Add Oracle (OracleSet + OracleDelete) transaction workloads

[manasip-prog]

Summary

• Adds full Antithesis workload coverage for Price Oracles (XLS-47) — OracleSet and OracleDelete transaction types with valid paths, fault injection, state tracking, and protocol-invariant assertions.
• No setup phase required — Oracle objects are created implicitly by OracleSet and the empty-state case is handled by early return in handlers.

New files

File	Purpose
transactions/oracle.py	OracleSet + OracleDelete handlers (valid + faulty paths)
parallel_driver_oracle_set_random.sh	Antithesis driver script for OracleSet
parallel_driver_oracle_delete_random.sh	Antithesis driver script for OracleDelete

Modified files

File	Change
models.py	Added Oracle dataclass (account, document_id, provider, asset_class)
params.py	Added 10 Oracle parameter generators (document_id, provider, asset_class, base/quote assets, asset_price, scale, price_data_count, last_update_time)
app.py	Added self.oracles = [] to Workload state
transactions/__init__.py	Added _on_oracle_set / _on_oracle_delete state updaters + registry entries
assertions.py	Added OracleSet / OracleDelete to _META_EXPECTATIONS table + OracleDocumentID to _OBJECT_ID_FIELDS
transactions/tickets.py	Added OracleSet to _TICKET_BUILDERS + OracleDelete to _TICKET_EXCLUDED
scripts/check-imports	Added oracle_set, oracle_delete, Oracle imports
scripts/check-endpoints	Added /oracle/set/random, /oracle/delete/random

Coverage

Valid paths

• OracleSet: 60% create new oracle (random doc_id, provider, asset_class, timestamp, 1–5 PriceData entries with unique base/quote pairs). 40% update existing oracle (reuses stored provider/asset_class, fresh timestamp + price data).
• OracleDelete: Picks a tracked Oracle from w.oracles and deletes it. Returns silently if none exist.

Faulty paths — OracleSet (4 mutations)

Mutation	Expected result
non_owner_submission	tefBAD_AUTH
invalid_flags (0x80000000)	temINVALID_FLAG
invalid_update_time (epoch + 1)	tecINVALID_UPDATE_TIME
provider_mismatch	Provider conflict on existing oracle

Faulty paths — OracleDelete (3 mutations)

Mutation	Expected result
delete_nonexistent	tecNO_ENTRY
non_owner_submission	tefBAD_AUTH
invalid_flags (0x80000000)	temINVALID_FLAG

Assertions (Always-invariants)

• _META_EXPECTATIONS["OracleSet"] = ("Created", "Modified", "Oracle") — on tesSUCCESS, metadata must contain an Oracle CreatedNode or ModifiedNode
• _META_EXPECTATIONS["OracleDelete"] = ("Deleted", "Oracle") — on tesSUCCESS, metadata must contain an Oracle DeletedNode

State tracking

• _on_oracle_set: Idempotent — updates existing oracle (by account + document_id) or appends new Oracle to w.oracles
• _on_oracle_delete: Removes oracle matching account + document_id from w.oracles

Design decisions

1. PriceData unique pairs — _random_price_data_series() deduplicates base/quote pairs within a single series and guarantees at least one entry, matching the spec requirement.
2. invalid_update_time uses epoch + 1 — xrpl-py blocks values before the XRPL epoch (946684800), so we use a value just after the epoch (~year 2000) which passes client validation but triggers tecINVALID_UPDATE_TIME server-side (>300s from ledger close).
3. No setup phase — Oracles don't require pre-existing ledger objects, avoiding additions to the setup dependency chain.
4. Ticket coverage — OracleSet is stateless (can create new oracle with any doc_id) so it's in _TICKET_BUILDERS. OracleDelete requires a pre-existing oracle so it's in _TICKET_EXCLUDED.
5. Hex-encoded Provider/AssetClass — xrpl-py calls bytes.fromhex() on these Blob fields, so generators return uppercase hex strings (e.g. "chainlink_123" → "636861696E6C696E6B5F313233").

Test plan

• scripts/check-imports passes
• scripts/check-endpoints passes (65 endpoints, 29 expected)
• ruff check passes
• ruff format --check passes
• Antithesis run (#1217) — all assertions passed:
  • workload::success : OracleSet ✅ reached
  • workload::success : OracleDelete ✅ reached
  • workload::failure : OracleSet ✅ reached (faulty paths exercised)
  • workload::seen : OracleSet ✅ reached
  • workload::endpoint_exception ✅ resolved (hex-encoding fix)
  • No existing assertions regressed

Known xrpl-py limitations

The following scenarios are blocked by xrpl-py client-side validation before reaching rippled:

Scenario	Ledger error	Why blocked
Empty PriceDataSeries	temARRAY_EMPTY	xrpl-py requires len(price_data_series) > 0
>10 PriceData entries	temARRAY_TOO_LARGE / tecARRAY_TOO_LARGE	xrpl-py enforces len <= 10
LastUpdateTime before XRPL epoch	temMALFORMED	xrpl-py enforces >= 946684800

References

• XLS-47 Price Oracle Specification
• OracleSet docs · OracleDelete docs

---

Pull Request opened by Augment Code with guidance from the PR author