Govern · inference

Model Gateway

Vendor-neutral routing across hosted and sovereign on-premise inference, with residency and tenant enforcement. Configuration is read-only in this console.

Providers

7

routes

Healthy

6

1 watch

p50 latency

767ms

live

Gateway rps

64

live

Sovereign

2

on-prem

Blocked

2

policy

Providers & routing

Fallback chain, residency allow-lists and immutable decision audit.

Read-only Agent OS

Providers and fallback chain

Ordered by fallback priority

FallbackProviderStatusLatencyError rateData residencyCost tierModelsAllowed tenants
#1Anthropichealthy690 ms0.3%US / EU routinghigh
claude-sonnet-4.5claude-haiku-4
Nordic Federated BankHelios Energy GridMeridian Health Systems
#2OpenAIhealthy780 ms0.4%US / EU routinghigh
gpt-4.1gpt-4.1-mini
Nordic Federated BankMeridian Health Systems
#3Azure OpenAIhealthy810 ms0.6%EU (Sweden Central)high
azure-gpt-4oazure-gpt-4o-mini
Nordic Federated BankHelios Energy GridAtlas Public Sector
#4AWS Bedrockhealthy940 ms0.9%EU (Frankfurt)medium
bedrock-claude-3.7bedrock-llama-3.1
Helios Energy GridAtlas Public Sector
#5Google Geminidegraded1420 ms3.1%USmedium
gemini-2.5-progemini-2.5-flash
Meridian Health Systems
#6Ollama (on-prem)healthy320 ms1.4%On-premiselow
llama-3.3-70bqwen2.5-32b
Atlas Public SectorHelios Energy Grid
#7vLLM Clusterhealthy410 ms1.1%On-premise (sovereign)low
llama-3.3-70b (vLLM)mistral-large
Nordic Federated BankAtlas Public Sector

Routing decision audit log

Immutable gateway decisions with correlation ID · 7 routed · 3 fallback · 2 blocked

Correlation IDTime (UTC)TenantAgentRequestedChosen providerDecisionFallback fromResidencyLatencyTokensOutcome
corr-4830-g22026-08-02 07:19:05Nordic Federated BankSupervisor Agent 01claude-sonnet-4.5Anthropicclaude-sonnet-4.5routedEU-West (Ireland)396 ms26,410completedPrimary route healthy
corr-4830-g12026-08-02 07:12:44Atlas Public SectorPlanner Agent 02gpt-4.1blockedn/arequest rejected at gatewayAgent passport expired, gateway refused issuance
corr-sec-90022026-08-02 06:58:00Helios Energy GridLinux Agent 02claude-sonnet-4.5Anthropicclaude-sonnet-4.5routedEU-West (Ireland)371 ms4,180completed, output redacted by guardrailPrimary route healthy
corr-4821-a42026-08-02 06:48:02Nordic Federated BankVerification Agent 01gemini-2.5-proOllama Edgemistral-small-3fallbackGoogle GeminiOn-premise (sovereign)742 ms3,260completed on edge routeResidency guard: EU-only tenant, provider region unavailable
corr-4821-a32026-08-02 06:45:33Nordic Federated BankNetwork Agent 01gpt-4.1Azure OpenAIazure-gpt-4ofallbackOpenAIEU-North (Sweden)512 ms12,980completed on fallback #2Primary p50 latency breached 1200 ms threshold
corr-4821-a12026-08-02 06:41:12Nordic Federated BankKubernetes Agent 01claude-sonnet-4.5Anthropicclaude-sonnet-4.5routedEU-West (Ireland)384 ms18,240completedPrimary route healthy, tenant allow-listed
corr-sec-90032026-08-02 06:31:00Meridian Health SystemsKubernetes Agent 03gemini-2.5-problockedUS-Central (denied)request rejected at gatewayTenant not on provider allow-list (clinical residency policy)
corr-4818-b12026-08-02 04:15:41Nordic Federated BankApplication Agent 02claude-sonnet-4.5Anthropicclaude-sonnet-4.5routedEU-West (Ireland)402 ms9,420completedPrimary route healthy
corr-4809-d12026-08-01 22:02:10Helios Energy GridNetwork Agent 01azure-gpt-4ovLLM Clusterllama-3.3-70bfallbackAzure OpenAIOn-premise (sovereign)688 ms21,460completed on sovereign routeProvider returned 429 rate limit twice
corr-4802-c12026-08-01 18:33:20Meridian Health SystemsDatabase Agent 01llama-3.3-70bvLLM Clusterllama-3.3-70broutedOn-premise (sovereign)610 ms15,380completedClinical tenant pinned to sovereign inference
corr-4795-e12026-07-31 11:06:44Atlas Public SectorLinux Agent 02gpt-4.1OpenAIgpt-4.1routedEU-West (Ireland)441 ms7,310completedPrimary route healthy
corr-4788-f12026-07-30 08:21:02Helios Energy GridApplication Agent 01bedrock-claude-3.7AWS Bedrockbedrock-claude-3.7routedEU-Central (Frankfurt)528 ms11,240completedCost-tier preference for batch correlation work