- Remove TEST_USER_ID checks from balance-calculator.ts (billing bypass)
- Remove TEST_USER_ID checks from stripe-metering.ts (already has CI check)
- Remove TEST_USER_ID checks from agent-runs API endpoints
- Replace TEST_USER_ID with local ADMIN_RELABEL_USER_ID in admin route
- Remove TEST_USER_ID export from old-constants.ts
- Remove TEST_USER_ID from test fixtures (use inline strings instead)
- Update 18 test files to use inline test-user-id string

The hardcoded test user bypass was a hack that coupled test infrastructure
to production billing code. Tests now use inline strings and production
code relies on environment checks (CI=true) to skip billing in test envs.

These tests were testing the old TEST_USER_ID skip behavior that was
removed in the previous commit. Tests now go through normal flow.

Instead of adding a new SKIP_BILLING env var, evals now set CI=true at the
top of their entry points (main.ts, main-nightly.ts, main-single-eval.ts,
main-hard-tasks.ts). This ensures billing is always skipped when running
evals, without requiring users to remember to set an env var.

The shouldAttemptStripeMetering() function already checks CI=true, so this
approach is simpler and follows the existing pattern.

- Add grantCreditOperation debt settlement tests
- Add revokeGrantByOperationId scenario tests
- Add usage-service balance calculation tests with multiple grant types
- Add usage-service error handling tests
- Add credit-delegation normalizeRepositoryUrl and extractOwnerAndRepo tests
- Add credit-delegation URL edge case tests

73 billing tests now pass with full DI coverage.

- Add monthly reset flow tests (future date skip, user not found error)
- Add usage data flow tests (reset -> auto-topup -> balance calculation)
- Add debt settlement flow tests (single/multiple debts, debt exceeds grant)
- Add credit delegation flow tests (no repo, empty repo, malformed URLs)
- Add credit fallback flow tests (no repo fallback, org delegation attempt)
- Add complete billing cycle tests (reset -> grant -> consume -> balance)
- Add balance calculation tests (multi-grant types, debt, personal context)
- Add error handling tests (user not found, DB errors, auto-topup graceful fail)

94 billing tests now pass with full DI coverage.

- Fix grantCreditOperation transaction bug: use dbClient instead of db
  when tx is provided to maintain transaction boundary
- Remove placeholder test and replace with real assertion
- Add DI support to revokeGrantByOperationId so tests can call the
  actual function with injected transaction
- Update revokeGrantByOperationId tests to call real function with DI
- Add TODO comments for consumeOrganizationCredits and
  grantOrganizationCredits noting they need integration tests

- Add DI support to consumeOrganizationCredits via deps.withSerializableTransaction
- Add DI support to grantOrganizationCredits via deps.db
- Remove TODO comments that were added in previous commit
- Add 11 integration tests for consumeOrganizationCredits:
  - Priority order consumption
  - Multi-grant consumption
  - No grants error handling
  - Purchased credits tracking
- Add 6 integration tests for grantOrganizationCredits:
  - Correct values creation
  - Default description
  - Idempotency (duplicate handling)
  - Error propagation
  - Priority setting
  - Expiration date

104 billing tests now pass with full DI coverage.

- Add typed query builder interfaces (SelectQueryBuilder, InsertQueryBuilder,
  UpdateQueryBuilder) with proper method chaining types
- Add WhereResult, FromResult, OrderByResult, GroupByResult interfaces
- Add TableQuery interface for findFirst operations with typed params
- Update BillingDbConnection to use generics instead of `any`
- Add JSDoc comments with usage examples throughout
- Add OnInsertCallback, OnUpdateCallback types for tracking operations
- Add TrackedMockDbResult interface for better autocomplete
- Convert MockDbConfig from type to interface with documentation

All 104 billing tests pass with improved type safety.

The callback parameter must use `any` because the real Drizzle transaction
type (PgTransaction) has many additional properties (schema, rollback, etc.)
that our minimal BillingDbConnection does not include. Using `any` allows
both the real transaction and mock implementations to work together.

- Change error: any to error: unknown in catch blocks with type guards
- Update isUniqueConstraintError helper to handle unknown type safely
- Change logContext/logData from any to Record<string, unknown>
- Improve org-monitoring.ts to avoid delete operator with destructuring

Files improved:
- grant-credits.ts: 4 error catch blocks
- stripe-metering.ts: 2 retry callbacks
- org-billing.ts: 1 error catch block + 2 log contexts
- auto-topup.ts: 1 log context
- org-monitoring.ts: 1 log object

The billing DI refactor is complete and validated with E2E tests.
Removing the planning documents as they are no longer needed.

Revert changes to:
- cli/src/utils/auth.ts (getConfigDirFromEnvironment extraction)
- sdk/src/tools/code-search.ts (spawn DI)
- web/src/lib/ban-conditions.ts (db/stripe DI)

These DI improvements can be done in a separate PR.
This PR now focuses solely on billing DI patterns.

Auto-topup functions:
- validateAutoTopupStatus: DI for db and stripeServer
- checkAndTriggerAutoTopup: DI for db, stripe, calculateUsageAndBalance, etc.
- checkAndTriggerOrgAutoTopup: DI for db, stripe, org billing functions

Stripe metering:
- reportPurchasedCreditsToStripe: DI for db, stripe, shouldAttemptStripeMetering

Grant credits helpers:
- getPreviousFreeGrantAmount: DI for db
- calculateTotalReferralBonus: DI for db
- processAndGrantCredit: DI for grantCreditFn and logSyncFailure

New test files:
- auto-topup.test.ts: 19 tests for auto-topup DI
- stripe-metering.test.ts: 12 tests for stripe metering DI
- Updated grant-credits.test.ts with helper function tests

Type improvements:
- Added BillingOrganization type to contracts/billing.ts
- Added org query to BillingDbConnection interface
- Added org support to mock-db.ts

Added dependency injection support to:
- syncOrganizationBillingCycle: DI for db and stripeServer
- findOrganizationForRepository: DI for db
- getOrderedActiveOrganizationGrants: already had conn param

New test file: org-billing.test.ts with 12 tests covering:
- Billing cycle sync error cases
- Stripe subscription syncing
- Organization grant ordering
- Repository lookup with DB mocking

All 180+ billing tests pass.

…ageThisCycle

Added dependency injection support to:
- consumeCreditsAndAddAgentStep: DI for withSerializableTransaction, trackEvent, reportPurchasedCreditsToStripe
- calculateUsageThisCycle: DI for db

New test file: balance-calculator.test.ts with 11 tests covering:
- Database query injection
- Transaction handling with mocked dependencies
- Analytics tracking verification
- Stripe reporting verification
- BYOK user handling
- Multi-grant consumption
- Error handling for missing grants and failed inserts
- Latency calculation

All 185+ billing tests pass.