This is a collection of shared data schema definitions across Sentry's services.
- If there is a definitive service that owns/produces the schema, it should be
used as folder name. Otherwise put the schema into
sentry/
. - Bot commits for auto-generated schema updates should reference the commit in the service repo that they were triggered by (e.g. reference commit of Relay master for schema change originating in Relay)
Right now the best way to use this repo is to include it as a submodule in your
project and set up Dependabot to periodically update
HEAD
, then use the following tools:
- For consuming JSON Schema as Python type hints,
jsonschema-typed
(
jsonschema-typed-v2
on PyPI) is recommended. It has been found by Snuba devs to produce decent type definitions. - For consuming JSON Schema as TypeScript type definitions, json-schema-to-typescript is recommended.
- Using a JSON Schema validation library in your SDK's test suite can help find inconsistencies between what you send and what the server accepts.
- Setting up something based on quicktype might work. Quicktype produces output for a very large variety of languages, but its output for TypeScript in particular has been found to be weakly typed. We use a fork of quicktype to generate schema documentation.
Contains a draft-07 JSON Schema of the Event payload as accepted, validated and
understood by Relay. The schema is automatically updated when master
in
getsentry/relay
is committed to.
Following caveats apply:
- Security reports and transaction events are not part of that schema.
- The top-level keys
metadata
andtitle
are missing because Relay does not produce them (Sentry does). - The key
project_id
andreceived
is missing as Relay does not accept those from SDK. - The keys
culprit
,template
and other deprecated fields are missing such that no producers or consumers can either send or rely on them.
In short, only the canonical form of an error event as SDKs should send them is reflected by that schema.