test: decouple integration contracts from E2E coverage registry

tests/e2e/coverage_registry/mgmt.yaml

@@ -90,4 +90,3 @@
 - {id: mgmt.mcp_toolset.update.persists, module: mgmt, tier: P0, surface: api, assertions: [persists], source: "mcp_management_endpoints.py:3098", rationale: "Narrowing the tools to one entry reads back exactly that entry"}
 - {id: mgmt.mcp_toolset.update.clear_persists, module: mgmt, tier: P0, surface: api, assertions: [clear_persists], source: "mcp_management_endpoints.py:3098", fail_before_fix: proven, rationale: "An explicit null clears the stored description; the update used to drop null and keep the old value"}
 - {id: mgmt.mcp_toolset.delete.persists, module: mgmt, tier: P0, surface: api, assertions: [persists], source: "mcp_management_endpoints.py:3149", rationale: "A deleted toolset is gone by id and from the list on every replica"}
-- {id: mgmt.key.update.preserves_independent_fields, module: mgmt, tier: P0, surface: api, assertions: [preserves_independent_fields], source: "proxy/management_endpoints/key_management_endpoints.py", rationale: "A partial key update preserves independent policy and metadata through serving"}

tests/e2e/coverage_registry/other.yaml

@@ -54,5 +54,3 @@
 
 - {id: other.auth.jwt.wrong_issuer_denied, module: other, tier: P0, area: auth, assertions: [wrong_issuer_denied], source: "auth/handle_jwt.py", rationale: "A signed token with the correct audience and an unexpected issuer is rejected"}
 - {id: other.auth.jwt.wrong_audience_denied, module: other, tier: P0, area: auth, assertions: [wrong_audience_denied], source: "auth/handle_jwt.py", rationale: "A signed token from the trusted issuer intended for another app is rejected"}
-- {id: other.provider_wire.internal_parameters_filtered, module: other, tier: P0, area: provider_wire, assertions: [internal_parameters_filtered], source: "main.py", rationale: "A real provider request preserves content and public parameters without internal limiter fields"}
-- {id: other.provider_wire.validator_rejects_corruption, module: other, tier: P0, area: provider_wire, assertions: [validator_rejects_corruption], source: "tests/integration/_support/upstream.py", rationale: "The controlled transport rejects missing messages and internal fields while accepting supported metadata"}

tests/e2e/coverage_registry/quota_management.yaml

@@ -63,6 +63,3 @@
 - {id: quota_management.spend_tracking.key_attribution.health_rows_keep_service_account, module: quota_management, tier: P1, behavior: spend_tracking, variant: key_attribution, assertions: [health_rows_keep_service_account], exercised_on: [chat_completions], source: "proxy/health_check.py", rationale: "A /health probe's spend row stays keyed by the literal litellm-internal-health-check service account rather than a hash of it, so health spend never appears as an unattributed key"}
 - {id: quota_management.spend_tracking.key_attribution.retrieve_batch_cost_joins_retrieving_key, module: quota_management, tier: P1, behavior: spend_tracking, variant: key_attribution, assertions: [retrieve_batch_cost_joins_retrieving_key], exercised_on: [batches], source: "proxy/batches_endpoints/endpoints.py", rationale: "The retrieve that first sees a batch in a terminal state prices it inline and writes its {provider_batch_id}_batch_cost row against the retrieving key, so the batch each run creates is one OpenAI fails at validation within seconds and the test retrieves it by its raw provider id with the same key until it is failed; a raw id is never owned by the CheckBatchCost poller, and the row must carry that key's token hash and alias"}
 - {id: quota_management.spend_tracking.key_attribution.poller_batch_cost_joins_creating_key, module: quota_management, tier: P1, behavior: spend_tracking, variant: key_attribution, assertions: [poller_batch_cost_joins_creating_key], exercised_on: [batches], source: "enterprise/litellm_enterprise/proxy/common_utils/check_batch_cost.py", rationale: "The CheckBatchCost poller bills a completed, positive-cost batch created through a unified id against the key that created it, a different writer from the inline retrieve. No test claims this cell yet: OpenAI's completion window is 24h and both e2e stacks boot a fresh Postgres per build, so a completed batch is out of one run's reach and the managed list never shows an earlier run's batch; the cell stays visible as a gap until a run can hand a completed batch to the poller"}
-- {id: quota_management.spend_tracking.custom_price.matches_input_rates, module: quota_management, tier: P0, behavior: spend_tracking, variant: custom_price, assertions: [matches_input_rates], exercised_on: [chat_completions], source: "router.py", rationale: "Configured deployment prices reach the response cost"}
-- {id: quota_management.spend_tracking.default_prices.survive_nullable_sibling_reload, module: quota_management, tier: P0, behavior: spend_tracking, variant: default_prices, assertions: [survive_nullable_sibling_reload], exercised_on: [chat_completions], source: "router.py", rationale: "Omitted and null prices retain defaults across sibling loading and reload, with persisted request charges"}
-- {id: quota_management.spend_tracking.default_prices.loaded_router_preserves_cached_defaults, module: quota_management, tier: P0, behavior: spend_tracking, variant: default_prices, assertions: [loaded_router_preserves_cached_defaults], exercised_on: [chat_completions], source: "router.py", rationale: "YAML-loaded omitted and null model-info prices preserve cached defaults across real SDK requests and reload order"}

tests/integration/README.md

@@ -8,7 +8,7 @@ Reuse the existing canned provider handlers through `_support/upstream.py`. It r
 
 The CircleCI workflow starts its own database and Redis, restricts test-phase egress to its owned services and writes JUnit plus an executed-node manifest. Missing setup, skipped tests, failed cleanup or a selected test without a passed call fail qualification. Existing GitHub Actions jobs do not own these tests
 
-Add contract definitions to the existing `tests/e2e/coverage_registry` and map canonical node IDs to those definitions in `contracts.json`. Every node must declare the same IDs with `covers`. The runner checks exact collected and passed selections against that mapping. Registry declarations alone do not mean a test passed
+Define integration contract IDs and their canonical test nodes in `contracts.json`. Every node must declare the same IDs with `covers`. The runner checks exact collected and passed selections against that mapping. These IDs belong to this CircleCI suite and must not be added to the separate E2E coverage registry. A manifest declaration alone does not mean a test passed
 
 Provider sentinels currently use the controlled server, not live recordings. The provider shard also runs the existing strict replay controls for changed requests, exhausted interactions, leftover interactions and no provider connection. Future recorded scenarios must use that replay-only implementation; missing recordings cannot fall back to a real provider. The observation endpoint is destructive and the current selection runs serially against one owned upstream
 

tests/integration/_support/manifest.py

@@ -4,8 +4,6 @@ from typing import Final
 
 from pydantic import TypeAdapter
 
-from e2e.coverage_registry.registry import load_registry
-
 MAPPING: Final = TypeAdapter(dict[str, tuple[str, ...]])
 OWNED_DIRECTORIES: Final = frozenset(
     {
@@ -28,10 +26,6 @@ OWNED_DIRECTORIES: Final = frozenset(
 def contracts() -> dict[str, tuple[str, ...]]:
     document: Final = json.loads((Path(__file__).resolve().parents[1] / "contracts.json").read_bytes())
     result: Final = MAPPING.validate_python(document["tests"])
-    if not result or any(not values for values in result.values()):
+    if not result or any(not values or any(not value.strip() for value in values) for values in result.values()):
         raise ValueError("Integration manifest must contain nodes with contract IDs")
-    registered: Final = {cell.id for cell in load_registry()}
-    unknown: Final = {identity for identities in result.values() for identity in identities} - registered
-    if unknown:
-        raise ValueError(f"Unknown coverage-registry contracts: {sorted(unknown)}")
     return result