Json-rpc to graphql migration

[11felix]

• Migrated SDK read paths from JSON-RPC to GraphQL while preserving existing public behavior and return contracts.
• Refactored blockchain/query parsing to handle GraphQL-native object/dynamic field/event shapes with stricter typing.
• Updated client architecture to keep transaction building reliable while removing unnecessary external SuiClient coupling.
• Added migration snapshot + semantic diff tooling to validate parity across all exposed reads and transaction builders.
• Documented migration + API updates and aligned quality gates (lint/type/test/pre-commit) for merge-ready stability.

[jangid]

Review: JSON-RPC to GraphQL Migration

Overall: Well-executed migration. Architecture is sound — all reads move to SuiGraphQLClient, public API simplifies, and the snapshot-diff tooling is a thoughtful approach to verifying parity. CI is green.

Issues to Address

1. Breaking change: getCoinObject for SUI with amount (behavior change)

In blockchain.ts, the new getCoinObject for SUI when amount is provided:

if (this.isCoinTypeSui(coinType)) {
  if (amount) {
    return tx.splitCoins(tx.gas, [amount]);
  }
  return tx.gas;
}

The old code in client.ts always returned tx.gas for SUI regardless of amount. The new code splits when amount is provided. This changes transaction shape for SUI supply/borrow/repay flows — the snapshot diff tool might mask this if it's categorized as noise. Verify this is intentional and tested end-to-end.

2. multiGetObjects query — ObjectKey type mismatch

In blockchain.ts:

variables: { objectIds: batch.map((id) => ({ address: id })) },

The Sui GraphQL schema's multiGetObjects takes [ObjectKey\!]\! which requires { address, version }, not just { address }. If this works at all, it's only because the server is lenient. Confirm this actually functions — or use individual getObject calls / the objectConnection query instead.

3. interestRates type assertion is unsafe

In parser.ts:

interestRates: config.interest_rates as unknown as number[],

interest_rates might also come as base64 from GraphQL (same as interest_rate_kinks). The type in queryTypes.ts says string[] | number[] but doesn't handle the base64 case. Should this also use decodeU8Vec?

4. Receipt type is a breaking change for downstream consumers

The old Receipt had receipt.content.type and receipt.content.fields.*. The new shape is { objectId, fields: ReceiptGql }. Any consumer accessing receipt.content will break. The README migration table doesn't mention this. Consider documenting it or keeping a compatibility shim.

5. Missing getEstimatedGasBudget signature change in exports

The old signature was getEstimatedGasBudget(suiClient, tx, address), the new is getEstimatedGasBudget(tx, address). Listed in README — good. But verify this is reflected in the published type declarations (index.ts exports).

Minor / Nits

• testGetAllMarkets.ts: const clientDuration = 0; is dead code left from removing SuiClient creation timing. Remove it and references in benchmark output.
• pythTest.ts: Changed from // run() to run().catch(...) — this makes the script auto-execute on import. Intentional?
• oracle-price-validation.test.ts: Still imports SuiClient and constructs one manually for raw dynamic field probing — the test still has a JSON-RPC dependency. Worth noting.
• tsconfig.json: New root tsconfig.json extends tsconfig.esm.json — ensure no conflict with tsconfig.cjs.json build path.

Questions

1. Was the snapshot diff tool actually run, and what was the result? (0 signal diffs?)
2. Is this intended as a semver major bump (2.0.0)? README references "2.x" but no package.json version change in the diff.
3. The priceIdentifier.type field removal — are there any known downstream consumers in alphafi-api, alphafi-crons, or alphalend-api?

[11felix]

1. It is intentional
2. confirmed and this works fine
3. confirmed that it comes as number[]
4. Updated in readme
5. Updated in readme



Questions - 


1. yes 0 signal diffs

2. Yes next version should be 2.0.0

3. None

[jangid]

Thanks for addressing the comments. Responses are satisfactory — intentional SUI split behavior, multiGetObjects confirmed working, interest_rates as number[] verified, README updated for Receipt and getEstimatedGasBudget changes. 0 signal diffs on snapshot tool is reassuring.

Minor nits (dead clientDuration, pythTest auto-execute, test JSON-RPC dep) can be cleaned up in a follow-up. Bump to 2.0.0 on merge.