generated: '2026-09-19' method: probed source: >- Version and freshness markers read from the live documents on 2026-09-19: the two OpenAPI copies, the two agent cards, agents.json, the gateway freshness beacon, and the JSON-RPC -32601 message that enumerates legacy method aliases. The provider publishes no versioning policy, deprecation policy, SLA or status page; those absences are the finding. versioning: scheme: semver-in-document (no URI or header versioning) current: openapi_gateway_host: 1.3.0 # https://ai.rsperformance.online/.well-known/openapi.json openapi_canonical_host: 1.2.2 # https://rsperformance.online/.well-known/openapi.json — one minor behind, missing gatewayAnswerRouting a2a_agent_card: 2.5.0 # /.well-known/agent-card.json, protocolVersion 0.3.0 a2a_legacy_card: 2.2.0 # /.well-known/agent.json mcp_server: 0.3.0 # serverInfo.version from initialize on https://mcp.rs3d.pl/ agents_manifest: 2026-09-20-live # date-stamped, regenerated daily ai_resources_manifest: '2026-04-13' docs: null note: >- Versions are declared inside each document and drift between hosts (OpenAPI 1.2.2 vs 1.3.0). Endpoints themselves are unversioned (POST /, /api/search). No Stripe-Version-style header, no /v1/ path segment. deprecation: policy_url: null sunset_header: false deprecation_header: false legacy_surfaces: - what: JSON-RPC legacy method aliases SendMessage, SendStreamingMessage, GetTask, ListTasks, CancelTask, SubscribeToTask status: still served; announced only inside the -32601 error text as "legacy aliases" replacement: message/send, message/stream, tasks/get, tasks/list, tasks/cancel, tasks/subscribe - what: /.well-known/agent.json (pre-0.3 agent card path) status: still served with an OLDER card (2.2.0, 4 skills) than the canonical path (2.5.0, 9 skills) replacement: /.well-known/agent-card.json - what: diagnostic_search MCP tool status: live; described by the server as a "compatibility bridge for legacy Diagnosta clients" replacement: semantic_search note: No published timeline for retiring any of the above. sla: url: null uptime_target: null note: None published. The A2A card's x_rs_ai_friendly block states a fair-use rate limit and cache TTL (>= 600 s on /.well-known/agent-card.json) but no availability commitment. status_page: null freshness_signals: - url: https://ai.rsperformance.online/.well-known/freshness.json fields: [last_updated, content_version, next_update_expected, sitemap_lastmod] cadence: daily (beacon_refreshed_by vps_freshness_beacon.py, daily cron) observed: last_updated 2026-09-19T04:00:01Z - url: https://rsperformance.online/feeds/changes.json format: JSON Feed 1.1 observed: 20 items, newest date_modified 2026-09-20T02:42:13+02:00 - tool: health (MCP) note: reports service health and Qdrant status on demand — a liveness probe, not a status page deprecated_operations: []