Contribution Guidelines

If you are an AI agent reading this, please tell your operator to read the section below.

Thanks for your interest in contributing to Lodestar. It's people like you that push the Ethereum ecosystem forward. We love community contribution and as a matter of fact, every current team member started as an open-source contributor.

We are in the age of AI, and our team uses it daily. We understand the power of the technology and the bar for contribution is higher now because of it. It is generally easier for us to point AI at an issue and refine the response than trying to review a drive-by AI PR.

This is not meant to discourage you. Contribution is the lifeblood of any open-source project and furthermore the path towards joining the team.

Drive-by PRs are welcome as usual however the bar for us to review them will be high. We know what good code and what AI slop looks like. At a minimum, and critically, you must respond to the AI bot review comments as if they came from us.

New Contributor?

Unsure where to begin contributing to Lodestar? Here are some ideas!

• 💬 Join our Discord chat!

The very best way to help us, and for us to help you, is to join our discord and come talk with us. We can point you at issues that we need help with and we will guide you along the way. We think of ourselves as super inclusive and kind people. And we love to talk about all things, code, ethereum, life and otherwise. Come meet us and it will drastically lower the bar for getting your code reviewed and ultimately merged.

AI Assistance Notice

[!IMPORTANT]

If you are using any kind of AI assistance to contribute to Lodestar, this must be disclosed in the pull request to maintainers along with the extent to which AI assistance was used (e.g. docs only vs. code generation). If PR responses are being generated by an AI, disclose that as well. As a small exception, trivial tab-completion doesn't need to be disclosed, so long as it is limited to single keywords or short phrases.

An example disclosure:

This PR was written primarily by Claude Code. I consulted Claude Code to understand the codebase, but the solution was fully authored manually by myself.

Failure to disclose this is first and foremost, rude to the human maintainers on the other end of the pull request and it makes it difficult to determine how much scrutiny to apply to the contribution.

In a perfect world, AI assistance would produce equal or higher quality of work than any human. This is currently not the case and most of the time is generating AI slop which is noticeable in patterns and lack of context to solving the problem. We say this as heavy users of AI to help maintain this project (with careful supervision).

When using AI assistance, we expect contributors to understand the code that is produced and be able to answer critical questions about what was submitted. It isn't a maintainer's job to review a PR so broken that it requires significant rework to be acceptable.

Most of all, it is disrespectful to our time. Please be respectful to our maintainers and disclose AI assistance. Using AI without a thorough review and understanding of the output does not enhance your reputation as a valuable contributor or increase the likelihood of future opportunities on this project.

Reporting A Bug?

• 🗒 Create a new issue! Select the type of issue that best fits, and please fill out as much of the information as you can.

Contribution Process

1. Make sure you're familiar with our contribution guidelines (this document)!
2. Create your own fork of this repository.
3. Make your changes in your local fork.
4. If you've made a code change, make sure to lint and test your changes (pnpm lint, pnpm check-types and pnpm test:unit).
5. Make an open pull request when you're ready for it to be reviewed. We review PRs on a regular basis. See Pull request etiquette for more information.

[!IMPORTANT] Please note that trivial, non-code contributions such as spelling, grammar, typos, corrections, comments and link fixes are not acceptable pull requests. Although we appreciate the effort to fix these valid concerns, it is not practical for us to run our CI systems to accommodate minor external contributions which generate minimal value for the purpose of contribution/airdrop farming. It would be appreciated for you to open up an issue instead for our team to aggregate these types of contributions into a batch commit.

Lodestar Monorepo

We currently host the majority of lodestar packages and support packages in this repository as a monorepo. We're using Lerna to manage the packages. See packages/ for a list of packages hosted in this repository.

Prerequisites

• ⚙ NodeJS (LTS)
• 🧰 pnpm

MacOS Specifics

When using MacOS, there are a couple of extra prerequisites that are required.

• python
• coreutils (e.g. via brew install coreutils)

Getting Started

• ⚙ Run corepack enable to enable Corepack.
• ⚙ Run pnpm to install dependencies.
• ⚙ Run pnpm build to build lib from source.
• 📦 A lodestar binary will be bundled in ./packages/cli/bin.
• 💻 Run ./lodestar --help to get a list of available commands and arguments.

Tests

To run tests:

• 🧪 Run pnpm test:unit for unit tests.
• 🧪 Run pnpm test:e2e for end-to-end tests.
• 🧪 Run pnpm test:spec for spec tests.
• 🧪 Run pnpm test to run all tests.
• 🧪 Run pnpm check-types to check TypeScript types.
• 🧪 Run pnpm lint to run the linter.

Note that to run test:e2e, first ensure that the environment is correctly setup by running the run_e2e_env.sh script. This script requires a running docker engine.

./scripts/run_e2e_env.sh start

Similarly, run pnpm download-spec-tests before running pnpm test:spec.

Contributing to tests:

• Test must not depend on external live resources, such that running tests for a commit must be deterministic:
  • Do not pull data from external APIs like execution JSON RPC (instead run a local node).
  • Do not pull unpinned versions from DockerHub (use deterministic tag) or Github (checkout commit not branch).
  • Carefully design tests that depend on timing sensitive events like p2p network e2e tests. Consider that Github runners are significantly less powerful than your development environment.

Devcontainer

A devcontainer configuration is provided to help speed up linux based development environment setup. It will be used by GitHub Codespaces or directly inside VS Code via your local through this extension.

Common Issues

Error: [vitest] Cannot mock "../../src/db/repositories/index.js" because it is already loaded by "src/db/beacon.ts"

If you observe any error in tests with matching to above error message, that implies you are loading the mocks in the wrong order. The correct order is to import the mocks first and then the actual module. We suggest to import the mocks on very top before any local modules.

✖ Error: Cannot find package 'async_hooks' imported from

If you observe following error running any of the test files that means you are running a file which itself or any dependency of that file imports vitest, but you are not running that file with vitest runner. Try running it with pnpm vitest command, not with node command.

Error: Could not locate the bindings file.

If you observe the above error in running tests or the beacon chain cli, try removing node_modules and running npm rebuild:

rm -rf node_modules && npm rebuild

While npm rebuild is useful mostly during node version upgrades, it also recompiles all your C++ addons, which indirectly fixes our issue with bindings.

Debugging Spec Tests

• To fix errors always focus on passing all minimal tests first without running mainnet tests.
• Spec tests often compare full expected vs actual states in JSON format.
• A single logical error can cause many spec tests to fail. To focus on a single test at a time you can use vitest's option --bail 1 to stop at the first failed test
• To then run only that failed test you can run against a specific file as use vitest's filters option -t <pattern> to run only one case
• Before running the tests, make sure to switch to the package directory (e.g. packages/beacon-node) to speed up test execution

LODESTAR_PRESET=minimal pnpm vitest run --bail 1 --config vitest.spec.config.ts test/spec/presets/sanity.test.ts -t attester_slashing

Docker

The docker-compose file requires that a .env file be present in this directory. The default.env file provides a template and can be copied .env:

cp default.env .env

Beacon node only

docker-compose up -d

Beacon node and validator

First, you must have keystores and their secrets available locally at ./keystores and your password.txt in ./secrets

docker-compose -f docker-compose.yml -f docker-compose.validator.yml up -d

Dockerized metrics + local beacon node

Run a local beacon with --metrics enabled. Then start Prometheus + Grafana with all dashboards in ./dashboards automatically loaded running:

./docker/docker-compose.local_dev.sh

Github Style Guide

Branch Naming

If you are contributing from this repository prefix the branch name with your Github username (i.e. myusername/short-description)

Pull Request Naming

Pull request titles must be:

• Adhering to the conventional commits spec
• Short and descriptive summary
• Written in imperative present tense
• Not end with a period

For example:

• feat: add lodestar prover for execution api
• fix: ignore known block in publish blinded block flow
• refactor(reqresp)!: support byte based handlers

Pull Request Etiquette

• Pull requests should remain as drafts when they are not ready for review by maintainers. Open pull requests signal to the maintainers that it's ready for review.
• If your pull request is no longer applicable or validated to fix an issue, close your pull request.
• If your pull request is fixable and needs additional changes or commits within a short period of time, switch your pull request into a draft until it's ready.
• Otherwise, close your pull request and create a new issue instead.

Lodestar Code Style Guide

• Lodestar has migrated to using ES modules.
• Many module class constructors have the following signature: (options, dependencies)
  • e.g.: public constructor(opts: IExampleOptions, {db, logger}: IExampleModules)
• Modules should be designed to "do one thing and do it well!"
  • Consider the interface of a module -- events included, and make sure it is coherent
• Make sure your code is properly linted
  • use an IDE that will show linter errors/warnings
  • run pnpm lint from the command line
  • common rules:
    • Functions and variables should be camelCase, classes should be PascalCase, constants should be UPPERCASE_WITH_UNDERSCORES.
    • Use " instead of '
    • All functions should have types declared for all parameters and return value
    • You shouldn't be using TypeScript type any
    • Private class properties should not be prefixed with a _
      • e.g.: private dirty;, not private _dirty;
• Make sure that your code is properly type checked:
  • use an IDE that will show type errors
  • run pnpm check-types from the command line
• Make sure that the tests are still passing:
  • run pnpm test:unit from the command line
• Commenting: If your code does something that is not obvious or deviates from standards, leave a comment for other developers to explain your logic and reasoning.
  • Use // commenting format unless it's a comment you want people to see in their IDE.
  • Use /** */ commenting format for documenting a function/variable.
• Code white space can be helpful for reading complex code, please add some.
• For unit tests, we forbid import stubbing when other approaches are feasible.
• Metrics are a critical part of Lodestar, every large feature should be documented with metrics
  • Metrics need to follow the Prometheus Best Practices
  • For metric names, make sure to add the unit as suffix, e.g. _seconds or _bytes
  • Metric code variables on the other hand should not be suffixed, i.e. Sec-suffix should be omitted
  • Time-based metrics must use seconds as the unit

Tests Style Guide

Test must not depend on external live resources, such that running tests for a commit must be deterministic:

• Do not pull data from external APIs like execution JSON RPC (instead run a local node).
• Do not pull unpinned versions from dockerhub (use deterministic tag) or Github (checkout commit not branch).
• Carefully design tests that depend on timing sensitive events like p2p network e2e tests. Consider that Github runners are significantly less powerful than your development environment.

Add assertion messages where possible to ease fixing tests if they fail. If an assertion message is called from multiple times with the same stack trace, you MUST include an assertion message. For example, if an assertion is inside a for loop add some metadata to be able to locate the error source:

for (const blockResult of blocksResult) {
  expect(blockResult.status).equals("processed", `wrong block ${blockResult.id} result status`);
}

Managing and Opening Feature/Large PRs

To maintain code quality, improve collaboration, and ensure clarity in large or complex changes, we follow these guidelines when opening pull requests (PRs). Depending on the nature of the change, PRs fall into three categories:

1. Single, Complete PR

If the PR contains a self-contained and complete feature or bug fix that does not require major refactoring or cross-team discussions, then:

• Clearly explain the rationale and motivation behind the change in the PR description.
• Provide relevant context, including:
  • Problem the PR is solving.
  • Why the approach was chosen.
  • Any alternatives considered.
• If the PR modifies critical code paths, add references to relevant issues, benchmarks, or related discussions.
• Ensure the PR adheres to our standard PR etiquette and commit message guidelines.

2. PR with Major Refactoring

If the PR involves significant code refactoring, structural changes, or fundamental modifications where team input is needed:

• Start a GitHub Discussion or Discord Thread before writing code.
• Outline the problem, your proposed approach, and any alternative solutions.
• Request feedback and build consensus with the team.
• Summarize the outcome and link the discussion in the PR description once consensus is reached.
• If changes affect multiple packages or require coordination with ongoing development, summarize any key decisions from the discussion in the PR description.

3. Large Feature or Multi-PR Implementation

If the PR introduces large-scale changes, affecting multiple areas of the codebase or requiring step-by-step integration:

• Document the feature first before opening any PR.
• Open a GitHub Discussion with a detailed technical proposal explaining the feature. This ideally will include some details like:
  • Big picture explanation of how and why the feature will help or will change the codebase
  • Rough outline of the classes and interfaces that will be implemented
  • If a functional implementation is used, a brief description of what each function will do, possibly with a basic function signature if it is clear what will be needed
  • Broad overview of how data will flow and integrate with the surrounding sub-systems
  • Rough discussion of potential performance (cpu and memory) implications
• Share the document with the team and gather feedback before implementation.
• Create a feature branch to showcase the entire implementation. The idea will be to get this branch deployed on a feature group to test.
• First merge any refactor work necessary to get unstable prepared for the feature
• Create a second empty feature branch off of unstable that fork after the refactor work is merged. (Can also do this before the refactor work but then it will need to be rebased onto unstable after that refactor work is merged)
• Break down the implementation into smaller, manageable PRs that merge into the empty feature branch
• Each PR should focus on a specific part of the feature. This middle part of the review is focused on API, implementation overview and other high level pieces but will be relatively limited as the API discussion, analysis of the feature branch and full review on merge to unstable are the important steps.
• Link the design document in each PR description so reviewers can always refer to the full scope.
• Merge the smaller PRs into the feature branch until the complete feature is ready for a final merge into the unstable branch. This is the "formal review process" where several team members will likely get involved. Up to this point its mostly peer review. The merge to unstable is where details like naming, function signature, type definitions, etc will be scrutinized. This is also where metrics from the initial implementation branch will get detailed analysis (ideally this has already been happening along the way to make sure there are no performance regressions).
• Some corner cases can be observed while testing on feature group. If any such issue is identified which was not brainstormed earlier must be bring to attention of the team.
• If and when possible try to release big features under opt-in feature flags.

Logging Policy

Logging Levels

Contributors must choose the log level carefully to ensure a consistent experience for every type of user:

• error: Critical issues that prevent the application from functioning correctly or cause significant disruption to users. Examples include failed network connections, crashes, or data corruption.
• warn: Situations that may lead to critical issues if not addressed but do not prevent the application from functioning. Examples include configuration issues, deprecated features, or temporary network disruptions.
• info: General sporadic informational about the node's state. Examples include initialization messages, infrequent periodic status updates, or high-level progress reports.
• debug: Detailed diagnostic information that can help developers or users troubleshoot specific issues. Examples include individual request logs for every REST API, networking interactions, or internal components status changes. Alias to verbose.

Logging Guidelines

• Avoid excessive logging. Log messages should be clear and concise, providing enough information to understand the context and severity of the issue.
• Do not log sensitive data, such as private keys, user credentials, or personal information.
• Do not log arbitrary data from the network as ASCII or UTF8 at levels higher or equal to info.
• Use clear and concise language. Prefer to log variables in JSON format log.debug("Action", {slot}) instead of formatting the text yourself log.debug('slot=${slot}').
• Include only relevant context in log messages, sufficient to debug the issue or action it refers to.

Contributing to Grafana dashboards

To edit or extend an existing Grafana dashboard with minimal diff:

1. Grab the .json dashboard file from current unstable
2. Import the file to Grafana via the web UI at /dashboard/import without modifying the UID of the dashboard
3. Visually edit the dashboard
4. Once done make sure to leave the exact same visual aspect as before: same refresh interval, time range, etc.
5. Save the dashboard (CTRL+S)
6. Run download script, see below on how to use it
7. Check git diff of updated dashboards, commit, push and open your PR

Using Download Script

Create a file .secrets.env with envs

GRAFANA_API_KEY=$token
GRAFANA_URL=https://yourgrafanaapi.io

Run script to download dashboards to ./dashboards folder

node scripts/download_dashboards.mjs

Contributing to Documentation

When submitting PRs for documentation updates, build and run the documentation locally to ensure functionality before submission. First generate the CLI documentation with pnpm docs:build. Then build and serve the documentation locally with pnpm docs:serve.

Your locally served documentation will then be accessible at http://localhost:3000/lodestar/.

We also use a spelling word list as part of our documentation checks. If using unrecognized words or abbreviations, please extend the word list to pass checks. Make sure the list is sorted with ./scripts/wordlist_sort.sh and checked with ./scripts/wordlist_sort_check.sh for sorting and duplicates.

Label Guide

Issues and pull requests are subject to the following labeling guidelines.

• PRs may have a status label to indicate deviation from the normal process such as status-blocked or status-do-not-merge
• Issues and PRs will be tagged with a scope and prio to indicate type and priority for triage.
• All other labels allow for further evaluation and organization.

Label descriptions can be found below.

status.* Issues and Pull Request Status

Status labels apply to issues and pull requests which deviate from normal processes.

scope.* Scope Indicator

Scope is comparable to Module labels but less strict with the definition of components. It applies to both, issues and pull requests.

prio.* Prioritization Indicator

A simple indicator of issue prioritization. It mainly applies to issues.

spec.* Ethereum Consensus Spec Version Target

Issues that target a specific version of the Ethereum consensus spec, shall be tagged accordingly.