You will need to have WebSocket endpoint to connect to the chain node and a Squid Archive. For a registry of Squid Archives, check this community-owned Archive Registry
If you don't find a suitable Squid Archive in the registry, set up your own Squid archive. There are multiple examples in this repo
Once set up, we encourage to contribute to the Squid community and make a PR to the Archive Registry.
Most chains publish their type bundles as an npm package. One of the best places to check for the latest version is the polkadot-js/app repo. Note, however, that a types bundle is only needed for pre-Metadata v14 blocks, so for recently deployed chains it may be not needed.
Note that the type bundle format for typegen is slightly different from OverrideBundleDefinition
of polkadot.js
. The structure is as follows, all the fields are optional.
{
types: {}, // top-level type definitions, as `.types` option of `ApiPromise`
typesAlias: {}, // top-level type alieases, as `.typesAlias` option of `ApiPromise`
versions: [ // spec version specific overrides, same as `OverrideBundleDefinition.types` of `polkadot.js`
{
minmax: [0, 1010] // spec range
types: {}, // type overrides for the spec range
typesAlias: {}, // type alias overrides for the spec range
}
]
}
The schema file defines the shape of the final GraphQL API and has very few limitations. Designing the schema file is very similar to the design of the database schema. As a rule of thumb, the schema should represent high level domain specific entities and relations between them, to make data fetching and filtering easy for the API consumers.
Typically, the API is consumed by the frontend and mobile apps, so it's a good idea to design the schema even before you go on with implementing the processor mappings. In fact, the processor is completely indpendent from the GraphQL server serving the API, so you can experiment how the API looks like.
TL;DR: If you're ok dropping the database, simply update schema.graphql
and run:
bash scripts/reset-schema.sh
OBS! The database will be wiped out, so if it's not an option, read below.
Here's a step-by-step instruction. First, generated the model files:
npx sqd codegen
npm run build
Now you have to options: either create a migration for an incremental schema update or recreate the whole schema from scratch.
During the development process, recreating the schema is often more convenient. However, if you already have a running API in production and don't want to resync it, having an incremental update is preferrable (but requires data backfilling).
Run
bash scripts/reset-db.sh
Generate a migration for the incremental changes and apply it
npx sqd db create-migration AddMyAwesomeNewField
npx sqd db migrate
You can find the newly generated and applied migration in db/migrations
.
Once the migrations are applied, simply run
npx squid-graphql-server
Observe the port (4350 by default) and navigate to localhost:4350/graphql
to explore your API. However, you need to run the processor to start populating the database.
First, make sure you have compiled your project with
npm run build
Then simply run
node -r dotenv/config lib/processor.js
Note that -r dotenv/config
ensures that the database settings are picked up from .env
. If you the environment variables them elsewhere, skip it.
Login to the Subsquid Hosted Service with your github handle to obtain a deployment key. Then create a Squid (that is, your deployment) and follow the instructions.
This part depends on the runtime business-logic of the chain. The primary and the most reliable source of information is thus the Rust sources for the pallets used by the chain.
For a quick lookup of the documentation and the data format it is often useful to check Runtime
section of Subscan, e.g. for Statemine. One can see the deployed pallets and drill down to events and extrinsics from there. One can also choose the spec version on the drop down.
Runtime upgrades may change the event data and even the event logic altogether, but Squid gets you covered with a first-class support for runtime upgrades.
Subsquid SDK comes with a tool called metadata explorer which makes it easy to keep track of all runtime upgrades happen so far.
The basic usage of the explorer is as follows (check README for details):
npx squid-substrate-metadata-explorer \
--chain <chain endpoint here> \
--archive <archive endpoint here> \
--out metadataVersions.json
Once the exploration is done, you should define all events and calls of interest in typegen.json
, then adjust the bundle and metadata history references and run:
npx squid-substrate-typegen typegen.json
A type-safe definition for each and every version of the event will be generated. Most of the times, one should be able to infer a normalized interface together with some glue code to make it fit the runtime specific versions. For example, for Kusama balances.Transfer
event, squid-substrate-typegen
generated three slightly different versions that can be reconciled as follows:
/**
* Normalized `balances.Transfer` event data
*/
interface TransferEvent {
from: Uint8Array
to: Uint8Array
amount: bigint
}
function getTransferEvent(ctx: EventHandlerContext): TransferEvent {
// instanciate type-safe facade around event data
let event = new BalancesTransferEvent(ctx)
// initial version, with runtime spec 1020
if (event.isV1020) {
let [from, to, amount, fee] = event.asV1020
return {from, to, amount}
// first upgrade at runtime spec version 1050
} else if (event.isV1050) {
let [from, to, amount] = event.asV1050
return {from, to, amount}
} else { // current version
// This cast will assert,
// that the type of a given event matches
// the type of generated facade.
return event.asLatest
}
}