From ea37596b884056b2c818f8945d71e394b5126c7d Mon Sep 17 00:00:00 2001 From: yassin Date: Fri, 18 Sep 2026 01:01:26 +0000 Subject: [PATCH 1/6] feat(mcp): let proxy admins force-close live MCP sessions and revoke stored user credentials Adds an admin-only DELETE /v1/mcp/sessions that terminates stateful MCP gateway sessions on the current worker by session id prefix and/or by the LiteLLM user that opened them, tombstones the terminated ids so a client reusing one gets 404 instead of a silently recreated stateless session, and lets PROXY_ADMIN name a user_id on the BYOK and OAuth credential delete routes. Full and view-only admins can list every user's stored credential metadata for a server (never the secret). The dashboard gains Disconnect controls on the Live Connections tab and a User Credentials tab with Revoke controls, both hidden from read-only admins. Resolves LIT-8001 Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 2 + litellm/proxy/_experimental/mcp_server/db.py | 32 ++ .../proxy/_experimental/mcp_server/server.py | 72 +++- litellm/proxy/_lazy_openapi_snapshot.json | 237 ++++++++++- litellm/proxy/_types.py | 10 + .../mcp_management_endpoints.py | 129 +++++- litellm/types/mcp.py | 8 + .../mcp_server/test_db_credentials.py | 47 +++ .../mcp_server/test_mcp_server.py | 177 ++++++++ .../test_mcp_management_endpoints.py | 398 ++++++++++++++++++ ...MCPGatewaySessionsTab.integration.test.tsx | 90 +++- .../_components/MCPGatewaySessionsTab.tsx | 146 ++++++- ...rUserCredentialsPanel.integration.test.tsx | 102 +++++ .../MCPServerUserCredentialsPanel.tsx | 212 ++++++++++ .../_components/mcp_server_view.tsx | 21 + .../mcp-servers/_components/mcp_servers.tsx | 4 +- .../src/components/mcp_tools/types.tsx | 20 + .../src/components/networking.tsx | 36 +- ui/litellm-dashboard/src/lib/http/schema.d.ts | 132 +++++- 19 files changed, 1826 insertions(+), 49 deletions(-) create mode 100644 ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.integration.test.tsx create mode 100644 ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.tsx diff --git a/litellm/constants.py b/litellm/constants.py index d4827bb7483..2ebc9beb632 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -183,6 +183,8 @@ MCP_TOOL_LISTING_TIMEOUT: Final = float(os.getenv("LITELLM_MCP_TOOL_LISTING_TIME MCP_METADATA_TIMEOUT: Final = float(os.getenv("LITELLM_MCP_METADATA_TIMEOUT", "10.0")) MCP_HEALTH_CHECK_TIMEOUT: Final = float(os.getenv("LITELLM_MCP_HEALTH_CHECK_TIMEOUT", "10.0")) MCP_TOOL_LISTING_MAX_PAGES: Final = 1000 +MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH: Final = 8 +MCP_ADMIN_TERMINATED_SESSION_IDS_MAX: Final = 1024 # Allowlist of commands permitted for MCP stdio transport. # Prevents arbitrary command execution via /mcp-rest/test/* endpoints or server creation. diff --git a/litellm/proxy/_experimental/mcp_server/db.py b/litellm/proxy/_experimental/mcp_server/db.py index 789b2ffaef4..a04e2f5c9b8 100644 --- a/litellm/proxy/_experimental/mcp_server/db.py +++ b/litellm/proxy/_experimental/mcp_server/db.py @@ -24,6 +24,7 @@ from litellm.proxy._types import ( MCPApprovalStatus, MCPEnvVar, MCPEnvVarScope, + MCPServerUserCredentialListItem, MCPSubmissionsSummary, NewMCPServerRequest, SpecialMCPServerName, @@ -1504,6 +1505,37 @@ async def get_user_oauth_credential( return _parse_oauth_payload(decoded) +def _server_user_credential_item( + row: "prisma_db_models.LiteLLM_MCPUserCredentials", +) -> MCPServerUserCredentialListItem: + oauth_payload: Final = _decode_oauth_payload(row.credential_b64) + if oauth_payload is None: + return MCPServerUserCredentialListItem( + user_id=row.user_id, + credential_type="byok", + updated_at=row.updated_at.isoformat(), + ) + return MCPServerUserCredentialListItem( + user_id=row.user_id, + credential_type="oauth2", + expires_at=oauth_payload.get("expires_at"), + connected_at=oauth_payload.get("connected_at"), + updated_at=row.updated_at.isoformat(), + ) + + +async def list_server_user_credentials( + prisma_client: PrismaClient, + server_id: str, +) -> tuple[MCPServerUserCredentialListItem, ...]: + """Every user's stored credential for one server, typed but without the secret, for admins.""" + rows: Final = await _db_find_user_credential_rows( + prisma_client, + {"server_id": server_id}, # mutable-ok: prisma where-inputs must be plain dicts + ) + return tuple(_server_user_credential_item(row) for row in rows) + + async def list_user_oauth_credentials( prisma_client: PrismaClient, user_id: str, diff --git a/litellm/proxy/_experimental/mcp_server/server.py b/litellm/proxy/_experimental/mcp_server/server.py index 524bac747ad..80d9274859d 100644 --- a/litellm/proxy/_experimental/mcp_server/server.py +++ b/litellm/proxy/_experimental/mcp_server/server.py @@ -14,7 +14,7 @@ import time import traceback import types import uuid -from collections import Counter +from collections import Counter, deque from collections.abc import AsyncIterator, Callable, Mapping, Sequence from datetime import datetime from typing import TYPE_CHECKING, Any, Final, NoReturn, Protocol @@ -28,7 +28,11 @@ from starlette.types import Message, Receive, Scope, Send from typing_extensions import ReadOnly, TypedDict from litellm._logging import verbose_logger -from litellm.constants import MAXIMUM_TRACEBACK_LINES_TO_LOG +from litellm.constants import ( + MAXIMUM_TRACEBACK_LINES_TO_LOG, + MCP_ADMIN_TERMINATED_SESSION_IDS_MAX, + MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH, +) from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj from litellm.llms.custom_httpx.http_handler import ( get_async_httpx_client, @@ -91,6 +95,7 @@ from litellm.types.mcp import ( MCPGatewaySession, MCPGatewaySessionGroupCount, MCPGatewaySessionsResponse, + MCPGatewaySessionsTerminateResponse, MCPSpecVersion, ) from litellm.types.mcp_server.mcp_server_manager import MCPInfo, MCPServer @@ -618,6 +623,9 @@ if MCP_AVAILABLE: _stateful_session_locks: Final[dict[str, asyncio.Lock]] = {} _stateful_session_active_request_counts: Final[dict[str, int]] = {} _stateful_session_client_info: Final[dict[str, Implementation]] = {} # mutable-ok: cleared on session teardown + _admin_terminated_session_ids: Final[deque[str]] = deque( # mutable-ok: bounded ring, appended on admin termination + maxlen=MCP_ADMIN_TERMINATED_SESSION_IDS_MAX + ) class _TerminableTransport(Protocol): async def terminate(self) -> None: ... @@ -3850,7 +3858,7 @@ if MCP_AVAILABLE: client_info: Final = _stateful_session_client_info.get(session_id) key_auth: Final = auth_user.user_api_key_auth return MCPGatewaySession( - session_id_prefix=session_id[:8], + session_id_prefix=session_id[:MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH], client_name=client_info.name if client_info is not None else None, client_version=client_info.version if client_info is not None else None, user_id=key_auth.user_id if key_auth is not None else None, @@ -3885,6 +3893,53 @@ if MCP_AVAILABLE: sessions=sessions, ) + def _session_matches_admin_selector( + session_id: str, + auth_user: MCPAuthenticatedUser, + session_id_prefix: str | None, + user_id: str | None, + ) -> bool: + if session_id_prefix is not None and not session_id.startswith(session_id_prefix): + return False + if user_id is None: + return True + key_auth: Final = auth_user.user_api_key_auth + return key_auth is not None and key_auth.user_id == user_id + + async def terminate_mcp_gateway_sessions( + *, + session_id_prefix: str | None = None, + user_id: str | None = None, + ) -> MCPGatewaySessionsTerminateResponse: + """Force-close every live stateful session on this worker matching the selector. + + The transport is terminated (open streams close), all per-session + tracking is dropped, and the id is remembered so a client that keeps + sending it receives 404 and has to ``initialize`` again, which re-runs + admission. Only sessions held by this worker process are affected. + """ + now: Final = time.monotonic() + server_instances: Final = _stateful_server_instances() + targets: Final = tuple( + (session_id, auth_user) + for session_id, auth_user in tuple(_stateful_session_auth_contexts.items()) + if session_id in server_instances + and _session_matches_admin_selector(session_id, auth_user, session_id_prefix, user_id) + ) + terminated: Final = tuple(_gateway_session_for(session_id, auth_user, now) for session_id, auth_user in targets) + for session_id, _ in targets: + _admin_terminated_session_ids.append(session_id) + transport = server_instances.pop(session_id, None) + _remove_stateful_session_tracking(session_id) + if transport is not None: + await transport.terminate() + verbose_logger.warning("MCP session '%s' terminated by an administrator.", session_id) + return MCPGatewaySessionsTerminateResponse( + worker_pid=os.getpid(), + terminated_sessions=len(terminated), + sessions=terminated, + ) + async def _read_request_body_for_routing( receive: Receive, ) -> tuple[list[Message], bytes]: @@ -4009,6 +4064,17 @@ if MCP_AVAILABLE: await success_response(scope, receive, send) return True + if _session_id in _admin_terminated_session_ids: + terminated_response: Final = JSONResponse( + status_code=404, + content={ # mutable-ok: JSONResponse content must be a plain dict + "error": "Not Found", + "details": "mcp-session-id was terminated by an administrator. Send initialize to start a new session.", + }, + ) + await terminated_response(scope, receive, send) + return True + # Non-DELETE: strip stale session ID to allow new session creation verbose_logger.warning( "MCP session ID '%s' not found in this worker's memory. " diff --git a/litellm/proxy/_lazy_openapi_snapshot.json b/litellm/proxy/_lazy_openapi_snapshot.json index 82b709feb92..1e9b1f5b743 100644 --- a/litellm/proxy/_lazy_openapi_snapshot.json +++ b/litellm/proxy/_lazy_openapi_snapshot.json @@ -27648,6 +27648,32 @@ "title": "MCPGatewaySessionsResponse", "type": "object" }, + "MCPGatewaySessionsTerminateResponse": { + "description": "Stateful sessions an administrator force-closed on this proxy worker.", + "properties": { + "sessions": { + "items": { + "$ref": "#/components/schemas/MCPGatewaySession" + }, + "title": "Sessions", + "type": "array" + }, + "terminated_sessions": { + "title": "Terminated Sessions", + "type": "integer" + }, + "worker_pid": { + "title": "Worker Pid", + "type": "integer" + } + }, + "required": [ + "worker_pid", + "terminated_sessions" + ], + "title": "MCPGatewaySessionsTerminateResponse", + "type": "object" + }, "MCPOAuthUserCredentialRequest": { "description": "Stores a user's OAuth2 token for an OpenAPI MCP server.", "properties": { @@ -27744,6 +27770,56 @@ "title": "MCPOAuthUserCredentialStatus", "type": "object" }, + "MCPServerUserCredentialListItem": { + "description": "One user's stored credential for an MCP server, as an admin sees it. Never carries the secret.", + "properties": { + "connected_at": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Connected At" + }, + "credential_type": { + "enum": [ + "oauth2", + "byok" + ], + "title": "Credential Type", + "type": "string" + }, + "expires_at": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Expires At" + }, + "updated_at": { + "title": "Updated At", + "type": "string" + }, + "user_id": { + "title": "User Id", + "type": "string" + } + }, + "required": [ + "user_id", + "credential_type", + "updated_at" + ], + "title": "MCPServerUserCredentialListItem", + "type": "object" + }, "MCPSubmissionsSummary": { "properties": { "active": { @@ -29920,7 +29996,7 @@ }, "/v1/mcp/server/{server_id}/oauth-user-credential": { "delete": { - "description": "Revoke the calling user's stored OAuth2 token for an MCP server", + "description": "Revoke the calling user's stored OAuth2 token for an MCP server. A proxy admin may pass user_id to revoke another user's stored token.", "operationId": "delete_mcp_oauth_user_credential_v1_mcp_server__server_id__oauth_user_credential_delete", "parameters": [ { @@ -29931,6 +30007,23 @@ "title": "Server Id", "type": "string" } + }, + { + "in": "query", + "name": "user_id", + "required": false, + "schema": { + "anyOf": [ + { + "minLength": 1, + "type": "string" + }, + { + "type": "null" + } + ], + "title": "User Id" + } } ], "responses": { @@ -30130,7 +30223,7 @@ }, "/v1/mcp/server/{server_id}/user-credential": { "delete": { - "description": "Delete the calling user's stored API key for a BYOK MCP server", + "description": "Delete the calling user's stored API key for a BYOK MCP server. A proxy admin may pass user_id to revoke another user's stored key.", "operationId": "delete_mcp_user_credential_v1_mcp_server__server_id__user_credential_delete", "parameters": [ { @@ -30141,6 +30234,23 @@ "title": "Server Id", "type": "string" } + }, + { + "in": "query", + "name": "user_id", + "required": false, + "schema": { + "anyOf": [ + { + "minLength": 1, + "type": "string" + }, + { + "type": "null" + } + ], + "title": "User Id" + } } ], "responses": { @@ -30232,6 +30342,58 @@ ] } }, + "/v1/mcp/server/{server_id}/user-credentials": { + "get": { + "description": "List every user's stored BYOK or OAuth2 credential for an MCP server (admin only, no secrets)", + "operationId": "list_mcp_server_user_credentials_v1_mcp_server__server_id__user_credentials_get", + "parameters": [ + { + "in": "path", + "name": "server_id", + "required": true, + "schema": { + "title": "Server Id", + "type": "string" + } + } + ], + "responses": { + "200": { + "content": { + "application/json": { + "schema": { + "items": { + "$ref": "#/components/schemas/MCPServerUserCredentialListItem" + }, + "title": "Response List Mcp Server User Credentials V1 Mcp Server Server Id User Credentials Get", + "type": "array" + } + } + }, + "description": "Successful Response" + }, + "422": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + }, + "description": "Validation Error" + } + }, + "security": [ + { + "APIKeyHeader": [] + } + ], + "summary": "List Mcp Server User Credentials", + "tags": [ + "mcp_management" + ] + } + }, "/v1/mcp/server/{server_id}/user-env-vars": { "delete": { "description": "Clear the calling user's per-user MCP env var values for this server.", @@ -30383,6 +30545,77 @@ } }, "/v1/mcp/sessions": { + "delete": { + "description": "Force-close live stateful MCP gateway sessions on this proxy worker, selected by session id prefix and/or by the LiteLLM user that opened them (proxy admin only).", + "operationId": "delete_mcp_gateway_sessions_v1_mcp_sessions_delete", + "parameters": [ + { + "in": "query", + "name": "session_id_prefix", + "required": false, + "schema": { + "anyOf": [ + { + "minLength": 8, + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id Prefix" + } + }, + { + "in": "query", + "name": "user_id", + "required": false, + "schema": { + "anyOf": [ + { + "minLength": 1, + "type": "string" + }, + { + "type": "null" + } + ], + "title": "User Id" + } + } + ], + "responses": { + "200": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/MCPGatewaySessionsTerminateResponse" + } + } + }, + "description": "Successful Response" + }, + "422": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + }, + "description": "Validation Error" + } + }, + "security": [ + { + "APIKeyHeader": [] + } + ], + "summary": "Delete Mcp Gateway Sessions", + "tags": [ + "mcp_management" + ] + }, "get": { "description": "Live stateful MCP gateway sessions on this proxy worker, grouped by AI client and by user.", "operationId": "get_mcp_gateway_sessions_v1_mcp_sessions_get", diff --git a/litellm/proxy/_types.py b/litellm/proxy/_types.py index ff32d4784df..ee92e64b046 100644 --- a/litellm/proxy/_types.py +++ b/litellm/proxy/_types.py @@ -1716,6 +1716,16 @@ class MCPUserCredentialListItem(LiteLLMPydanticObjectBase): connected_at: str | None = None # ISO-8601 +class MCPServerUserCredentialListItem(LiteLLMPydanticObjectBase): + """One user's stored credential for an MCP server, as an admin sees it. Never carries the secret.""" + + user_id: str + credential_type: Literal["oauth2", "byok"] + expires_at: str | None = None + connected_at: str | None = None + updated_at: str + + class MCPUserEnvVarsRequest(LiteLLMPydanticObjectBase): """Payload for storing the calling user's per-user env var values.""" diff --git a/litellm/proxy/management_endpoints/mcp_management_endpoints.py b/litellm/proxy/management_endpoints/mcp_management_endpoints.py index 82a1cdcdd00..728ba9568e8 100644 --- a/litellm/proxy/management_endpoints/mcp_management_endpoints.py +++ b/litellm/proxy/management_endpoints/mcp_management_endpoints.py @@ -47,7 +47,7 @@ except ImportError: import litellm from litellm._logging import verbose_logger, verbose_proxy_logger from litellm._uuid import uuid -from litellm.constants import LITELLM_PROXY_ADMIN_NAME +from litellm.constants import LITELLM_PROXY_ADMIN_NAME, MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH from litellm.proxy._experimental.mcp_server.utils import ( LITELLM_MCP_SERVER_DESCRIPTION, LITELLM_MCP_SERVER_NAME, @@ -145,6 +145,7 @@ if MCP_AVAILABLE: get_user_env_vars, get_user_env_vars_bulk, get_user_oauth_credential, + list_server_user_credentials, list_user_oauth_credentials, mcp_oauth_token_identity, merge_user_env_vars, @@ -180,6 +181,7 @@ if MCP_AVAILABLE: MCPApprovalStatus, MCPOAuthUserCredentialRequest, MCPOAuthUserCredentialStatus, + MCPServerUserCredentialListItem, MCPSubmissionsSummary, MCPTransport, MCPUserCredentialListItem, @@ -221,6 +223,7 @@ if MCP_AVAILABLE: MCPAuth, MCPCredentials, MCPGatewaySessionsResponse, + MCPGatewaySessionsTerminateResponse, normalize_upstream_header_name, ) from litellm.types.mcp_server.mcp_server_manager import MCPServer @@ -662,6 +665,31 @@ if MCP_AVAILABLE: """ return user_api_key_dict.user_role == LitellmUserRoles.PROXY_ADMIN + def _resolve_credential_target_user_id(user_api_key_dict: UserAPIKeyAuth, requested_user_id: str | None) -> str: + """The user whose stored MCP credential a request acts on. + + Defaults to the caller. Naming another user is a revocation and needs + ``PROXY_ADMIN``; a read-only admin or a regular user gets 403. + """ + caller_user_id: Final = user_api_key_dict.user_id or "" + if requested_user_id is not None and requested_user_id != caller_user_id: + if not _user_is_full_admin(user_api_key_dict): + raise HTTPException( + status_code=status.HTTP_403_FORBIDDEN, + detail={ # mutable-ok: FastAPI HTTPException detail requires a plain dict + "error": "Proxy admin access required to revoke another user's MCP credential.", + }, + ) + return requested_user_id + if not caller_user_id: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail={ + "error": "User ID not found in token" + }, # mutable-ok: FastAPI HTTPException detail requires a plain dict + ) + return caller_user_id + def _is_restricted_virtual_key_request(user_api_key_dict: UserAPIKeyAuth) -> bool: """Best-effort detection for route-restricted virtual keys. @@ -1373,6 +1401,41 @@ if MCP_AVAILABLE: return get_mcp_gateway_sessions_report() + @router.delete( + "/sessions", + description=( + "Force-close live stateful MCP gateway sessions on this proxy worker, selected by session id prefix " + "and/or by the LiteLLM user that opened them (proxy admin only)." + ), + dependencies=(Depends(user_api_key_auth),), + response_model=MCPGatewaySessionsTerminateResponse, + ) + @management_endpoint_wrapper + async def delete_mcp_gateway_sessions( + user_api_key_dict: Annotated[UserAPIKeyAuth, Depends(user_api_key_auth)], + session_id_prefix: Annotated[str | None, Query(min_length=MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH)] = None, + user_id: Annotated[str | None, Query(min_length=1)] = None, + ) -> MCPGatewaySessionsTerminateResponse: + if not _user_is_full_admin(user_api_key_dict): + raise HTTPException( + status_code=status.HTTP_403_FORBIDDEN, + detail={ # mutable-ok: FastAPI HTTPException detail requires a plain dict + "error": "Proxy admin access required to terminate MCP gateway sessions.", + }, + ) + if session_id_prefix is None and user_id is None: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail={ # mutable-ok: FastAPI HTTPException detail requires a plain dict + "error": "Provide session_id_prefix and/or user_id to select the sessions to terminate.", + }, + ) + from litellm.proxy._experimental.mcp_server.server import ( + terminate_mcp_gateway_sessions, + ) + + return await terminate_mcp_gateway_sessions(session_id_prefix=session_id_prefix, user_id=user_id) + @router.get( "/server/submissions", description="Returns all MCP servers submitted by non-admin users (admin review queue). Mirrors GET /guardrails/submissions.", @@ -2261,7 +2324,10 @@ if MCP_AVAILABLE: @router.delete( "/server/{server_id}/user-credential", - description="Delete the calling user's stored API key for a BYOK MCP server", + description=( + "Delete the calling user's stored API key for a BYOK MCP server. " + "A proxy admin may pass user_id to revoke another user's stored key." + ), dependencies=[Depends(user_api_key_auth)], response_model=MCPUserCredentialResponse, ) @@ -2269,24 +2335,20 @@ if MCP_AVAILABLE: async def delete_mcp_user_credential( server_id: str, user_api_key_dict: UserAPIKeyAuth = Depends(user_api_key_auth), + user_id: Annotated[str | None, Query(min_length=1)] = None, ): - """Remove the calling user's BYOK credential.""" + """Remove the target user's BYOK credential (the caller unless an admin names another user).""" prisma_client: Final = get_prisma_client_or_throw("Database not connected. Connect a database to your proxy") - user_id: Final = user_api_key_dict.user_id or "" - if not user_id: - raise HTTPException( - status_code=status.HTTP_400_BAD_REQUEST, - detail={"error": "User ID not found in token"}, - ) + target_user_id: Final = _resolve_credential_target_user_id(user_api_key_dict, user_id) try: - await delete_user_credential(prisma_client, user_id, server_id) + await delete_user_credential(prisma_client, target_user_id, server_id) except RecordNotFoundError: pass # Already deleted or didn't exist from litellm.proxy._experimental.mcp_server.server import ( _invalidate_byok_cred_cache, ) - _invalidate_byok_cred_cache(user_id, server_id) + _invalidate_byok_cred_cache(target_user_id, server_id) return MCPUserCredentialResponse(server_id=server_id, has_credential=False) # ── OAuth2 user-credential endpoints ────────────────────────────────────── @@ -2362,7 +2424,10 @@ if MCP_AVAILABLE: @router.delete( "/server/{server_id}/oauth-user-credential", - description="Revoke the calling user's stored OAuth2 token for an MCP server", + description=( + "Revoke the calling user's stored OAuth2 token for an MCP server. " + "A proxy admin may pass user_id to revoke another user's stored token." + ), dependencies=[Depends(user_api_key_auth)], response_model=MCPOAuthUserCredentialStatus, ) @@ -2370,29 +2435,25 @@ if MCP_AVAILABLE: async def delete_mcp_oauth_user_credential( server_id: str, user_api_key_dict: UserAPIKeyAuth = Depends(user_api_key_auth), + user_id: Annotated[str | None, Query(min_length=1)] = None, ): - """Revoke/delete the user's OAuth2 credential.""" + """Revoke the target user's OAuth2 credential (the caller unless an admin names another user).""" prisma_client: Final = get_prisma_client_or_throw("Database not connected. Connect a database to your proxy") - user_id: Final = user_api_key_dict.user_id or "" - if not user_id: - raise HTTPException( - status_code=status.HTTP_400_BAD_REQUEST, - detail={"error": "User ID not found in token"}, - ) + target_user_id: Final = _resolve_credential_target_user_id(user_api_key_dict, user_id) # Only delete if the stored credential is actually an OAuth2 token. # This prevents accidentally deleting a BYOK credential if one exists # for the same (user_id, server_id) pair. - cred_to_delete: Final = await get_user_oauth_credential(prisma_client, user_id, server_id) + cred_to_delete: Final = await get_user_oauth_credential(prisma_client, target_user_id, server_id) if cred_to_delete is not None: try: - await delete_user_credential(prisma_client, user_id, server_id) + await delete_user_credential(prisma_client, target_user_id, server_id) except RecordNotFoundError: pass # Already gone — treat as a successful delete from litellm.proxy._experimental.mcp_server.mcp_server_manager import ( # noqa: PLC0415 global_mcp_server_manager, ) - await global_mcp_server_manager.invalidate_user_oauth_token_cache(user_id, server_id) + await global_mcp_server_manager.invalidate_user_oauth_token_cache(target_user_id, server_id) return MCPOAuthUserCredentialStatus( server_id=server_id, has_credential=False, @@ -2481,6 +2542,30 @@ if MCP_AVAILABLE: ) return items + @router.get( + "/server/{server_id}/user-credentials", + description="List every user's stored BYOK or OAuth2 credential for an MCP server (admin only, no secrets)", + dependencies=(Depends(user_api_key_auth),), + response_model=list[MCPServerUserCredentialListItem], + ) + @management_endpoint_wrapper + async def list_mcp_server_user_credentials( + server_id: str, + user_api_key_dict: Annotated[UserAPIKeyAuth, Depends(user_api_key_auth)], + ) -> tuple[MCPServerUserCredentialListItem, ...]: + if user_api_key_dict.user_role not in ( + LitellmUserRoles.PROXY_ADMIN, + LitellmUserRoles.PROXY_ADMIN_VIEW_ONLY, + ): + raise HTTPException( + status_code=status.HTTP_403_FORBIDDEN, + detail={ # mutable-ok: FastAPI HTTPException detail requires a plain dict + "error": "Admin access required to view MCP server user credentials.", + }, + ) + prisma_client: Final = get_prisma_client_or_throw("Database not connected. Connect a database to your proxy") + return await list_server_user_credentials(prisma_client, server_id) + # ── Per-user MCP env var endpoints ──────────────────────────────────────── async def _authorize_and_fetch_mcp_server( diff --git a/litellm/types/mcp.py b/litellm/types/mcp.py index 2d06bb9a009..c5a26c997b7 100644 --- a/litellm/types/mcp.py +++ b/litellm/types/mcp.py @@ -464,3 +464,11 @@ class MCPGatewaySessionsResponse(BaseModel): by_client: list[MCPGatewaySessionGroupCount] = Field(default_factory=list) by_user: list[MCPGatewaySessionGroupCount] = Field(default_factory=list) sessions: list[MCPGatewaySession] = Field(default_factory=list) + + +class MCPGatewaySessionsTerminateResponse(BaseModel): + """Stateful sessions an administrator force-closed on this proxy worker.""" + + worker_pid: int + terminated_sessions: int + sessions: list[MCPGatewaySession] = Field(default_factory=list) diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_db_credentials.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_db_credentials.py index 60a5e1a22bb..cfcff73b857 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_db_credentials.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_db_credentials.py @@ -212,6 +212,53 @@ async def test_purge_user_oauth_credentials_for_server_invalidates_each_user(): assert set(invalidations) == {("alice", "srv-1"), ("bob", "srv-1")} +@pytest.mark.asyncio +async def test_list_server_user_credentials_types_each_row_without_leaking_the_secret(): + """The admin view of one server's stored credentials names the user and the kind of + credential (OAuth2 vs BYOK) and echoes OAuth expiry, but never the token or key itself.""" + from litellm.proxy._experimental.mcp_server.db import list_server_user_credentials + + oauth_row = _legacy_row( + json.dumps( + { + "type": "oauth2", + "access_token": "tok-alice", + "expires_at": "2026-12-31T00:00:00+00:00", + "connected_at": "2026-01-01T00:00:00+00:00", + } + ) + ) + oauth_row.user_id = "alice" + oauth_row.updated_at = datetime(2026, 1, 1, tzinfo=timezone.utc) + byok_row = _byok_row("carol") + byok_row.updated_at = datetime(2026, 2, 1, tzinfo=timezone.utc) + prisma = MagicMock() + prisma.db.litellm_mcpusercredentials.find_many = AsyncMock(return_value=[oauth_row, byok_row]) + + items = await list_server_user_credentials(prisma, "srv-1") + + prisma.db.litellm_mcpusercredentials.find_many.assert_awaited_once_with(where={"server_id": "srv-1"}) + assert [item.model_dump() for item in items] == [ + { + "user_id": "alice", + "credential_type": "oauth2", + "expires_at": "2026-12-31T00:00:00+00:00", + "connected_at": "2026-01-01T00:00:00+00:00", + "updated_at": "2026-01-01T00:00:00+00:00", + }, + { + "user_id": "carol", + "credential_type": "byok", + "expires_at": None, + "connected_at": None, + "updated_at": "2026-02-01T00:00:00+00:00", + }, + ] + serialized = "".join(item.model_dump_json() for item in items) + assert "tok-alice" not in serialized + assert "sk-byok-carol" not in serialized + + @pytest.mark.asyncio async def test_purge_user_oauth_credentials_for_server_spares_byok_rows(): """Regression: the purge used to delete_many on server_id alone, wiping BYOK API keys that share diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py index ef424255f04..4311a69d465 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py @@ -2871,6 +2871,183 @@ def test_remove_stateful_session_tracking_drops_client_info(): assert session_id not in mcp_server._stateful_session_client_info +def _admin_terminate_fixture(mcp_server): + def auth_user(user_id: str): + return mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key=f"key-{user_id}", user_id=user_id), + ) + + contexts = { + "alice-session-1": auth_user("alice"), + "alice-session-2": auth_user("alice"), + "bob-session-1": auth_user("bob"), + "anon-session-1": mcp_server.MCPAuthenticatedUser(user_api_key_auth=None), + "gone-session-1": auth_user("alice"), + } + transports = { + session_id: MagicMock(terminate=AsyncMock()) + for session_id in ("alice-session-1", "alice-session-2", "bob-session-1", "anon-session-1") + } + return contexts, transports + + +@pytest.mark.asyncio +async def test_terminate_mcp_gateway_sessions_by_user_closes_every_live_session_of_that_user(): + try: + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy._experimental.mcp_server.server import session_manager_stateful + except ImportError: + pytest.skip("MCP server not available") + + contexts, transports = _admin_terminate_fixture(mcp_server) + live_transports = dict(transports) + last_seen = {session_id: 100.0 for session_id in contexts} + locks = {session_id: asyncio.Lock() for session_id in contexts} + + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + session_manager_stateful, "_server_instances", live_transports + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, contexts, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_context_last_seen, last_seen, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_locks, locks, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_owners, {session_id: "owner" for session_id in contexts}, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_active_request_counts, {}, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_client_info, {}, clear=True + ), + ): + result = await mcp_server.terminate_mcp_gateway_sessions(user_id="alice") + + assert set(live_transports) == {"bob-session-1", "anon-session-1"} + assert set(mcp_server._stateful_session_auth_contexts) == {"bob-session-1", "anon-session-1", "gone-session-1"} + assert set(mcp_server._stateful_session_locks) == {"bob-session-1", "anon-session-1", "gone-session-1"} + assert set(mcp_server._stateful_session_owners) == {"bob-session-1", "anon-session-1", "gone-session-1"} + assert set(mcp_server._stateful_session_auth_context_last_seen) == { + "bob-session-1", + "anon-session-1", + "gone-session-1", + } + + transports["alice-session-1"].terminate.assert_awaited_once() + transports["alice-session-2"].terminate.assert_awaited_once() + transports["bob-session-1"].terminate.assert_not_awaited() + transports["anon-session-1"].terminate.assert_not_awaited() + assert result.terminated_sessions == 2 + assert sorted(session.session_id_prefix for session in result.sessions) == ["alice-se", "alice-se"] + assert {session.user_id for session in result.sessions} == {"alice"} + assert "key-alice" not in result.model_dump_json() + assert "alice-session-1" not in result.model_dump_json() + + +@pytest.mark.asyncio +async def test_terminate_mcp_gateway_sessions_prefix_and_user_must_both_match(): + try: + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy._experimental.mcp_server.server import session_manager_stateful + except ImportError: + pytest.skip("MCP server not available") + + contexts, transports = _admin_terminate_fixture(mcp_server) + live_transports = dict(transports) + + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + session_manager_stateful, "_server_instances", live_transports + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, contexts, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_client_info, {}, clear=True + ), + ): + mismatch = await mcp_server.terminate_mcp_gateway_sessions(session_id_prefix="alice-session-1", user_id="bob") + assert mismatch.terminated_sessions == 0 + assert set(live_transports) == set(transports) + + stale = await mcp_server.terminate_mcp_gateway_sessions(session_id_prefix="gone-session-1") + assert stale.terminated_sessions == 0 + + exact = await mcp_server.terminate_mcp_gateway_sessions(session_id_prefix="alice-session-1", user_id="alice") + assert exact.terminated_sessions == 1 + assert set(live_transports) == {"alice-session-2", "bob-session-1", "anon-session-1"} + + +@pytest.mark.asyncio +async def test_admin_terminated_session_id_gets_404_instead_of_a_fresh_stateless_session(): + """Once an admin closes a session, a client replaying its id must not be silently upgraded to a + new stateless session by the stale-header path; it gets 404 and has to initialize again.""" + try: + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy._experimental.mcp_server.server import session_manager_stateful + from starlette.types import Scope + except ImportError: + pytest.skip("MCP server not available") + + session_id = "admin-closed-session-1" + live_transports = {session_id: MagicMock(terminate=AsyncMock())} + contexts = { + session_id: mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key="key-alice", user_id="alice"), + ) + } + + def scope_with_session_header() -> Scope: + return { + "type": "http", + "method": "POST", + "headers": [(b"content-type", b"application/json"), (b"mcp-session-id", session_id.encode())], + } + + try: + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + session_manager_stateful, "_server_instances", live_transports + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, contexts, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_client_info, {}, clear=True + ), + ): + await mcp_server.terminate_mcp_gateway_sessions(session_id_prefix=session_id) + + terminated_scope = scope_with_session_header() + send = AsyncMock() + handled = await mcp_server._handle_stale_mcp_session( + terminated_scope, AsyncMock(), send, session_manager_stateful + ) + + assert handled is True + statuses = [m["status"] for (m,), _ in send.await_args_list if m["type"] == "http.response.start"] + assert statuses == [404] + assert [k for k, _ in terminated_scope["headers"]] == [b"content-type", b"mcp-session-id"] + + unknown_scope = scope_with_session_header() + unknown_scope["headers"][1] = (b"mcp-session-id", b"never-seen-session") + assert ( + await mcp_server._handle_stale_mcp_session( + unknown_scope, AsyncMock(), AsyncMock(), session_manager_stateful + ) + is False + ) + assert [k for k, _ in unknown_scope["headers"]] == [b"content-type"] + finally: + mcp_server._admin_terminated_session_ids.clear() + + @pytest.mark.asyncio async def test_initialize_request_with_existing_session_tracks_new_session(): try: diff --git a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py index 7c874aff3df..2f33018599f 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py @@ -5136,6 +5136,261 @@ async def test_delete_mcp_oauth_user_credential_invalidates_when_record_already_ assert result.has_credential is False +def _make_admin_auth(role: LitellmUserRoles = LitellmUserRoles.PROXY_ADMIN) -> "UserAPIKeyAuth": + return UserAPIKeyAuth(api_key="sk-admin", user_id="admin-user", user_role=role) + + +@pytest.mark.asyncio +async def test_admin_revokes_another_users_byok_credential(): + """A proxy admin naming user_id deletes and cache-invalidates that user's stored key, not their own.""" + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_user_credential, + ) + + delete_mock = AsyncMock(return_value=None) + invalidate_mock = MagicMock() + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the credential row delete + "litellm.proxy.management_endpoints.mcp_management_endpoints.delete_user_credential", + new=delete_mock, + ), + patch.object( # test-quality-ok: the cache invalidator is module scoped; the suite's only seam + mcp_server, "_invalidate_byok_cred_cache", new=invalidate_mock + ), + ): + result = await delete_mcp_user_credential( + server_id="srv-byok-admin", + user_api_key_dict=_make_admin_auth(), + user_id="mallory", + ) + + delete_mock.assert_awaited_once() + assert delete_mock.await_args.args[1:] == ("mallory", "srv-byok-admin") + invalidate_mock.assert_called_once_with("mallory", "srv-byok-admin") + assert result.has_credential is False + + +@pytest.mark.asyncio +@pytest.mark.parametrize("role", [LitellmUserRoles.INTERNAL_USER, LitellmUserRoles.PROXY_ADMIN_VIEW_ONLY]) +async def test_non_full_admin_cannot_revoke_another_users_byok_credential(role): + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_user_credential, + ) + + delete_mock = AsyncMock(return_value=None) + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the credential row delete + "litellm.proxy.management_endpoints.mcp_management_endpoints.delete_user_credential", + new=delete_mock, + ), + ): + with pytest.raises(HTTPException) as exc_info: + await delete_mcp_user_credential( + server_id="srv-byok-forbidden", + user_api_key_dict=_make_admin_auth(role), + user_id="mallory", + ) + + assert exc_info.value.status_code == 403 + delete_mock.assert_not_awaited() + + +@pytest.mark.asyncio +async def test_user_naming_themselves_still_deletes_own_byok_credential(): + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_user_credential, + ) + + delete_mock = AsyncMock(return_value=None) + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the credential row delete + "litellm.proxy.management_endpoints.mcp_management_endpoints.delete_user_credential", + new=delete_mock, + ), + patch.object( # test-quality-ok: the cache invalidator is module scoped; the suite's only seam + mcp_server, "_invalidate_byok_cred_cache", new=MagicMock() + ), + ): + await delete_mcp_user_credential( + server_id="srv-byok-self", + user_api_key_dict=_make_user_auth("user-self"), + user_id="user-self", + ) + + assert delete_mock.await_args.args[1:] == ("user-self", "srv-byok-self") + + +@pytest.mark.asyncio +async def test_admin_revokes_another_users_oauth_credential(): + """A proxy admin naming user_id reads, deletes, and cache-invalidates that user's OAuth token.""" + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy._experimental.mcp_server import mcp_server_manager as manager_module + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_oauth_user_credential, + ) + + get_mock = AsyncMock(return_value={"type": "oauth2", "access_token": "mallory-tok"}) + delete_mock = AsyncMock(return_value=None) + invalidate_mock = AsyncMock(return_value=None) + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the stored OAuth token read + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_user_oauth_credential", + new=get_mock, + ), + patch( # test-quality-ok: endpoint test stubs the credential row delete + "litellm.proxy.management_endpoints.mcp_management_endpoints.delete_user_credential", + new=delete_mock, + ), + patch.object( # test-quality-ok: the OAuth cache lives on the global manager; the suite's only seam + manager_module.global_mcp_server_manager, + "invalidate_user_oauth_token_cache", + new=invalidate_mock, + ), + ): + result = await delete_mcp_oauth_user_credential( + server_id="srv-oauth-admin", + user_api_key_dict=_make_admin_auth(), + user_id="mallory", + ) + + assert get_mock.await_args.args[1:] == ("mallory", "srv-oauth-admin") + assert delete_mock.await_args.args[1:] == ("mallory", "srv-oauth-admin") + invalidate_mock.assert_awaited_once_with("mallory", "srv-oauth-admin") + assert result.has_credential is False + + +@pytest.mark.asyncio +@pytest.mark.parametrize("role", [LitellmUserRoles.INTERNAL_USER, LitellmUserRoles.PROXY_ADMIN_VIEW_ONLY]) +async def test_non_full_admin_cannot_revoke_another_users_oauth_credential(role): + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_oauth_user_credential, + ) + + get_mock = AsyncMock(return_value={"type": "oauth2", "access_token": "mallory-tok"}) + delete_mock = AsyncMock(return_value=None) + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the stored OAuth token read + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_user_oauth_credential", + new=get_mock, + ), + patch( # test-quality-ok: endpoint test stubs the credential row delete + "litellm.proxy.management_endpoints.mcp_management_endpoints.delete_user_credential", + new=delete_mock, + ), + ): + with pytest.raises(HTTPException) as exc_info: + await delete_mcp_oauth_user_credential( + server_id="srv-oauth-forbidden", + user_api_key_dict=_make_admin_auth(role), + user_id="mallory", + ) + + assert exc_info.value.status_code == 403 + get_mock.assert_not_awaited() + delete_mock.assert_not_awaited() + + +@pytest.mark.asyncio +@pytest.mark.parametrize("role", [LitellmUserRoles.PROXY_ADMIN, LitellmUserRoles.PROXY_ADMIN_VIEW_ONLY]) +async def test_admin_lists_every_users_credential_for_a_server(role): + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy._types import MCPServerUserCredentialListItem + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + list_mcp_server_user_credentials, + ) + + items = ( + MCPServerUserCredentialListItem(user_id="alice", credential_type="byok", updated_at="2026-01-01T00:00:00"), + MCPServerUserCredentialListItem(user_id="bob", credential_type="oauth2", updated_at="2026-01-02T00:00:00"), + ) + list_mock = AsyncMock(return_value=items) + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the credential row listing + "litellm.proxy.management_endpoints.mcp_management_endpoints.list_server_user_credentials", + new=list_mock, + ), + ): + result = await list_mcp_server_user_credentials( + server_id="srv-list-admin", + user_api_key_dict=_make_admin_auth(role), + ) + + assert list_mock.await_args.args[1:] == ("srv-list-admin",) + assert [(item.user_id, item.credential_type) for item in result] == [("alice", "byok"), ("bob", "oauth2")] + + +@pytest.mark.asyncio +async def test_non_admin_cannot_list_a_servers_user_credentials(): + if not mgmt_endpoints.MCP_AVAILABLE: + pytest.skip("MCP module not installed") + + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + list_mcp_server_user_credentials, + ) + + list_mock = AsyncMock(return_value=()) + with ( + patch( # test-quality-ok: endpoint test stubs the Prisma client lookup + "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", + return_value=_make_prisma_client(), + ), + patch( # test-quality-ok: endpoint test stubs the credential row listing + "litellm.proxy.management_endpoints.mcp_management_endpoints.list_server_user_credentials", + new=list_mock, + ), + ): + with pytest.raises(HTTPException) as exc_info: + await list_mcp_server_user_credentials( + server_id="srv-list-forbidden", + user_api_key_dict=_make_user_auth("user-plain"), + ) + + assert exc_info.value.status_code == 403 + list_mock.assert_not_awaited() + + @pytest.mark.asyncio async def test_list_mcp_user_credentials_batch_server_fetch(): """list_mcp_user_credentials uses a single batch DB call, not N+1 queries.""" @@ -7321,3 +7576,146 @@ class TestGetMCPGatewaySessions: assert [(group.label, group.count) for group in result.by_client] == [("cursor", 1)] assert [(group.label, group.count) for group in result.by_user] == [("alice", 1)] assert "sk-live-secret" not in result.model_dump_json() + + +class TestDeleteMCPGatewaySessions: + @pytest.fixture(autouse=True) + def _forget_admin_terminated_ids(self): + from litellm.proxy._experimental.mcp_server import server as mcp_server + + yield + mcp_server._admin_terminated_session_ids.clear() + + @pytest.mark.asyncio + @pytest.mark.parametrize("role", [LitellmUserRoles.INTERNAL_USER, LitellmUserRoles.PROXY_ADMIN_VIEW_ONLY]) + async def test_non_full_admin_forbidden_before_any_session_is_touched(self, role): + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_gateway_sessions, + ) + + session_id = "gateway-terminate-forbidden-1" + transport = MagicMock(terminate=AsyncMock()) + auth_user = mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key="sk-live", user_id="alice"), + ) + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + mcp_server.session_manager_stateful, "_server_instances", {session_id: transport} + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, {session_id: auth_user}, clear=True + ), + ): + with pytest.raises(HTTPException) as exc_info: + await delete_mcp_gateway_sessions( + user_api_key_dict=generate_mock_user_api_key_auth(user_role=role), + session_id_prefix=session_id, + user_id=None, + ) + assert exc_info.value.status_code == 403 + transport.terminate.assert_not_awaited() + assert session_id in mcp_server._stateful_session_auth_contexts + + @pytest.mark.asyncio + async def test_requires_a_selector(self): + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_gateway_sessions, + ) + + with pytest.raises(HTTPException) as exc_info: + await delete_mcp_gateway_sessions( + user_api_key_dict=generate_mock_user_api_key_auth(user_role=LitellmUserRoles.PROXY_ADMIN), + session_id_prefix=None, + user_id=None, + ) + assert exc_info.value.status_code == 400 + + @pytest.mark.asyncio + async def test_admin_terminates_only_the_selected_session(self): + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_gateway_sessions, + ) + from litellm.types.mcp import MCPGatewaySessionsTerminateResponse + + target_id = "11111111-target-session" + other_id = "22222222-other-session" + target_transport = MagicMock(terminate=AsyncMock()) + other_transport = MagicMock(terminate=AsyncMock()) + transports = {target_id: target_transport, other_id: other_transport} + contexts = { + target_id: mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key="sk-live-target", user_id="alice"), + ), + other_id: mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key="sk-live-other", user_id="bob"), + ), + } + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + mcp_server.session_manager_stateful, "_server_instances", transports + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, contexts, clear=True + ), + ): + result = await delete_mcp_gateway_sessions( + user_api_key_dict=generate_mock_user_api_key_auth(user_role=LitellmUserRoles.PROXY_ADMIN), + session_id_prefix=target_id[:8], + user_id=None, + ) + assert target_id not in transports + assert other_id in transports + assert target_id not in mcp_server._stateful_session_auth_contexts + assert other_id in mcp_server._stateful_session_auth_contexts + + target_transport.terminate.assert_awaited_once() + other_transport.terminate.assert_not_awaited() + assert isinstance(result, MCPGatewaySessionsTerminateResponse) + assert result.terminated_sessions == 1 + assert [(s.session_id_prefix, s.user_id) for s in result.sessions] == [(target_id[:8], "alice")] + assert target_id not in result.model_dump_json() + assert "sk-live-target" not in result.model_dump_json() + + @pytest.mark.asyncio + async def test_admin_terminates_every_session_of_the_selected_user(self): + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + delete_mcp_gateway_sessions, + ) + + def auth_user(user_id: str): + return mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key=f"sk-live-{user_id}", user_id=user_id), + ) + + transports = { + "bob-session-1": MagicMock(terminate=AsyncMock()), + "bob-session-2": MagicMock(terminate=AsyncMock()), + "alice-session-1": MagicMock(terminate=AsyncMock()), + } + contexts = { + "bob-session-1": auth_user("bob"), + "bob-session-2": auth_user("bob"), + "alice-session-1": auth_user("alice"), + } + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + mcp_server.session_manager_stateful, "_server_instances", transports + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, contexts, clear=True + ), + ): + result = await delete_mcp_gateway_sessions( + user_api_key_dict=generate_mock_user_api_key_auth(user_role=LitellmUserRoles.PROXY_ADMIN), + session_id_prefix=None, + user_id="bob", + ) + assert set(transports) == {"alice-session-1"} + assert set(mcp_server._stateful_session_auth_contexts) == {"alice-session-1"} + + assert result.terminated_sessions == 2 + assert {s.user_id for s in result.sessions} == {"bob"} + assert "sk-live-bob" not in result.model_dump_json() diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.integration.test.tsx index 11328ff1a3d..f1ddf709038 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.integration.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.integration.test.tsx @@ -1,13 +1,15 @@ import React from "react"; import { render, screen, within } from "@testing-library/react"; +import userEvent from "@testing-library/user-event"; import { describe, it, expect, vi, beforeEach } from "vitest"; import { QueryClient, QueryClientProvider } from "@tanstack/react-query"; -import { MCPGatewaySessionsTab, formatIdleSeconds } from "./MCPGatewaySessionsTab"; +import { MCPGatewaySessionsTab, describeTerminateResult, formatIdleSeconds } from "./MCPGatewaySessionsTab"; import * as networking from "@/components/networking"; -import type { MCPGatewaySessionsResponse } from "@/components/mcp_tools/types"; +import type { MCPGatewaySessionsResponse, MCPGatewaySessionsTerminateResponse } from "@/components/mcp_tools/types"; vi.mock("@/components/networking", () => ({ fetchMCPGatewaySessions: vi.fn(), + terminateMCPGatewaySessions: vi.fn(), })); const REPORT: MCPGatewaySessionsResponse = { @@ -64,11 +66,11 @@ const REPORT: MCPGatewaySessionsResponse = { ], }; -const renderTab = () => { +const renderTab = ({ canTerminate = false }: { canTerminate?: boolean } = {}) => { const queryClient = new QueryClient({ defaultOptions: { queries: { retry: false, gcTime: 0 } } }); return render( - + , ); }; @@ -83,6 +85,17 @@ describe("formatIdleSeconds", () => { }); }); +describe("describeTerminateResult", () => { + it("pluralizes the session count and names the worker", () => { + expect(describeTerminateResult({ worker_pid: 9, terminated_sessions: 1, sessions: [] })).toBe( + "Disconnected 1 session on worker pid 9.", + ); + expect(describeTerminateResult({ worker_pid: 9, terminated_sessions: 0, sessions: [] })).toBe( + "Disconnected 0 sessions on worker pid 9.", + ); + }); +}); + describe("MCPGatewaySessionsTab", () => { beforeEach(() => { vi.clearAllMocks(); @@ -135,4 +148,73 @@ describe("MCPGatewaySessionsTab", () => { expect(alert).toHaveTextContent("Could not load live connections"); expect(alert).toHaveTextContent("Admin access required"); }); + + it("hides every disconnect control from a read-only admin", async () => { + vi.mocked(networking.fetchMCPGatewaySessions).mockResolvedValue(REPORT); + renderTab({ canTerminate: false }); + + await screen.findByRole("region", { name: "Live sessions" }); + expect(screen.queryByRole("button", { name: /^Disconnect/ })).not.toBeInTheDocument(); + }); + + it("disconnects one session by its displayed prefix after confirmation and refetches", async () => { + const user = userEvent.setup(); + const terminated: MCPGatewaySessionsTerminateResponse = { + worker_pid: 4242, + terminated_sessions: 1, + sessions: [REPORT.sessions[1]], + }; + vi.mocked(networking.fetchMCPGatewaySessions).mockResolvedValue(REPORT); + vi.mocked(networking.terminateMCPGatewaySessions).mockResolvedValue(terminated); + renderTab({ canTerminate: true }); + + await user.click(await screen.findByRole("button", { name: "Disconnect session bbbb2222" })); + expect(networking.terminateMCPGatewaySessions).not.toHaveBeenCalled(); + const dialog = await screen.findByRole("alertdialog"); + expect(dialog).toHaveTextContent("session bbbb2222"); + await user.click(within(dialog).getByRole("button", { name: "Disconnect" })); + + const status = await screen.findByText("Disconnected 1 session on worker pid 4242.", { exact: false }); + expect(status).toBeInTheDocument(); + expect(networking.terminateMCPGatewaySessions).toHaveBeenCalledWith("token", { session_id_prefix: "bbbb2222" }); + expect(networking.fetchMCPGatewaySessions).toHaveBeenCalledTimes(2); + }); + + it("disconnects every session of a user from the by-user table", async () => { + const user = userEvent.setup(); + vi.mocked(networking.fetchMCPGatewaySessions).mockResolvedValue(REPORT); + vi.mocked(networking.terminateMCPGatewaySessions).mockResolvedValue({ + worker_pid: 4242, + terminated_sessions: 2, + sessions: [REPORT.sessions[0], REPORT.sessions[1]], + }); + renderTab({ canTerminate: true }); + + const byUser = await screen.findByRole("region", { name: "Sessions by user" }); + expect(within(byUser).queryByRole("button", { name: /\(unknown\)/ })).not.toBeInTheDocument(); + await user.click(within(byUser).getByRole("button", { name: "Disconnect all sessions for user alice" })); + const dialog = await screen.findByRole("alertdialog"); + expect(dialog).toHaveTextContent("every live session opened by user alice"); + await user.click(within(dialog).getByRole("button", { name: "Disconnect" })); + + expect(await screen.findByText(/Disconnected 2 sessions on worker pid 4242\./)).toBeInTheDocument(); + expect(networking.terminateMCPGatewaySessions).toHaveBeenCalledWith("token", { user_id: "alice" }); + }); + + it("shows the API error when a disconnect is refused", async () => { + const user = userEvent.setup(); + vi.mocked(networking.fetchMCPGatewaySessions).mockResolvedValue(REPORT); + vi.mocked(networking.terminateMCPGatewaySessions).mockRejectedValue( + new Error("Proxy admin access required to terminate MCP gateway sessions."), + ); + renderTab({ canTerminate: true }); + + await user.click(await screen.findByRole("button", { name: "Disconnect session aaaa1111" })); + await user.click(within(await screen.findByRole("alertdialog")).getByRole("button", { name: "Disconnect" })); + + const alert = await screen.findByRole("alert"); + expect(alert).toHaveTextContent("Could not disconnect"); + expect(alert).toHaveTextContent("Proxy admin access required to terminate MCP gateway sessions."); + expect(screen.getByRole("region", { name: "Live sessions" })).toBeInTheDocument(); + }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.tsx index 18f44d5d090..04a095f8792 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPGatewaySessionsTab.tsx @@ -1,14 +1,27 @@ "use client"; -import React from "react"; -import { useQuery } from "@tanstack/react-query"; -import { RefreshCw } from "lucide-react"; +import React, { useState } from "react"; +import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query"; +import { RefreshCw, Unplug } from "lucide-react"; import { Alert, AlertDescription, AlertTitle } from "@/components/ui/alert"; +import { + AlertDialog, + AlertDialogContent, + AlertDialogDescription, + AlertDialogFooter, + AlertDialogHeader, + AlertDialogTitle, +} from "@/components/ui/alert-dialog"; import { Button } from "@/components/ui/button"; import { Table, TableBody, TableCell, TableHead, TableHeader, TableRow } from "@/components/ui/table"; import { UiLoadingSpinner } from "@/components/ui/ui-loading-spinner"; -import { fetchMCPGatewaySessions } from "@/components/networking"; -import type { MCPGatewaySessionGroupCount, MCPGatewaySessionsResponse } from "@/components/mcp_tools/types"; +import { fetchMCPGatewaySessions, terminateMCPGatewaySessions } from "@/components/networking"; +import type { + MCPGatewaySessionGroupCount, + MCPGatewaySessionSelector, + MCPGatewaySessionsResponse, + MCPGatewaySessionsTerminateResponse, +} from "@/components/mcp_tools/types"; import { createQueryKeys } from "@/app/(dashboard)/hooks/common/queryKeysFactory"; const mcpGatewaySessionKeys = createQueryKeys("mcpGatewaySessions"); @@ -28,6 +41,16 @@ function groupLabel(label: string | null): string { return label === "" ? '""' : label; } +export function describeSelector(selector: MCPGatewaySessionSelector): string { + if (selector.user_id !== undefined) return `every live session opened by user ${groupLabel(selector.user_id)}`; + return `session ${selector.session_id_prefix}`; +} + +export function describeTerminateResult(result: MCPGatewaySessionsTerminateResponse): string { + const noun = result.terminated_sessions === 1 ? "session" : "sessions"; + return `Disconnected ${result.terminated_sessions} ${noun} on worker pid ${result.worker_pid}.`; +} + function StatCard({ label, value }: { label: string; value: number }) { return (
@@ -37,14 +60,37 @@ function StatCard({ label, value }: { label: string; value: number }) { ); } +function DisconnectUserButton({ + userId, + onDisconnectUser, +}: { + userId: string | null; + onDisconnectUser: (userId: string) => void; +}) { + if (userId === null || userId === "") return null; + return ( + + ); +} + function GroupCountTable({ title, groups, labelHeader, + onDisconnectUser, }: { title: string; groups: MCPGatewaySessionGroupCount[]; labelHeader: string; + onDisconnectUser?: (userId: string) => void; }) { return (
@@ -54,6 +100,7 @@ function GroupCountTable({ {labelHeader} Sessions + {onDisconnectUser ? Actions : null} @@ -61,6 +108,11 @@ function GroupCountTable({ {groupLabel(group.label)} {group.count} + {onDisconnectUser ? ( + + + + ) : null} ))} @@ -73,10 +125,12 @@ function SessionsBody({ data, error, isLoading, + onDisconnect, }: { data: MCPGatewaySessionsResponse | undefined; error: Error | null; isLoading: boolean; + onDisconnect: ((selector: MCPGatewaySessionSelector) => void) | null; }) { if (isLoading) { return ( @@ -117,7 +171,12 @@ function SessionsBody({
- + onDisconnect({ user_id: userId }) : undefined} + />

@@ -134,11 +193,12 @@ function SessionsBody({ Client IP Idle In flight + {onDisconnect ? Actions : null} - {data.sessions.map((session) => ( - + {data.sessions.map((session, index) => ( + {session.session_id_prefix} {session.client_name === null ? ( @@ -169,6 +229,19 @@ function SessionsBody({ {session.client_ip || "-"} {formatIdleSeconds(session.idle_seconds)} {session.in_flight_requests} + {onDisconnect ? ( + + + + ) : null} ))} @@ -180,9 +253,12 @@ function SessionsBody({ interface MCPGatewaySessionsTabProps { accessToken: string | null; + canTerminate: boolean; } -export function MCPGatewaySessionsTab({ accessToken }: MCPGatewaySessionsTabProps) { +export function MCPGatewaySessionsTab({ accessToken, canTerminate }: MCPGatewaySessionsTabProps) { + const queryClient = useQueryClient(); + const [pendingSelector, setPendingSelector] = useState(null); const queryOptions = { queryKey: mcpGatewaySessionKeys.lists(), queryFn: () => fetchMCPGatewaySessions(accessToken!), @@ -190,6 +266,15 @@ export function MCPGatewaySessionsTab({ accessToken }: MCPGatewaySessionsTabProp refetchInterval: REFETCH_INTERVAL_MS, }; const { data, error, isLoading, isFetching, refetch } = useQuery(queryOptions); + const terminate = useMutation({ + mutationFn: (selector) => terminateMCPGatewaySessions(accessToken!, selector), + onSettled: () => queryClient.invalidateQueries({ queryKey: mcpGatewaySessionKeys.lists() }), + }); + const confirmDisconnect = () => { + if (pendingSelector === null) return; + terminate.mutate(pendingSelector); + setPendingSelector(null); + }; return (
@@ -214,7 +299,48 @@ export function MCPGatewaySessionsTab({ accessToken }: MCPGatewaySessionsTabProp
- + {terminate.isError ? ( + + Could not disconnect + {terminate.error.message} + + ) : null} + {terminate.isSuccess ? ( + + Disconnected + + {describeTerminateResult(terminate.data)} Clients holding those sessions must send a new initialize request, + which re-runs authentication. Sessions on other proxy workers are not affected. + + + ) : null} + + + + !open && setPendingSelector(null)}> + + + Disconnect MCP session + + {pendingSelector ? `This force-closes ${describeSelector(pendingSelector)} on this proxy worker. ` : ""} + In-flight requests fail and the client must initialize again before it can call tools. + + + + + + + + ); } diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.integration.test.tsx new file mode 100644 index 00000000000..a20dd032d33 --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.integration.test.tsx @@ -0,0 +1,102 @@ +import React from "react"; +import { render, screen, within } from "@testing-library/react"; +import userEvent from "@testing-library/user-event"; +import { describe, it, expect, vi, beforeEach } from "vitest"; +import { QueryClient, QueryClientProvider } from "@tanstack/react-query"; +import { MCPServerUserCredentialsPanel } from "./MCPServerUserCredentialsPanel"; +import * as networking from "@/components/networking"; +import type { MCPServerUserCredentialListItem } from "@/components/mcp_tools/types"; + +vi.mock("@/components/networking", () => ({ + fetchMCPServerUserCredentials: vi.fn(), + revokeMCPServerUserCredential: vi.fn(), +})); + +const ITEMS: MCPServerUserCredentialListItem[] = [ + { + user_id: "alice", + credential_type: "oauth2", + expires_at: "2026-12-31T00:00:00+00:00", + connected_at: "2026-01-01T00:00:00+00:00", + updated_at: "2026-01-01T00:00:00+00:00", + }, + { + user_id: "carol", + credential_type: "byok", + expires_at: null, + connected_at: null, + updated_at: "2026-02-01T00:00:00+00:00", + }, +]; + +const renderPanel = ({ canRevoke = false }: { canRevoke?: boolean } = {}) => { + const queryClient = new QueryClient({ defaultOptions: { queries: { retry: false, gcTime: 0 } } }); + return render( + + + , + ); +}; + +describe("MCPServerUserCredentialsPanel", () => { + beforeEach(() => { + vi.clearAllMocks(); + }); + + it("lists each user's credential type without a revoke control for a read-only admin", async () => { + vi.mocked(networking.fetchMCPServerUserCredentials).mockResolvedValue(ITEMS); + renderPanel({ canRevoke: false }); + + const table = await screen.findByRole("region", { name: "Stored user credentials" }); + expect(within(table).getByRole("row", { name: /alice/ })).toHaveTextContent("OAuth2"); + expect(within(table).getByRole("row", { name: /carol/ })).toHaveTextContent("BYOK API key"); + expect(screen.queryByRole("button", { name: /^Revoke credential/ })).not.toBeInTheDocument(); + expect(networking.fetchMCPServerUserCredentials).toHaveBeenCalledWith("token", "srv-1"); + }); + + it("revokes the selected user's credential through the route for its type and refetches", async () => { + const user = userEvent.setup(); + vi.mocked(networking.fetchMCPServerUserCredentials).mockResolvedValueOnce(ITEMS).mockResolvedValueOnce([ITEMS[1]]); + vi.mocked(networking.revokeMCPServerUserCredential).mockResolvedValue(undefined); + renderPanel({ canRevoke: true }); + + await user.click(await screen.findByRole("button", { name: "Revoke credential for user alice" })); + expect(networking.revokeMCPServerUserCredential).not.toHaveBeenCalled(); + const dialog = await screen.findByRole("alertdialog"); + expect(dialog).toHaveTextContent("OAuth2 credential stored for user alice"); + await user.click(within(dialog).getByRole("button", { name: "Revoke" })); + + expect(await screen.findByText(/OAuth2 credential for user alice was deleted/)).toBeInTheDocument(); + expect(networking.revokeMCPServerUserCredential).toHaveBeenCalledWith("token", "srv-1", "alice", "oauth2"); + const table = await screen.findByRole("region", { name: "Stored user credentials" }); + expect(within(table).queryByRole("row", { name: /alice/ })).not.toBeInTheDocument(); + expect(within(table).getByRole("row", { name: /carol/ })).toBeInTheDocument(); + }); + + it("shows the API error when a revoke is refused and keeps the list", async () => { + const user = userEvent.setup(); + vi.mocked(networking.fetchMCPServerUserCredentials).mockResolvedValue(ITEMS); + vi.mocked(networking.revokeMCPServerUserCredential).mockRejectedValue( + new Error("Proxy admin access required to revoke another user's MCP credential."), + ); + renderPanel({ canRevoke: true }); + + await user.click(await screen.findByRole("button", { name: "Revoke credential for user carol" })); + await user.click(within(await screen.findByRole("alertdialog")).getByRole("button", { name: "Revoke" })); + + const alert = await screen.findByRole("alert"); + expect(alert).toHaveTextContent("Could not revoke credential"); + expect(alert).toHaveTextContent("Proxy admin access required to revoke another user's MCP credential."); + expect(networking.revokeMCPServerUserCredential).toHaveBeenCalledWith("token", "srv-1", "carol", "byok"); + expect(screen.getByRole("region", { name: "Stored user credentials" })).toBeInTheDocument(); + }); + + it("shows the API error when the list cannot be loaded", async () => { + vi.mocked(networking.fetchMCPServerUserCredentials).mockRejectedValue(new Error("Admin access required")); + renderPanel(); + + const alert = await screen.findByRole("alert"); + expect(alert).toHaveTextContent("Could not load user credentials"); + expect(alert).toHaveTextContent("Admin access required"); + }); +}); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.tsx new file mode 100644 index 00000000000..20b679450c3 --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/MCPServerUserCredentialsPanel.tsx @@ -0,0 +1,212 @@ +"use client"; + +import React, { useState } from "react"; +import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query"; +import { RefreshCw, ShieldOff } from "lucide-react"; +import { Alert, AlertDescription, AlertTitle } from "@/components/ui/alert"; +import { + AlertDialog, + AlertDialogContent, + AlertDialogDescription, + AlertDialogFooter, + AlertDialogHeader, + AlertDialogTitle, +} from "@/components/ui/alert-dialog"; +import { Badge } from "@/components/ui/badge"; +import { Button } from "@/components/ui/button"; +import { Table, TableBody, TableCell, TableHead, TableHeader, TableRow } from "@/components/ui/table"; +import { UiLoadingSpinner } from "@/components/ui/ui-loading-spinner"; +import { fetchMCPServerUserCredentials, revokeMCPServerUserCredential } from "@/components/networking"; +import type { MCPServerUserCredentialListItem } from "@/components/mcp_tools/types"; +import { createQueryKeys } from "@/app/(dashboard)/hooks/common/queryKeysFactory"; + +const mcpServerUserCredentialKeys = createQueryKeys("mcpServerUserCredentials"); + +export function credentialTypeLabel(credentialType: MCPServerUserCredentialListItem["credential_type"]): string { + return credentialType === "oauth2" ? "OAuth2" : "BYOK API key"; +} + +export function formatTimestamp(value: string | null): string { + if (value === null) return "-"; + const parsed = new Date(value); + return Number.isNaN(parsed.getTime()) ? value : parsed.toLocaleString(); +} + +function CredentialsBody({ + items, + error, + isLoading, + onRevoke, +}: { + items: MCPServerUserCredentialListItem[] | undefined; + error: Error | null; + isLoading: boolean; + onRevoke: ((item: MCPServerUserCredentialListItem) => void) | null; +}) { + if (isLoading) { + return ( +
+ +

Loading user credentials...

+
+ ); + } + if (error) { + return ( + + Could not load user credentials + {error.message} + + ); + } + if (!items) return null; + if (items.length === 0) { + return ( +
+

No user has a stored credential for this server.

+
+ ); + } + return ( +
+ + + + User + Type + Connected + Expires + Updated + {onRevoke ? Actions : null} + + + + {items.map((item) => ( + + {item.user_id} + + {credentialTypeLabel(item.credential_type)} + + {formatTimestamp(item.connected_at)} + {formatTimestamp(item.expires_at)} + {formatTimestamp(item.updated_at)} + {onRevoke ? ( + + + + ) : null} + + ))} + +
+
+ ); +} + +interface MCPServerUserCredentialsPanelProps { + serverId: string; + accessToken: string | null; + canRevoke: boolean; +} + +export function MCPServerUserCredentialsPanel({ + serverId, + accessToken, + canRevoke, +}: MCPServerUserCredentialsPanelProps) { + const queryClient = useQueryClient(); + const [pendingItem, setPendingItem] = useState(null); + const queryKey = mcpServerUserCredentialKeys.detail(serverId); + const { data, error, isLoading, isFetching, refetch } = useQuery({ + queryKey, + queryFn: () => fetchMCPServerUserCredentials(accessToken!, serverId), + enabled: !!accessToken, + }); + const revoke = useMutation({ + mutationFn: (item) => revokeMCPServerUserCredential(accessToken!, serverId, item.user_id, item.credential_type), + onSettled: () => queryClient.invalidateQueries({ queryKey }), + }); + const confirmRevoke = () => { + if (pendingItem === null) return; + revoke.mutate(pendingItem); + setPendingItem(null); + }; + + return ( +
+
+
+

User Credentials

+

+ Per-user OAuth2 tokens and BYOK API keys stored for this server. Revoking one deletes it from the database + and clears the cached copy, so the user must connect again before the gateway will call this server for + them. +

+
+ +
+ + {revoke.isError ? ( + + Could not revoke credential + {revoke.error.message} + + ) : null} + {revoke.isSuccess ? ( + + Credential revoked + + The stored {credentialTypeLabel(revoke.variables.credential_type)} credential for user{" "} + {revoke.variables.user_id} was deleted. + + + ) : null} + + + + !open && setPendingItem(null)}> + + + Revoke stored credential + + {pendingItem + ? `This deletes the ${credentialTypeLabel(pendingItem.credential_type)} credential stored for user ${pendingItem.user_id}. ` + : ""} + Their next MCP request to this server fails until they connect again. + + + + + + + + +
+ ); +} + +export default MCPServerUserCredentialsPanel; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx index 475392620d4..c23b48ef672 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx @@ -9,7 +9,9 @@ import { MCPServer, handleTransport, handleAuth } from "@/components/mcp_tools/t // TODO: Move Tools viewer from index file import { MCPToolsViewer } from "."; import MCPServerEdit, { EDIT_OAUTH_UI_STATE_KEY } from "./mcp_server_edit"; +import { MCPServerUserCredentialsPanel } from "./MCPServerUserCredentialsPanel"; import { getSecureItem } from "@/utils/secureStorage"; +import { isProxyAdminRole, isProxyAdminTierRole } from "@/utils/roles"; import MCPServerCostDisplay from "./mcp_server_cost_display"; import { getMaskedAndFullUrl } from "./utils"; import { copyToClipboard as utilCopyToClipboard } from "@/utils/dataUtils"; @@ -63,6 +65,8 @@ export const MCPServerView: React.FC = ({ const [showFullUrl, setShowFullUrl] = useState(false); const [copiedStates, setCopiedStates] = useState>({}); const [selectedTabIndex, setSelectedTabIndex] = useState(returningFromEditOAuth ? 2 : initialTabIndex); + const canViewUserCredentials = userRole !== null && isProxyAdminTierRole(userRole); + const canRevokeUserCredentials = userRole !== null && isProxyAdminRole(userRole); const handleSuccess = (updated: MCPServer) => { setEditing(false); @@ -142,6 +146,11 @@ export const MCPServerView: React.FC = ({ Settings )} + {canViewUserCredentials && ( + + User Credentials + + )} {/* Overview Panel */} @@ -387,6 +396,18 @@ export const MCPServerView: React.FC = ({ )} + + {canViewUserCredentials && ( + + + + + + )} ); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx index 00d79022103..3a197786774 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx @@ -1,4 +1,4 @@ -import { isAdminRole, isProxyAdminTierRole } from "@/utils/roles"; +import { isAdminRole, isProxyAdminRole, isProxyAdminTierRole } from "@/utils/roles"; import { CircleHelp, Search } from "lucide-react"; import { Badge } from "@/components/ui/badge"; import { Button } from "@/components/ui/button"; @@ -755,7 +755,7 @@ const MCPServers: React.FC = ({ accessToken, userRole, userID }) )} {isProxyAdminTierRole(userRole) && ( - + )} diff --git a/ui/litellm-dashboard/src/components/mcp_tools/types.tsx b/ui/litellm-dashboard/src/components/mcp_tools/types.tsx index fe04eb4969b..2f6a3f1d0d9 100644 --- a/ui/litellm-dashboard/src/components/mcp_tools/types.tsx +++ b/ui/litellm-dashboard/src/components/mcp_tools/types.tsx @@ -587,3 +587,23 @@ export interface MCPGatewaySessionsResponse { by_user: MCPGatewaySessionGroupCount[]; sessions: MCPGatewaySession[]; } + +export interface MCPGatewaySessionsTerminateResponse { + worker_pid: number; + terminated_sessions: number; + sessions: MCPGatewaySession[]; +} + +export type MCPGatewaySessionSelector = + | { session_id_prefix: string; user_id?: undefined } + | { user_id: string; session_id_prefix?: undefined }; + +export type MCPServerUserCredentialType = "oauth2" | "byok"; + +export interface MCPServerUserCredentialListItem { + user_id: string; + credential_type: MCPServerUserCredentialType; + expires_at: string | null; + connected_at: string | null; + updated_at: string; +} diff --git a/ui/litellm-dashboard/src/components/networking.tsx b/ui/litellm-dashboard/src/components/networking.tsx index a8ea1b66488..9381e2c07c9 100644 --- a/ui/litellm-dashboard/src/components/networking.tsx +++ b/ui/litellm-dashboard/src/components/networking.tsx @@ -97,7 +97,14 @@ import type { ModelBudgetUsage, ModelMaxBudget } from "./key_team_helpers/ModelM import type { ObjectPermission } from "./object_permission_types"; import type { components } from "@/lib/http/schema"; import { jsonFields } from "./common_components/check_openapi_schema"; -import type { MCPGatewaySessionsResponse, MCPUserEnvVarsStatus } from "./mcp_tools/types"; +import type { + MCPGatewaySessionSelector, + MCPGatewaySessionsResponse, + MCPGatewaySessionsTerminateResponse, + MCPServerUserCredentialListItem, + MCPServerUserCredentialType, + MCPUserEnvVarsStatus, +} from "./mcp_tools/types"; import type { CoordinationRedisSettings, CoordinationRedisSettingsResponse, @@ -5112,6 +5119,33 @@ export const fetchMCPSubmissions = async (accessToken: string) => { export const fetchMCPGatewaySessions = async (accessToken: string): Promise => apiClient.get(`/v1/mcp/sessions`, { accessToken }); +export const terminateMCPGatewaySessions = async ( + accessToken: string, + selector: MCPGatewaySessionSelector, +): Promise => + apiClient.delete(`/v1/mcp/sessions`, { accessToken, query: { ...selector } }); + +export const fetchMCPServerUserCredentials = async ( + accessToken: string, + serverId: string, +): Promise => + apiClient.get(`/v1/mcp/server/${encodeURIComponent(serverId)}/user-credentials`, { + accessToken, + }); + +export const revokeMCPServerUserCredential = async ( + accessToken: string, + serverId: string, + userId: string, + credentialType: MCPServerUserCredentialType, +): Promise => { + const route = credentialType === "oauth2" ? "oauth-user-credential" : "user-credential"; + await apiClient.delete(`/v1/mcp/server/${encodeURIComponent(serverId)}/${route}`, { + accessToken, + query: { user_id: userId }, + }); +}; + export const approveMCPServer = async (accessToken: string, serverId: string) => { try { const url = (proxyBaseUrl ? `${proxyBaseUrl}` : "") + `/v1/mcp/server/${encodeURIComponent(serverId)}/approve`; diff --git a/ui/litellm-dashboard/src/lib/http/schema.d.ts b/ui/litellm-dashboard/src/lib/http/schema.d.ts index f1779c7cb5a..43e1e16125c 100644 --- a/ui/litellm-dashboard/src/lib/http/schema.d.ts +++ b/ui/litellm-dashboard/src/lib/http/schema.d.ts @@ -18902,7 +18902,7 @@ export interface paths { post: operations["store_mcp_oauth_user_credential_v1_mcp_server__server_id__oauth_user_credential_post"]; /** * Delete Mcp Oauth User Credential - * @description Revoke the calling user's stored OAuth2 token for an MCP server + * @description Revoke the calling user's stored OAuth2 token for an MCP server. A proxy admin may pass user_id to revoke another user's stored token. */ delete: operations["delete_mcp_oauth_user_credential_v1_mcp_server__server_id__oauth_user_credential_delete"]; options?: never; @@ -18966,7 +18966,7 @@ export interface paths { post: operations["store_mcp_user_credential_v1_mcp_server__server_id__user_credential_post"]; /** * Delete Mcp User Credential - * @description Delete the calling user's stored API key for a BYOK MCP server + * @description Delete the calling user's stored API key for a BYOK MCP server. A proxy admin may pass user_id to revoke another user's stored key. */ delete: operations["delete_mcp_user_credential_v1_mcp_server__server_id__user_credential_delete"]; options?: never; @@ -18974,6 +18974,26 @@ export interface paths { patch?: never; trace?: never; }; + "/v1/mcp/server/{server_id}/user-credentials": { + parameters: { + query?: never; + header?: never; + path?: never; + cookie?: never; + }; + /** + * List Mcp Server User Credentials + * @description List every user's stored BYOK or OAuth2 credential for an MCP server (admin only, no secrets) + */ + get: operations["list_mcp_server_user_credentials_v1_mcp_server__server_id__user_credentials_get"]; + put?: never; + post?: never; + delete?: never; + options?: never; + head?: never; + patch?: never; + trace?: never; + }; "/v1/mcp/server/{server_id}/user-env-vars": { parameters: { query?: never; @@ -19016,7 +19036,11 @@ export interface paths { get: operations["get_mcp_gateway_sessions_v1_mcp_sessions_get"]; put?: never; post?: never; - delete?: never; + /** + * Delete Mcp Gateway Sessions + * @description Force-close live stateful MCP gateway sessions on this proxy worker, selected by session id prefix and/or by the LiteLLM user that opened them (proxy admin only). + */ + delete: operations["delete_mcp_gateway_sessions_v1_mcp_sessions_delete"]; options?: never; head?: never; patch?: never; @@ -32383,6 +32407,18 @@ export interface components { /** Worker Pid */ worker_pid: number; }; + /** + * MCPGatewaySessionsTerminateResponse + * @description Stateful sessions an administrator force-closed on this proxy worker. + */ + MCPGatewaySessionsTerminateResponse: { + /** Sessions */ + sessions?: components["schemas"]["MCPGatewaySession"][]; + /** Terminated Sessions */ + terminated_sessions: number; + /** Worker Pid */ + worker_pid: number; + }; /** * MCPOAuthUserCredentialRequest * @description Stores a user's OAuth2 token for an OpenAPI MCP server. @@ -32487,6 +32523,25 @@ export interface components { [key: string]: unknown; }; }; + /** + * MCPServerUserCredentialListItem + * @description One user's stored credential for an MCP server, as an admin sees it. Never carries the secret. + */ + MCPServerUserCredentialListItem: { + /** Connected At */ + connected_at?: string | null; + /** + * Credential Type + * @enum {string} + */ + credential_type: "oauth2" | "byok"; + /** Expires At */ + expires_at?: string | null; + /** Updated At */ + updated_at: string; + /** User Id */ + user_id: string; + }; /** MCPSubmissionsSummary */ MCPSubmissionsSummary: { /** Active */ @@ -65109,7 +65164,9 @@ export interface operations { }; delete_mcp_oauth_user_credential_v1_mcp_server__server_id__oauth_user_credential_delete: { parameters: { - query?: never; + query?: { + user_id?: string | null; + }; header?: never; path: { server_id: string; @@ -65241,7 +65298,9 @@ export interface operations { }; delete_mcp_user_credential_v1_mcp_server__server_id__user_credential_delete: { parameters: { - query?: never; + query?: { + user_id?: string | null; + }; header?: never; path: { server_id: string; @@ -65270,6 +65329,37 @@ export interface operations { }; }; }; + list_mcp_server_user_credentials_v1_mcp_server__server_id__user_credentials_get: { + parameters: { + query?: never; + header?: never; + path: { + server_id: string; + }; + cookie?: never; + }; + requestBody?: never; + responses: { + /** @description Successful Response */ + 200: { + headers: { + [name: string]: unknown; + }; + content: { + "application/json": components["schemas"]["MCPServerUserCredentialListItem"][]; + }; + }; + /** @description Validation Error */ + 422: { + headers: { + [name: string]: unknown; + }; + content: { + "application/json": components["schemas"]["HTTPValidationError"]; + }; + }; + }; + }; get_mcp_user_env_vars_v1_mcp_server__server_id__user_env_vars_get: { parameters: { query?: never; @@ -65387,6 +65477,38 @@ export interface operations { }; }; }; + delete_mcp_gateway_sessions_v1_mcp_sessions_delete: { + parameters: { + query?: { + session_id_prefix?: string | null; + user_id?: string | null; + }; + header?: never; + path?: never; + cookie?: never; + }; + requestBody?: never; + responses: { + /** @description Successful Response */ + 200: { + headers: { + [name: string]: unknown; + }; + content: { + "application/json": components["schemas"]["MCPGatewaySessionsTerminateResponse"]; + }; + }; + /** @description Validation Error */ + 422: { + headers: { + [name: string]: unknown; + }; + content: { + "application/json": components["schemas"]["HTTPValidationError"]; + }; + }; + }; + }; get_mcp_tools_v1_mcp_tools_get: { parameters: { query?: never; From aea33b4b507d139bc8b5a1d93aa055205602a67f Mon Sep 17 00:00:00 2001 From: yassin Date: Fri, 18 Sep 2026 01:37:12 +0000 Subject: [PATCH 2/6] fix(ui): hide MCP disconnect and revoke controls from view-only admin sessions The auth hook normalizes proxy_admin_viewer to Admin for page access, so the role check alone let a view-only admin see Disconnect and Revoke buttons that the backend refuses with 403. Thread isViewOnly from useAuthorized into the MCP servers page and gate both mutation controls on it Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../_components/mcp_server_view.test.tsx | 62 +++++++++++++++---- .../_components/mcp_server_view.tsx | 4 +- .../_components/mcp_servers.test.tsx | 48 ++++++++++++++ .../mcp-servers/_components/mcp_servers.tsx | 8 ++- .../src/app/(dashboard)/mcp-servers/page.tsx | 4 +- .../src/components/mcp_tools/types.tsx | 1 + 6 files changed, 110 insertions(+), 17 deletions(-) diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.test.tsx index 02d168bf7f4..da564f23de5 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.test.tsx @@ -1,7 +1,9 @@ -import { render, screen } from "@testing-library/react"; +import { render, screen, within } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { describe, it, expect, vi, beforeEach } from "vitest"; +import { QueryClient, QueryClientProvider } from "@tanstack/react-query"; import { MCPServerView } from "./mcp_server_view"; +import * as networking from "@/components/networking"; import type { MCPServer } from "@/components/mcp_tools/types"; vi.mock(".", () => ({ @@ -13,6 +15,12 @@ vi.mock("./mcp_server_edit", () => ({ EDIT_OAUTH_UI_STATE_KEY: "litellm-mcp-oauth-edit-state", })); +vi.mock("@/components/networking", async (importOriginal) => ({ + ...(await importOriginal()), + fetchMCPServerUserCredentials: vi.fn(), + revokeMCPServerUserCredential: vi.fn(), +})); + const baseServer = { server_id: "srv-1", server_name: "demo server", @@ -25,19 +33,38 @@ const baseServer = { const renderView = (overrides: Partial = {}, props: Record = {}) => render( - , + + + , ); +const openUserCredentials = async (props: Record) => { + vi.mocked(networking.fetchMCPServerUserCredentials).mockResolvedValue([ + { + user_id: "alice", + credential_type: "byok", + expires_at: null, + connected_at: null, + updated_at: "2026-01-01T00:00:00+00:00", + }, + ]); + renderView({}, props); + await userEvent.click(screen.getByRole("tab", { name: "User Credentials" })); + return within(await screen.findByRole("region", { name: "Stored user credentials" })).getByRole("row", { + name: /alice/, + }); +}; + describe("MCPServerView", () => { beforeEach(() => { vi.clearAllMocks(); @@ -149,4 +176,15 @@ describe("MCPServerView", () => { expect(await screen.findByText("All tools enabled")).toBeInTheDocument(); }); + + it("lets a full admin revoke a stored user credential", async () => { + const row = await openUserCredentials({}); + expect(within(row).getByRole("button", { name: "Revoke credential for user alice" })).toBeInTheDocument(); + }); + + it("shows stored credentials to a view-only admin session without a revoke control", async () => { + const row = await openUserCredentials({ isViewOnly: true }); + expect(row).toHaveTextContent("BYOK API key"); + expect(within(row).queryByRole("button", { name: /^Revoke credential/ })).not.toBeInTheDocument(); + }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx index c23b48ef672..a346c7d986b 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_server_view.tsx @@ -25,6 +25,7 @@ interface MCPServerViewProps { accessToken: string | null; userRole: string | null; userID: string | null; + isViewOnly?: boolean; availableAccessGroups: string[]; initialTabIndex?: number; } @@ -55,6 +56,7 @@ export const MCPServerView: React.FC = ({ accessToken, userRole, userID, + isViewOnly = false, availableAccessGroups, initialTabIndex = 0, }) => { @@ -66,7 +68,7 @@ export const MCPServerView: React.FC = ({ const [copiedStates, setCopiedStates] = useState>({}); const [selectedTabIndex, setSelectedTabIndex] = useState(returningFromEditOAuth ? 2 : initialTabIndex); const canViewUserCredentials = userRole !== null && isProxyAdminTierRole(userRole); - const canRevokeUserCredentials = userRole !== null && isProxyAdminRole(userRole); + const canRevokeUserCredentials = userRole !== null && isProxyAdminRole(userRole) && !isViewOnly; const handleSuccess = (updated: MCPServer) => { setEditing(false); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx index 8abb8855e3d..1394a923174 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx @@ -17,6 +17,8 @@ vi.mock("@/components/networking", () => ({ updateConfigFieldSetting: vi.fn().mockResolvedValue(undefined), deleteConfigFieldSetting: vi.fn().mockResolvedValue(undefined), listMCPUserEnvVarStatus: vi.fn().mockResolvedValue([]), + fetchMCPGatewaySessions: vi.fn(), + terminateMCPGatewaySessions: vi.fn(), })); const createQueryClient = () => @@ -400,4 +402,50 @@ describe("MCPServers", () => { // The server list refresh must NOT trigger a second health check expect(networking.fetchMCPServerHealth).toHaveBeenCalledTimes(1); }); + + const liveSessionsReport = { + worker_pid: 4242, + total_sessions: 1, + by_client: [{ label: "claude-code", count: 1 }], + by_user: [{ label: "alice", count: 1 }], + sessions: [ + { + session_id_prefix: "aaaa1111", + client_name: "claude-code", + client_version: "1.0.0", + user_id: "alice", + user_email: "alice@example.com", + key_alias: "alice-key", + team_id: null, + team_alias: null, + client_ip: "10.0.0.1", + idle_seconds: 5, + in_flight_requests: 0, + }, + ], + }; + + const openLiveConnections = async (props: { isViewOnly?: boolean }) => { + vi.mocked(networking.fetchMCPServers).mockResolvedValue([]); + vi.mocked(networking.fetchMCPGatewaySessions).mockResolvedValue(liveSessionsReport); + render( + + + , + ); + await userEvent.click(await screen.findByRole("tab", { name: "Live Connections" })); + return within(await screen.findByRole("region", { name: "Live sessions" })).getByRole("row", { name: /aaaa1111/ }); + }; + + it("lets a full admin disconnect a live session", async () => { + const row = await openLiveConnections({ isViewOnly: false }); + expect(within(row).getByRole("button", { name: "Disconnect session aaaa1111" })).toBeInTheDocument(); + }); + + it("shows live sessions to a view-only admin session without any disconnect control", async () => { + const row = await openLiveConnections({ isViewOnly: true }); + expect(row).toHaveTextContent("alice@example.com"); + expect(within(row).queryByRole("button", { name: /^Disconnect/ })).not.toBeInTheDocument(); + expect(screen.queryByRole("button", { name: /^Disconnect all/ })).not.toBeInTheDocument(); + }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx index 3a197786774..aa0a031c55c 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.tsx @@ -109,7 +109,7 @@ const readToolsOAuthServerId = (): string | null => { } }; -const MCPServers: React.FC = ({ accessToken, userRole, userID }) => { +const MCPServers: React.FC = ({ accessToken, userRole, userID, isViewOnly = false }) => { const { data: mcpServers, isLoading: isLoadingServers, refetch } = useMCPServers(); // Fetch health status for all servers @@ -578,6 +578,7 @@ const MCPServers: React.FC = ({ accessToken, userRole, userID }) accessToken={accessToken} userID={userID} userRole={userRole} + isViewOnly={isViewOnly} availableAccessGroups={uniqueMcpAccessGroups} initialTabIndex={selectedServerId === toolsTabServerId ? 1 : 0} /> @@ -755,7 +756,10 @@ const MCPServers: React.FC = ({ accessToken, userRole, userID }) )} {isProxyAdminTierRole(userRole) && ( - + )} diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/page.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/page.tsx index 462c48360cd..c297dcb8d33 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/page.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/page.tsx @@ -4,6 +4,6 @@ import { MCPServers } from "./_components"; import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; export default function McpServers() { - const { accessToken, userRole, userId } = useAuthorized(); - return ; + const { accessToken, userRole, userId, isViewOnly } = useAuthorized(); + return ; } diff --git a/ui/litellm-dashboard/src/components/mcp_tools/types.tsx b/ui/litellm-dashboard/src/components/mcp_tools/types.tsx index 2f6a3f1d0d9..f009429693b 100644 --- a/ui/litellm-dashboard/src/components/mcp_tools/types.tsx +++ b/ui/litellm-dashboard/src/components/mcp_tools/types.tsx @@ -517,6 +517,7 @@ export interface MCPServerProps { accessToken: string | null; userRole: string | null; userID: string | null; + isViewOnly?: boolean; } export interface MCPToolsetTool { From fd834f6f8b18c5b2a5d642627631f1d4bf566ab4 Mon Sep 17 00:00:00 2001 From: yassin Date: Fri, 18 Sep 2026 02:10:50 +0000 Subject: [PATCH 3/6] fix(mcp): broadcast BYOK and OAuth credential eviction to peer workers and expire admin session tombstones Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 3 +- .../mcp_server/byok_credential_cache.py | 38 ++++++ .../mcp_server/byok_oauth_endpoints.py | 2 +- .../mcp_server/oauth2_token_cache.py | 7 +- .../proxy/_experimental/mcp_server/server.py | 121 +++++++++--------- .../mcp_management_endpoints.py | 4 +- litellm/proxy/proxy_server.py | 3 +- .../mcp_server/test_byok_credential_cache.py | 57 +++++++++ .../mcp_server/test_byok_oauth_endpoints.py | 32 ++++- .../mcp_server/test_mcp_server.py | 70 +++++++++- .../mcp_server/test_oauth2_token_cache.py | 26 ++++ .../test_mcp_management_endpoints.py | 6 +- tests/test_litellm/proxy/test_proxy_server.py | 71 ++++++++++ 13 files changed, 366 insertions(+), 74 deletions(-) create mode 100644 litellm/proxy/_experimental/mcp_server/byok_credential_cache.py create mode 100644 tests/test_litellm/proxy/_experimental/mcp_server/test_byok_credential_cache.py diff --git a/litellm/constants.py b/litellm/constants.py index 2ebc9beb632..72fc6ed5160 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -184,7 +184,8 @@ MCP_METADATA_TIMEOUT: Final = float(os.getenv("LITELLM_MCP_METADATA_TIMEOUT", "1 MCP_HEALTH_CHECK_TIMEOUT: Final = float(os.getenv("LITELLM_MCP_HEALTH_CHECK_TIMEOUT", "10.0")) MCP_TOOL_LISTING_MAX_PAGES: Final = 1000 MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH: Final = 8 -MCP_ADMIN_TERMINATED_SESSION_IDS_MAX: Final = 1024 +MCP_BYOK_CREDENTIAL_CACHE_TTL_SECONDS: Final = 60 +MCP_BYOK_CREDENTIAL_CACHE_MAX_SIZE: Final = 4096 # Allowlist of commands permitted for MCP stdio transport. # Prevents arbitrary command execution via /mcp-rest/test/* endpoints or server creation. diff --git a/litellm/proxy/_experimental/mcp_server/byok_credential_cache.py b/litellm/proxy/_experimental/mcp_server/byok_credential_cache.py new file mode 100644 index 00000000000..3892015c405 --- /dev/null +++ b/litellm/proxy/_experimental/mcp_server/byok_credential_cache.py @@ -0,0 +1,38 @@ +"""Per-worker cache of stored BYOK credentials, keyed so peer workers can evict it over the auth cache pub/sub.""" + +from dataclasses import dataclass +from typing import Final + +from litellm.caching.in_memory_cache import InMemoryCache +from litellm.constants import MCP_BYOK_CREDENTIAL_CACHE_MAX_SIZE, MCP_BYOK_CREDENTIAL_CACHE_TTL_SECONDS + +_CACHE_KEY_PREFIX: Final = "mcp_byok_credential" + + +@dataclass(frozen=True, slots=True) +class CachedByokCredential: + credential: str | None + + +byok_credential_cache: Final = InMemoryCache( + max_size_in_memory=MCP_BYOK_CREDENTIAL_CACHE_MAX_SIZE, + default_ttl=MCP_BYOK_CREDENTIAL_CACHE_TTL_SECONDS, +) + + +def byok_credential_cache_key(user_id: str, server_id: str) -> str: + return f"{_CACHE_KEY_PREFIX}:{user_id}:{server_id}" + + +def get_cached_byok_credential(user_id: str, server_id: str) -> CachedByokCredential | None: + cached: Final = byok_credential_cache.get_cache( # pyright: ignore[reportUnknownMemberType, reportUnknownVariableType] # InMemoryCache is untyped + byok_credential_cache_key(user_id, server_id) + ) + return cached if isinstance(cached, CachedByokCredential) else None + + +def cache_byok_credential(user_id: str, server_id: str, credential: str | None) -> None: + byok_credential_cache.set_cache( # pyright: ignore[reportUnknownMemberType] # InMemoryCache is untyped + byok_credential_cache_key(user_id, server_id), + CachedByokCredential(credential=credential), + ) diff --git a/litellm/proxy/_experimental/mcp_server/byok_oauth_endpoints.py b/litellm/proxy/_experimental/mcp_server/byok_oauth_endpoints.py index 0ab76588b1f..2c63e0a96d8 100644 --- a/litellm/proxy/_experimental/mcp_server/byok_oauth_endpoints.py +++ b/litellm/proxy/_experimental/mcp_server/byok_oauth_endpoints.py @@ -865,7 +865,7 @@ async def byok_token( _invalidate_byok_cred_cache, ) - _invalidate_byok_cred_cache(user_id, server_id) + await _invalidate_byok_cred_cache(user_id, server_id) except Exception as exc: verbose_proxy_logger.error( "byok_token: failed to store user credential for user=%s server=%s: %s", diff --git a/litellm/proxy/_experimental/mcp_server/oauth2_token_cache.py b/litellm/proxy/_experimental/mcp_server/oauth2_token_cache.py index 42edc2999ab..3742d7b4ccc 100644 --- a/litellm/proxy/_experimental/mcp_server/oauth2_token_cache.py +++ b/litellm/proxy/_experimental/mcp_server/oauth2_token_cache.py @@ -295,12 +295,15 @@ class MCPPerUserTokenCache: ) async def delete(self, user_id: str, server_id: str) -> None: - """Invalidate the cached token (removes from both in-memory and Redis layers).""" + """Invalidate the cached token in Redis, here, and in every peer worker's in-memory layer.""" try: + from litellm.proxy.common_utils.auth_cache_invalidation_pubsub import ( # noqa: PLC0415 # proxy import cycle + evict_and_broadcast, + ) from litellm.proxy.proxy_server import user_api_key_cache # noqa: PLC0415 key: Final = self._cache_key(user_id, server_id) - await user_api_key_cache.async_delete_cache(key) + await evict_and_broadcast((key,), user_api_key_cache) except Exception as exc: verbose_logger.debug( "MCPPerUserTokenCache.delete failed for user=%s server=%s: %s", diff --git a/litellm/proxy/_experimental/mcp_server/server.py b/litellm/proxy/_experimental/mcp_server/server.py index 80d9274859d..a7ca2775fb1 100644 --- a/litellm/proxy/_experimental/mcp_server/server.py +++ b/litellm/proxy/_experimental/mcp_server/server.py @@ -14,7 +14,7 @@ import time import traceback import types import uuid -from collections import Counter, deque +from collections import Counter from collections.abc import AsyncIterator, Callable, Mapping, Sequence from datetime import datetime from typing import TYPE_CHECKING, Any, Final, NoReturn, Protocol @@ -30,7 +30,6 @@ from typing_extensions import ReadOnly, TypedDict from litellm._logging import verbose_logger from litellm.constants import ( MAXIMUM_TRACEBACK_LINES_TO_LOG, - MCP_ADMIN_TERMINATED_SESSION_IDS_MAX, MCP_GATEWAY_SESSION_ID_PREFIX_LENGTH, ) from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj @@ -42,6 +41,12 @@ from litellm.proxy._experimental.mcp_server.auth.user_api_key_auth_mcp import ( MCPRequestHandler, _is_mcp_admitted_user_subject, ) +from litellm.proxy._experimental.mcp_server.byok_credential_cache import ( + byok_credential_cache, + byok_credential_cache_key, + cache_byok_credential, + get_cached_byok_credential, +) from litellm.proxy._experimental.mcp_server.discoverable_endpoints import ( get_request_base_url, ) @@ -86,6 +91,9 @@ from litellm.proxy._types import ( UserAPIKeyAuth, ) from litellm.proxy.auth.ip_address_utils import IPAddressUtils +from litellm.proxy.common_utils.auth_cache_invalidation_pubsub import ( + publish_auth_cache_invalidation, +) from litellm.proxy.litellm_pre_call_utils import ( LiteLLMProxyRequestSetup, get_chain_id_from_headers, @@ -107,13 +115,6 @@ if TYPE_CHECKING: from litellm.proxy._experimental.mcp_server.db import OAuthCredentialPayload -# Short-lived in-memory cache for BYOK credentials. -# Keyed by (user_id, server_id); value is (credential_or_None, monotonic_timestamp). -# Storing the credential value (not just a bool) means _get_byok_credential and -# _check_byok_credential share a single DB round-trip per TTL window. -_byok_cred_cache: Final[dict[tuple[str, str], tuple[str | None, float]]] = {} -_BYOK_CRED_CACHE_TTL: Final = 60 # seconds -_BYOK_CRED_CACHE_MAX_SIZE: Final = 4096 # cap to prevent unbounded growth _STATEFUL_SESSION_IDLE_TIMEOUT_SECONDS: Final = 30 * 60 # Upper bound on concurrent stateful sessions a single caller may hold. Each # `initialize` creates a session that survives until the idle timeout, so @@ -132,20 +133,11 @@ _MCP_TRANSPORT_SPAN_SCOPE_KEY: Final = "litellm_otel_transport_span" _MCP_DESTINATIONS_SCOPE_KEY: Final = "litellm_otel_request_destinations" -def _invalidate_byok_cred_cache(user_id: str, server_id: str) -> None: - """Remove a (user_id, server_id) entry from the BYOK credential cache. - - Call this after storing or deleting a credential so subsequent calls - see the fresh value rather than a stale cached result. - """ - _byok_cred_cache.pop((user_id, server_id), None) - - -def _write_byok_cred_cache(user_id: str, server_id: str, credential: str | None) -> None: - """Write a credential value to the cache, evicting all entries if at capacity.""" - if len(_byok_cred_cache) >= _BYOK_CRED_CACHE_MAX_SIZE: - _byok_cred_cache.clear() - _byok_cred_cache[(user_id, server_id)] = (credential, time.monotonic()) +async def _invalidate_byok_cred_cache(user_id: str, server_id: str) -> None: + """Drop a stored-or-deleted BYOK credential from this worker's cache and from every peer worker's.""" + cache_key: Final = byok_credential_cache_key(user_id, server_id) + byok_credential_cache.delete_cache(cache_key) + await publish_auth_cache_invalidation(cache_key=cache_key) # Check if MCP is available @@ -623,9 +615,7 @@ if MCP_AVAILABLE: _stateful_session_locks: Final[dict[str, asyncio.Lock]] = {} _stateful_session_active_request_counts: Final[dict[str, int]] = {} _stateful_session_client_info: Final[dict[str, Implementation]] = {} # mutable-ok: cleared on session teardown - _admin_terminated_session_ids: Final[deque[str]] = deque( # mutable-ok: bounded ring, appended on admin termination - maxlen=MCP_ADMIN_TERMINATED_SESSION_IDS_MAX - ) + _admin_terminated_session_ids: Final[dict[str, float]] = {} # mutable-ok: admin-closed id -> last replay class _TerminableTransport(Protocol): async def terminate(self) -> None: ... @@ -697,6 +687,7 @@ if MCP_AVAILABLE: for session_id in list(_stateful_session_auth_context_last_seen): if session_id not in _stateful_session_auth_contexts: _remove_stateful_session_tracking(session_id) + _forget_expired_admin_terminated_session_ids(now) async def _enforce_stateful_session_cap_for_owner(owner: str) -> bool: """ @@ -2819,35 +2810,28 @@ if MCP_AVAILABLE: mcp_server: MCPServer, user_api_key_auth: UserAPIKeyAuth | None, ) -> str | None: - """Retrieve the stored BYOK credential for a user+server pair. - - Uses the shared _byok_cred_cache to avoid a DB round-trip on every - tool call within the TTL window. - """ + """Retrieve the stored BYOK credential for a user+server pair, served from the worker cache within its TTL.""" if not mcp_server.is_byok: return None user_id: Final = (user_api_key_auth.user_id if user_api_key_auth else None) or "" if not user_id: return None - cache_key: Final = (user_id, mcp_server.server_id) - cached: Final = _byok_cred_cache.get(cache_key) + cached: Final = get_cached_byok_credential(user_id, mcp_server.server_id) if cached is not None: - credential, ts = cached - if time.monotonic() - ts < _BYOK_CRED_CACHE_TTL: - return credential + return cached.credential from litellm.proxy._experimental.mcp_server.db import get_user_credential from litellm.proxy.proxy_server import prisma_client if prisma_client is None: return None - credential = await get_user_credential( + credential: Final = await get_user_credential( prisma_client=prisma_client, user_id=user_id, server_id=mcp_server.server_id, ) - _write_byok_cred_cache(user_id, mcp_server.server_id, credential) + cache_byok_credential(user_id, mcp_server.server_id, credential) return credential async def _check_byok_credential( @@ -2876,27 +2860,23 @@ if MCP_AVAILABLE: headers={"WWW-Authenticate": get_byok_www_authenticate()}, ) - # Check shared credential cache before hitting the DB. - cache_key: Final = (user_id, mcp_server.server_id) - cached: Final = _byok_cred_cache.get(cache_key) + cached: Final = get_cached_byok_credential(user_id, mcp_server.server_id) if cached is not None: - cached_cred, ts = cached - if time.monotonic() - ts < _BYOK_CRED_CACHE_TTL: - if cached_cred is None: - raise HTTPException( - status_code=401, - detail={ - "error": "byok_auth_required", - "server_id": mcp_server.server_id, - "server_name": mcp_server.server_name or mcp_server.name, - "message": ( - "No stored credential found for this BYOK server. " - "Complete the OAuth authorization flow to provide your API key." - ), - }, - headers={"WWW-Authenticate": get_byok_www_authenticate()}, - ) - return + if cached.credential is None: + raise HTTPException( + status_code=401, + detail={ + "error": "byok_auth_required", + "server_id": mcp_server.server_id, + "server_name": mcp_server.server_name or mcp_server.name, + "message": ( + "No stored credential found for this BYOK server. " + "Complete the OAuth authorization flow to provide your API key." + ), + }, + headers={"WWW-Authenticate": get_byok_www_authenticate()}, + ) + return from litellm.proxy._experimental.mcp_server.db import get_user_credential from litellm.proxy.proxy_server import prisma_client @@ -2920,7 +2900,7 @@ if MCP_AVAILABLE: user_id=user_id, server_id=mcp_server.server_id, ) - _write_byok_cred_cache(user_id, mcp_server.server_id, credential) + cache_byok_credential(user_id, mcp_server.server_id, credential) if credential is None: raise HTTPException( status_code=401, @@ -3906,6 +3886,24 @@ if MCP_AVAILABLE: key_auth: Final = auth_user.user_api_key_auth return key_auth is not None and key_auth.user_id == user_id + def _forget_expired_admin_terminated_session_ids(now: float) -> None: + for session_id in [ + session_id + for session_id, last_replayed in _admin_terminated_session_ids.items() + if now - last_replayed >= _STATEFUL_SESSION_IDLE_TIMEOUT_SECONDS + ]: + del _admin_terminated_session_ids[session_id] + + def _is_admin_terminated_session_id(session_id: str, now: float) -> bool: + last_replayed: Final = _admin_terminated_session_ids.get(session_id) + if last_replayed is None: + return False + if now - last_replayed >= _STATEFUL_SESSION_IDLE_TIMEOUT_SECONDS: + del _admin_terminated_session_ids[session_id] + return False + _admin_terminated_session_ids[session_id] = now + return True + async def terminate_mcp_gateway_sessions( *, session_id_prefix: str | None = None, @@ -3919,6 +3917,7 @@ if MCP_AVAILABLE: admission. Only sessions held by this worker process are affected. """ now: Final = time.monotonic() + _forget_expired_admin_terminated_session_ids(now) server_instances: Final = _stateful_server_instances() targets: Final = tuple( (session_id, auth_user) @@ -3928,7 +3927,7 @@ if MCP_AVAILABLE: ) terminated: Final = tuple(_gateway_session_for(session_id, auth_user, now) for session_id, auth_user in targets) for session_id, _ in targets: - _admin_terminated_session_ids.append(session_id) + _admin_terminated_session_ids[session_id] = now transport = server_instances.pop(session_id, None) _remove_stateful_session_tracking(session_id) if transport is not None: @@ -4064,7 +4063,7 @@ if MCP_AVAILABLE: await success_response(scope, receive, send) return True - if _session_id in _admin_terminated_session_ids: + if _is_admin_terminated_session_id(_session_id, time.monotonic()): terminated_response: Final = JSONResponse( status_code=404, content={ # mutable-ok: JSONResponse content must be a plain dict diff --git a/litellm/proxy/management_endpoints/mcp_management_endpoints.py b/litellm/proxy/management_endpoints/mcp_management_endpoints.py index 728ba9568e8..5326cf3415f 100644 --- a/litellm/proxy/management_endpoints/mcp_management_endpoints.py +++ b/litellm/proxy/management_endpoints/mcp_management_endpoints.py @@ -2317,7 +2317,7 @@ if MCP_AVAILABLE: _invalidate_byok_cred_cache, ) - _invalidate_byok_cred_cache(user_id, server_id) + await _invalidate_byok_cred_cache(user_id, server_id) return MCPUserCredentialResponse(server_id=server_id, has_credential=True) # save=False: credential not persisted return MCPUserCredentialResponse(server_id=server_id, has_credential=False) @@ -2348,7 +2348,7 @@ if MCP_AVAILABLE: _invalidate_byok_cred_cache, ) - _invalidate_byok_cred_cache(target_user_id, server_id) + await _invalidate_byok_cred_cache(target_user_id, server_id) return MCPUserCredentialResponse(server_id=server_id, has_credential=False) # ── OAuth2 user-credential endpoints ────────────────────────────────────── diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index d7d8413d2ce..48122549aea 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -307,6 +307,7 @@ from litellm.litellm_core_utils.sensitive_data_masker import ( from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, HTTPHandler from litellm.llms.openai_like.model_info import MODEL_INFO_REFRESH_SECONDS from litellm.llms.vertex_ai.vertex_llm_base import VertexBase +from litellm.proxy._experimental.mcp_server.byok_credential_cache import byok_credential_cache from litellm.proxy._lazy_features import attach_lazy_features, reserve_lazy_slot from litellm.proxy._types import * from litellm.proxy.analytics_endpoints.analytics_endpoints import ( @@ -7535,7 +7536,7 @@ class ProxyConfig: subscriber: Final = AuthCacheInvalidationSubscriber( redis_cache=redis_cache, user_api_key_cache=user_api_key_cache, - additional_in_memory_caches=(spend_counter_cache.in_memory_cache,), + additional_in_memory_caches=(spend_counter_cache.in_memory_cache, byok_credential_cache), ) self.auth_cache_invalidation_subscriber = subscriber subscriber.start() diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_credential_cache.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_credential_cache.py new file mode 100644 index 00000000000..8ec5b8642bc --- /dev/null +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_credential_cache.py @@ -0,0 +1,57 @@ +import json + +import pytest + +from litellm.proxy._experimental.mcp_server.byok_credential_cache import ( + CachedByokCredential, + byok_credential_cache, + byok_credential_cache_key, + cache_byok_credential, + get_cached_byok_credential, +) +from litellm.proxy.common_utils.auth_cache_invalidation_pubsub import AuthCacheInvalidationSubscriber +from litellm.proxy.common_utils.user_api_key_cache import UserApiKeyCache + + +class _FakeRedisCache: + namespace = None + + def init_async_client(self) -> object: + return object() + + +@pytest.fixture(autouse=True) +def _empty_cache(): + byok_credential_cache.flush_cache() + yield + byok_credential_cache.flush_cache() + + +def test_a_cached_negative_lookup_is_distinguishable_from_a_miss(): + assert get_cached_byok_credential("u-1", "srv-1") is None + cache_byok_credential("u-1", "srv-1", None) + assert get_cached_byok_credential("u-1", "srv-1") == CachedByokCredential(credential=None) + cache_byok_credential("u-1", "srv-1", "sk-stored") + assert get_cached_byok_credential("u-1", "srv-1") == CachedByokCredential(credential="sk-stored") + assert get_cached_byok_credential("u-1", "srv-2") is None + + +def test_peer_worker_invalidation_message_evicts_the_cached_credential(): + """The key a mutating worker broadcasts must be the key every other worker caches under.""" + cache_byok_credential("mallory", "srv-byok", "sk-revoked") + cache_byok_credential("alice", "srv-byok", "sk-kept") + subscriber = AuthCacheInvalidationSubscriber( + redis_cache=_FakeRedisCache(), # pyright: ignore[reportArgumentType] # subscriber is never started; only its message handler runs + user_api_key_cache=UserApiKeyCache(), + additional_in_memory_caches=(byok_credential_cache,), + ) + + subscriber._apply_message( # pyright: ignore[reportPrivateUsage] # exercising the real cross-worker message handler + { + "type": "message", + "data": json.dumps({"cache_key": byok_credential_cache_key("mallory", "srv-byok")}).encode(), + } + ) + + assert get_cached_byok_credential("mallory", "srv-byok") is None + assert get_cached_byok_credential("alice", "srv-byok") == CachedByokCredential(credential="sk-kept") diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py index 55accfb169d..df0d7916baf 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py @@ -592,7 +592,7 @@ async def test_check_byok_credential_missing_credential(monkeypatch): monkeypatch.delenv("PROXY_BASE_URL", raising=False) monkeypatch.delenv("SERVER_ROOT_PATH", raising=False) - monkeypatch.setattr(server_module, "_byok_cred_cache", {}) + server_module.byok_credential_cache.flush_cache() mock_prisma = MagicMock() with ( @@ -628,7 +628,7 @@ async def test_execute_byok_tool_missing_credential_advertises_api_key_flow(monk from litellm.types.mcp_server.mcp_server_manager import MCPServer monkeypatch.setenv("PROXY_BASE_URL", "https://gateway.example.com/proxy") - monkeypatch.setattr(mcp_module, "_byok_cred_cache", {}) + mcp_module.byok_credential_cache.flush_cache() server = MCPServer(server_id="byok-discovery", name="byok-discovery", transport=MCPTransport.http, is_byok=True) prisma = MagicMock() prisma.db.litellm_mcpusercredentials.find_unique = AsyncMock(return_value=None) @@ -677,6 +677,34 @@ async def test_check_byok_credential_has_credential(): await _check_byok_credential(server, user_auth) +@pytest.mark.asyncio +async def test_invalidate_byok_cred_cache_evicts_locally_and_broadcasts_the_same_key(): + """A revoked credential must stop being served here and on every peer worker within the TTL.""" + from litellm.proxy._experimental.mcp_server import server as server_module + from litellm.proxy._experimental.mcp_server.byok_credential_cache import byok_credential_cache_key + from litellm.proxy._types import UserAPIKeyAuth + from litellm.types.mcp_server.mcp_server_manager import MCPServer + + server = MCPServer(server_id="byok-revoke", name="byok-server", transport=MCPTransport.http, is_byok=True) + user_auth = UserAPIKeyAuth(user_id="mallory", api_key="sk-test") + server_module.byok_credential_cache.flush_cache() + db_lookup = AsyncMock(side_effect=["sk-before-revoke", None]) + publish = AsyncMock() + + with ( + patch("litellm.proxy._experimental.mcp_server.db.get_user_credential", new=db_lookup), + patch("litellm.proxy.proxy_server.prisma_client", MagicMock()), + patch.object(server_module, "publish_auth_cache_invalidation", new=publish), + ): + assert await server_module._get_byok_credential(server, user_auth) == "sk-before-revoke" + assert await server_module._get_byok_credential(server, user_auth) == "sk-before-revoke" + await server_module._invalidate_byok_cred_cache("mallory", "byok-revoke") + assert await server_module._get_byok_credential(server, user_auth) is None + + assert db_lookup.await_count == 2 + publish.assert_awaited_once_with(cache_key=byok_credential_cache_key("mallory", "byok-revoke")) + + @pytest.mark.asyncio async def test_check_byok_credential_db_unavailable_fails_closed(): """BYOK server with no prisma_client → 503, not silent pass. diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py index 4311a69d465..f496f22d5d2 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py @@ -2989,9 +2989,10 @@ async def test_admin_terminated_session_id_gets_404_instead_of_a_fresh_stateless """Once an admin closes a session, a client replaying its id must not be silently upgraded to a new stateless session by the stale-header path; it gets 404 and has to initialize again.""" try: + from starlette.types import Scope + from litellm.proxy._experimental.mcp_server import server as mcp_server from litellm.proxy._experimental.mcp_server.server import session_manager_stateful - from starlette.types import Scope except ImportError: pytest.skip("MCP server not available") @@ -3048,6 +3049,73 @@ async def test_admin_terminated_session_id_gets_404_instead_of_a_fresh_stateless mcp_server._admin_terminated_session_ids.clear() +@pytest.mark.asyncio +async def test_admin_terminated_session_id_stays_refused_while_replayed_and_is_forgotten_like_an_idle_session(): + """The refusal window slides on every replay, so a client that keeps retrying is never silently + upgraded to a stateless session no matter how many other sessions an admin closes later; an id + nobody has replayed for a full idle timeout is dropped from the table by the idle sweep.""" + try: + from starlette.types import Scope + + from litellm.proxy._experimental.mcp_server import server as mcp_server + from litellm.proxy._experimental.mcp_server.server import session_manager_stateful + except ImportError: + pytest.skip("MCP server not available") + + idle_timeout = mcp_server._STATEFUL_SESSION_IDLE_TIMEOUT_SECONDS + retrying_id, silent_id = "admin-closed-retrying", "admin-closed-silent" + contexts = { + session_id: mcp_server.MCPAuthenticatedUser( + user_api_key_auth=UserAPIKeyAuth(api_key="key-alice", user_id="alice"), + ) + for session_id in (retrying_id, silent_id) + } + live_transports = {session_id: MagicMock(terminate=AsyncMock()) for session_id in contexts} + + async def replay(session_id: str, now: float) -> tuple[bool, list[bytes]]: + scope: Scope = { + "type": "http", + "method": "POST", + "headers": [(b"content-type", b"application/json"), (b"mcp-session-id", session_id.encode())], + } + with patch.object(mcp_server.time, "monotonic", return_value=now): + handled = await mcp_server._handle_stale_mcp_session( + scope, AsyncMock(), AsyncMock(), session_manager_stateful + ) + return handled, [k for k, _ in scope["headers"]] + + try: + with ( + patch.object( # test-quality-ok: the transport registry is a module-level singleton; the suite's only seam + session_manager_stateful, "_server_instances", live_transports + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_contexts, contexts, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_client_info, {}, clear=True + ), + patch.dict( # test-quality-ok: the session tables are module-level singletons; the suite's only seam + mcp_server._stateful_session_auth_context_last_seen, {}, clear=True + ), + ): + with patch.object(mcp_server.time, "monotonic", return_value=1000.0): + closed = await mcp_server.terminate_mcp_gateway_sessions(user_id="alice") + assert closed.terminated_sessions == 2 + + for elapsed in (idle_timeout - 1, 2 * idle_timeout - 2, 3 * idle_timeout - 3): + assert await replay(retrying_id, 1000.0 + elapsed) == (True, [b"content-type", b"mcp-session-id"]) + + await mcp_server._purge_expired_stateful_session_auth_contexts(now=1000.0 + idle_timeout) + assert set(mcp_server._admin_terminated_session_ids) == {retrying_id} + + assert await replay(silent_id, 1000.0 + idle_timeout) == (False, [b"content-type"]) + assert await replay(retrying_id, 1000.0 + 4 * idle_timeout) == (False, [b"content-type"]) + assert mcp_server._admin_terminated_session_ids == {} + finally: + mcp_server._admin_terminated_session_ids.clear() + + @pytest.mark.asyncio async def test_initialize_request_with_existing_session_tracks_new_session(): try: diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py index f7567efcabc..301b3887922 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py @@ -395,6 +395,32 @@ async def test_invalidate_clears_every_identity_for_a_server(): assert mock_client.post.call_count == 3 +@pytest.mark.asyncio +async def test_per_user_token_delete_evicts_locally_and_broadcasts_to_peer_workers(): + """Revoking a user's OAuth token must not leave peer workers serving it from their in-memory layer.""" + from litellm.proxy import proxy_server + from litellm.proxy._experimental.mcp_server.oauth2_token_cache import MCPPerUserTokenCache + from litellm.proxy.common_utils.user_api_key_cache import UserApiKeyCache + + local_cache = UserApiKeyCache() + publish = AsyncMock() + token_cache = MCPPerUserTokenCache() + key = token_cache._cache_key("mallory", "srv-oauth") # pyright: ignore[reportPrivateUsage] # asserting the broadcast names the stored key + local_cache.in_memory_cache.set_cache(key, "encrypted-token") + + with ( + patch.object(proxy_server, "user_api_key_cache", local_cache), + patch( + "litellm.proxy.common_utils.auth_cache_invalidation_pubsub.publish_auth_cache_invalidation", + new=publish, + ), + ): + await token_cache.delete("mallory", "srv-oauth") + + assert local_cache.in_memory_cache.get_cache(key) is None + publish.assert_awaited_once_with(cache_key=key) + + @pytest.mark.asyncio async def test_m2m_mint_uses_admin_entered_token_url_when_issuer_yield_empties_resolved(): """A pinned issuer empties the resolved token_url while configured_token_url keeps the diff --git a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py index 2f33018599f..a69cfe0ac7d 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py @@ -5152,7 +5152,7 @@ async def test_admin_revokes_another_users_byok_credential(): ) delete_mock = AsyncMock(return_value=None) - invalidate_mock = MagicMock() + invalidate_mock = AsyncMock() with ( patch( # test-quality-ok: endpoint test stubs the Prisma client lookup "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", @@ -5174,7 +5174,7 @@ async def test_admin_revokes_another_users_byok_credential(): delete_mock.assert_awaited_once() assert delete_mock.await_args.args[1:] == ("mallory", "srv-byok-admin") - invalidate_mock.assert_called_once_with("mallory", "srv-byok-admin") + invalidate_mock.assert_awaited_once_with("mallory", "srv-byok-admin") assert result.has_credential is False @@ -5231,7 +5231,7 @@ async def test_user_naming_themselves_still_deletes_own_byok_credential(): new=delete_mock, ), patch.object( # test-quality-ok: the cache invalidator is module scoped; the suite's only seam - mcp_server, "_invalidate_byok_cred_cache", new=MagicMock() + mcp_server, "_invalidate_byok_cred_cache", new=AsyncMock() ), ): await delete_mcp_user_credential( diff --git a/tests/test_litellm/proxy/test_proxy_server.py b/tests/test_litellm/proxy/test_proxy_server.py index 41c4956dba6..acd13fef9a6 100644 --- a/tests/test_litellm/proxy/test_proxy_server.py +++ b/tests/test_litellm/proxy/test_proxy_server.py @@ -13919,3 +13919,74 @@ async def test_token_counter_loads_a_custom_tokenizer_once_per_identifier_revisi ] finally: litellm.utils._select_custom_tokenizer_helper.cache_clear() + + +@pytest.mark.asyncio +async def test_auth_cache_invalidation_subscriber_evicts_byok_credentials_cached_by_this_worker(): + """A peer worker's BYOK revocation broadcast must reach this worker's BYOK credential cache.""" + from redis.asyncio import Redis + + from litellm.proxy._experimental.mcp_server.byok_credential_cache import ( + byok_credential_cache, + byok_credential_cache_key, + cache_byok_credential, + get_cached_byok_credential, + ) + from litellm.proxy.common_utils.user_api_key_cache import UserApiKeyCache + + class _QueuePubSub: + def __init__(self, messages: list[object]) -> None: + self.queue: asyncio.Queue[object] = asyncio.Queue() + for message in messages: + self.queue.put_nowait(message) + + async def subscribe(self, *channels: str) -> None: + return None + + async def get_message(self, *, ignore_subscribe_messages: bool, timeout: float) -> object | None: + try: + return await asyncio.wait_for(self.queue.get(), timeout) + except asyncio.TimeoutError: + return None + + async def aclose(self) -> None: + return None + + class _PubSubRedisClient(Redis): + def __init__(self, pubsub: _QueuePubSub) -> None: + self._scripted_pubsub = pubsub + + def pubsub(self) -> _QueuePubSub: + return self._scripted_pubsub + + class _FakeRedisCache: + namespace = None + + def __init__(self, client: object) -> None: + self._client = client + + def init_async_client(self) -> object: + return self._client + + byok_credential_cache.flush_cache() + cache_byok_credential("mallory", "srv-byok", "sk-revoked-elsewhere") + message: Final = { + "type": "message", + "data": json.dumps({"cache_key": byok_credential_cache_key("mallory", "srv-byok")}).encode(), + } + proxy_config: Final = proxy_server_module.ProxyConfig() + proxy_config.start_auth_cache_invalidation_subscriber( + redis_cache=_FakeRedisCache(_PubSubRedisClient(_QueuePubSub([message]))), # pyright: ignore[reportArgumentType] # fake pub/sub capable redis; no live redis in this unit test + user_api_key_cache=UserApiKeyCache(), + ) + try: + for _ in range(200): + if get_cached_byok_credential("mallory", "srv-byok") is None: + break + await asyncio.sleep(0.01) + evicted: Final = get_cached_byok_credential("mallory", "srv-byok") is None + finally: + await proxy_config.stop_auth_cache_invalidation_subscriber() + byok_credential_cache.flush_cache() + + assert evicted, "the subscriber does not evict the BYOK credential cache on a peer worker's broadcast" From ce48a3fbcce70cc17f877de30b9c2a64ed4bc278 Mon Sep 17 00:00:00 2001 From: yassin Date: Fri, 18 Sep 2026 02:49:41 +0000 Subject: [PATCH 4/6] test(mcp): give the new cache and tombstone patches TQ008 reasons and match the keyword eviction call Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- tests/mcp_tests/test_per_user_oauth_cache.py | 4 +--- .../mcp_server/test_byok_oauth_endpoints.py | 12 +++++++++--- .../_experimental/mcp_server/test_mcp_server.py | 8 ++++++-- .../mcp_server/test_oauth2_token_cache.py | 6 ++++-- 4 files changed, 20 insertions(+), 10 deletions(-) diff --git a/tests/mcp_tests/test_per_user_oauth_cache.py b/tests/mcp_tests/test_per_user_oauth_cache.py index 141b906fce9..ac453df8fa5 100644 --- a/tests/mcp_tests/test_per_user_oauth_cache.py +++ b/tests/mcp_tests/test_per_user_oauth_cache.py @@ -331,9 +331,7 @@ class TestMCPPerUserTokenCache: with patch("litellm.proxy.proxy_server.user_api_key_cache", mock_dual_cache): await cache.delete("alice", "slack-test") - mock_dual_cache.async_delete_cache.assert_called_once_with( - "mcp:per_user_token:alice:slack-test" - ) + mock_dual_cache.async_delete_cache.assert_called_once_with(key="mcp:per_user_token:alice:slack-test") mock_dual_cache.async_set_cache.assert_not_called() @pytest.mark.asyncio diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py index df0d7916baf..87e23893616 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_byok_oauth_endpoints.py @@ -692,9 +692,15 @@ async def test_invalidate_byok_cred_cache_evicts_locally_and_broadcasts_the_same publish = AsyncMock() with ( - patch("litellm.proxy._experimental.mcp_server.db.get_user_credential", new=db_lookup), - patch("litellm.proxy.proxy_server.prisma_client", MagicMock()), - patch.object(server_module, "publish_auth_cache_invalidation", new=publish), + patch( # test-quality-ok: the DB row lookup is the only seam below the credential resolver; no Prisma fake exists + "litellm.proxy._experimental.mcp_server.db.get_user_credential", new=db_lookup + ), + patch( # test-quality-ok: the resolver reads the module-level prisma_client singleton; the suite's only seam + "litellm.proxy.proxy_server.prisma_client", MagicMock() + ), + patch.object( # test-quality-ok: the redis publisher is module-level; asserting the broadcast without a redis + server_module, "publish_auth_cache_invalidation", new=publish + ), ): assert await server_module._get_byok_credential(server, user_auth) == "sk-before-revoke" assert await server_module._get_byok_credential(server, user_auth) == "sk-before-revoke" diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py index f496f22d5d2..0e2ceb98987 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py @@ -3078,7 +3078,9 @@ async def test_admin_terminated_session_id_stays_refused_while_replayed_and_is_f "method": "POST", "headers": [(b"content-type", b"application/json"), (b"mcp-session-id", session_id.encode())], } - with patch.object(mcp_server.time, "monotonic", return_value=now): + with patch.object( # test-quality-ok: the stale-session handler reads the clock directly; no injectable now + mcp_server.time, "monotonic", return_value=now + ): handled = await mcp_server._handle_stale_mcp_session( scope, AsyncMock(), AsyncMock(), session_manager_stateful ) @@ -3099,7 +3101,9 @@ async def test_admin_terminated_session_id_stays_refused_while_replayed_and_is_f mcp_server._stateful_session_auth_context_last_seen, {}, clear=True ), ): - with patch.object(mcp_server.time, "monotonic", return_value=1000.0): + with patch.object( # test-quality-ok: termination stamps the tombstone from the clock directly; no injectable now + mcp_server.time, "monotonic", return_value=1000.0 + ): closed = await mcp_server.terminate_mcp_gateway_sessions(user_id="alice") assert closed.terminated_sessions == 2 diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py index 301b3887922..30d0f17a099 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_oauth2_token_cache.py @@ -409,8 +409,10 @@ async def test_per_user_token_delete_evicts_locally_and_broadcasts_to_peer_worke local_cache.in_memory_cache.set_cache(key, "encrypted-token") with ( - patch.object(proxy_server, "user_api_key_cache", local_cache), - patch( + patch.object( # test-quality-ok: the token cache reads the module-level user_api_key_cache singleton; the suite's only seam + proxy_server, "user_api_key_cache", local_cache + ), + patch( # test-quality-ok: the redis publisher is module-level; asserting the broadcast without a redis "litellm.proxy.common_utils.auth_cache_invalidation_pubsub.publish_auth_cache_invalidation", new=publish, ), From 307df09792c1088a2d7b63e424c8040fa7746595 Mon Sep 17 00:00:00 2001 From: yassin Date: Fri, 18 Sep 2026 18:51:43 +0000 Subject: [PATCH 5/6] test(mcp): assert the self-revoke response instead of echoing the delete mock Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../test_mcp_management_endpoints.py | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py index a69cfe0ac7d..afadd6f3d19 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py @@ -24,6 +24,7 @@ from litellm.proxy._types import ( LiteLLM_MCPServerTable, LitellmUserRoles, MCPTransport, + MCPUserCredentialResponse, NewMCPServerRequest, UpdateMCPServerRequest, UserAPIKeyAuth, @@ -5220,7 +5221,11 @@ async def test_user_naming_themselves_still_deletes_own_byok_credential(): delete_mcp_user_credential, ) - delete_mock = AsyncMock(return_value=None) + deleted_rows: list[tuple[str, str]] = [] # mutable-ok: test-local recorder for the fake delete boundary + + async def _fake_delete_user_credential(_prisma_client: object, user_id: str, server_id: str) -> None: + deleted_rows.append((user_id, server_id)) + with ( patch( # test-quality-ok: endpoint test stubs the Prisma client lookup "litellm.proxy.management_endpoints.mcp_management_endpoints.get_prisma_client_or_throw", @@ -5228,19 +5233,20 @@ async def test_user_naming_themselves_still_deletes_own_byok_credential(): ), patch( # test-quality-ok: endpoint test stubs the credential row delete "litellm.proxy.management_endpoints.mcp_management_endpoints.delete_user_credential", - new=delete_mock, + new=_fake_delete_user_credential, ), patch.object( # test-quality-ok: the cache invalidator is module scoped; the suite's only seam mcp_server, "_invalidate_byok_cred_cache", new=AsyncMock() ), ): - await delete_mcp_user_credential( + result = await delete_mcp_user_credential( server_id="srv-byok-self", user_api_key_dict=_make_user_auth("user-self"), user_id="user-self", ) - assert delete_mock.await_args.args[1:] == ("user-self", "srv-byok-self") + assert deleted_rows == [("user-self", "srv-byok-self")] + assert result == MCPUserCredentialResponse(server_id="srv-byok-self", has_credential=False) @pytest.mark.asyncio From cf22e4b171da3b923765a7e597d78378f03fb9c9 Mon Sep 17 00:00:00 2001 From: yassin Date: Fri, 18 Sep 2026 19:38:38 +0000 Subject: [PATCH 6/6] fix(proxy): make SettingsStore.clear terminate when the config file owns a key MutableMapping.clear pops items until the mapping is empty, but __delitem__ keeps config-owned keys, so clear spun forever on any store that had loaded a config file. unittest.mock.patch.dict calls clear on exit, which is why the proxy-infra and proxy-endpoints shards hung at 99 percent until the 20 minute job timeout on every run since the store started refusing config-owned writes Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/config_resolvers/settings_store.py | 4 ++++ .../config_resolvers/test_settings_store.py | 16 ++++++++++++++++ 2 files changed, 20 insertions(+) diff --git a/litellm/proxy/config_resolvers/settings_store.py b/litellm/proxy/config_resolvers/settings_store.py index 079d262319f..a3d5da4b343 100644 --- a/litellm/proxy/config_resolvers/settings_store.py +++ b/litellm/proxy/config_resolvers/settings_store.py @@ -98,6 +98,10 @@ class SettingsStore(MutableMapping[str, JsonValue]): def __len__(self) -> int: return sum(1 for _ in self) + def clear(self) -> None: + self._runtime_values = _EMPTY_VALUES + self._deleted_runtime_keys = frozenset(key for key in self._keys() if not self.owned_by_config(key)) + def _clear_runtime(self) -> None: self._runtime_values = _EMPTY_VALUES self._deleted_runtime_keys = frozenset() diff --git a/tests/test_litellm/proxy/config_resolvers/test_settings_store.py b/tests/test_litellm/proxy/config_resolvers/test_settings_store.py index aa410deac43..0b472938b74 100644 --- a/tests/test_litellm/proxy/config_resolvers/test_settings_store.py +++ b/tests/test_litellm/proxy/config_resolvers/test_settings_store.py @@ -168,6 +168,22 @@ def test_settings_store_refuses_a_runtime_write_to_a_config_owned_key() -> None: assert store.source("max_parallel_requests") == "config" +@pytest.mark.timeout(10) +def test_settings_store_clear_terminates_and_keeps_config_owned_keys() -> None: + store: Final = SettingsStore("general_settings") + store.load_yaml({"max_parallel_requests": 3}) + store.apply_db_row("general_settings", {"max_file_size_mb": 9}) + store["ui_access_mode"] = "admin_only" + before: Final = dict(store) + + store.clear() + cleared: Final = dict(store) + store.update(before) + + assert cleared == {"max_parallel_requests": 3} + assert dict(store) == before + + def test_settings_store_reports_the_config_owned_keys_a_write_would_change() -> None: store: Final = SettingsStore("general_settings") store.load_yaml({"max_parallel_requests": 3, "ui_access_mode": "admin_only"})