInteractive

      public static boolean isInteractive()

    • isDryRun

      public static boolean isDryRun()

    • contributingGuidelines

      public static String contributingGuidelines()
      Reads the repository's CONTRIBUTING.md (path overridable via CONTRIBUTING_MD_PATH) so the agent can check pull requests against the real contribution guidelines, mirroring the Python agent's read_file(CONTRIBUTING.md). Returns an empty string if the file cannot be read (e.g. in unit tests or when the working directory is not the repo root), so callers never need to handle an exception.

    • parseNumberString

      public static int parseNumberString(@Nullable String value, int defaultValue)
      Parses a number from a string, falling back to defaultValue on null/blank/invalid input. Mirrors parse_number_string in the Python utils.