Modify docs using Git or the Github UI (All docs live in the
docs/ folder in the Airbyte repository)
If you're adding new files, update
If you're moving existing pages, add redirects in the
.gitbook.yaml file in the Airbyte repository root directory
Create a Pull Request
Directly edit the docs you want to edit in the Github UI
Create a Pull Request
Fork the repository.
Modify the documentation.
Create a pull request
Connectors typically have the following documentation elements:
Github Issues & Pull Requests
Source code comments
Below are some best practices related to each of these.
Every module should have a README containing:
A brief description of the module
development pre-requisites (like which language or binaries are required for development)
how to install dependencies
how to build and run the code locally & via Docker
any other information needed for local iteration
Core changelogs should be updated in the
Each connector should have a CHANGELOG.md section in its public facing docs in the
docs/integrations/<sources OR destinations>/<name> at the bottom of the page. Inside, each new connector version should have a section whose title is the connector's version number. The body of this section should describe the changes added in the new version. For example:
| Version | Date | Pull Request | Subject || :------ | :-------- | :----- | :------ || 0.2.0 | 20XX-05-XX | [PR2#](https://github.com/airbytehq/airbyte/pull/PR2#) | Fixed bug with schema generation <br><br> Added a better description for the `password` input parameter || 0.1.0 | 20XX-04-XX | [PR#](https://github.com/airbytehq/airbyte/pull/PR#) | Added incremental sync |
It's hard to pin down exactly what to do around source code comments, but there are two (very subjective) and rough guidelines:
If something is not obvious, write it down. Examples include:
non-trivial class definitions should have docstrings
magic variables should have comments explaining why those values are used (e.g: if using a page size of 10 in a connector, describe why if possible. If there is no reason, that's also fine, just mention in a comment).
Complicated subroutines/logic which cannot be refactored should have comments explaining what they are doing and why
If something is obvious, don't write it down since it's probably more likely to go out of date. For example, a comment like
x = 42; // sets x to 42 is not adding any new information and is therefore better omitted.
Describe outputs, not implementation: An issue or PR title should describe the desired end result, not the implementation. The exception is child issues/subissues of an epic. Be specific about the domain. Airbyte operates a monorepo, so being specific about what is being changed in the PR or issue title is important.
Some examples: subpar issue title:
Remove airbyteCdk.dependsOn("unrelatedPackage"). This describes a solution not a problem.
good issue title:
Building the Airbyte Python CDK should not build unrelated packages. Describes desired end state and the intent is understandable without reading the full issue.
subpar PR title:
Update tests. Which tests? What was the update?
good PR title:
Source MySQL: update acceptance tests to connect to SSL-enabled database. Specific about the domain and change that was made.
PR title conventions When creating a PR, follow the naming conventions depending on the change being made:
Notable updates to Airbyte Core: "🎉"
🎉 enable configuring un-nesting in normalization
New connectors: “🎉 New source or destination: ” e.g:
🎉 New Source: Okta
New connector features: “🎉 : E.g:
🎉 Destination Redshift: write JSONs as SUPER type instead of VARCHAR
🎉 Source MySQL: enable logical replication
Bugfixes should start with the 🐛 emoji
🐛 Source Facebook Marketing: fix incorrect parsing of lookback window
Documentation improvements should start with any of the book/paper emojis: 📚 📝 etc…
Any refactors, cleanups, etc.. that are not visible improvements to the user should not have emojis
The emojis help us identify which commits should be included in the product release notes.
Context: Provide enough information (or a link to enough information) in the description so team members with no context can understand what the issue or PR is trying to accomplish. This usually means you should include two things:
Some background information motivating the problem
A description of the problem itself
Good places to start reading and file changes that can be skipped
Create an OpenAPI to JSON schema generator. Unclear what the value or problem being solved here is.
When creating or updating connectors, we spend a lot of time manually transcribing JSON Schema files based on OpenAPI docs. This is ncessary because OpenAPI and JSON schema are very similar but not perfectly compatible. This process is automatable. Therefore we should create a program which converts from OpenAPI to JSONSchema format.