All posts tagged

Validate AISIX decision-model requests, answer ordering, refusals, token usage, and failure paths with a runnable application-side contract checker.

AISIX 1.7.0 adds inherited cost-attribution tags so teams can connect AI usage, budgets, and rate-limit policy to the organization that owns the work.

Route Codex through AISIX with scoped credentials and a tested Responses contract, while preserving tools, streaming, continuation, and useful errors.

Build LLM monitoring with Prometheus using terminal request outcomes, streaming latency, TTFT, fallback, and policy signals from AISIX 1.6.0.

AISIX 1.6.0 records supported streamed requests when they end, so request metrics and duration histograms reflect terminal outcomes instead of response start.

Route Claude Code through AISIX with scoped caller keys, stable model aliases, limits, guardrails, and usage visibility.

Secure MCP with resource-bound authentication, least-privilege tools, separate credentials, correctly timed guardrails, and egress controls.

Design an AI gateway that safely serves accepted policy through control-plane outages, restarts, stale state, shared-service failures, and recovery drills.

Use stable model aliases, contract tests, canary traffic, and rollback gates to migrate LLM providers without coupling applications to constant model churn.

AISIX 1.5.0 extends Model Group routing, retry, failover, balancing, and per-attempt usage records to supported single-request AI endpoints.

Connect Dify, n8n, and Open WebUI to AISIX with stable model aliases, protected provider keys, policy, usage attribution, and workload tests.

AISIX 1.4.0 can serve in degraded mode when Redis is unavailable at startup and can retry usage delivery within strict limits when deduplication evidence permits.

Test an OpenAI-compatible AI gateway across auth, model discovery, streaming, tool calls, errors, usage, and provider translation.

AISIX 1.3.0 expands structured output and Responses translation, with explicit guardrail boundaries and clearer records of completed agent requests.

Learn how a private LLM gateway secures vLLM, SGLang, and Ollama with stable model aliases, traffic policy, routing, and observability.

Learn how an MCP Registry catalogs servers, differs from a marketplace and gateway, and supports governed MCP discovery and tool execution.

Learn how an AI Gateway governs Files, Batch, and Fine-tuning APIs with secure routing, credential isolation, job lifecycle control, and usage attribution.

Learn how a multimodal AI Gateway governs image, audio, video, and realtime APIs across identity, policy, provider routing, cost, and telemetry.

AISIX 1.0.0 adds semantic guardrail testing and decision scores, with clearer failure policies and controls for production AI traffic.

Learn how AI Gateway audit logging connects requests, policy outcomes, configuration changes, retention, and exports for AI governance and investigations.

Learn how AI Gateway PII redaction detects, masks, or blocks sensitive data in LLM requests and responses before it crosses a trust boundary.

Use AISIX AI Gateway guardrails safely with monitor and block modes, input and output hooks, scoped policies, streaming controls, and failure handling.

Rotate LLM provider credentials in AISIX Cloud or open-source AISIX while keeping caller API keys and model aliases stable, tested, and recoverable.

AISIX AI Gateway now serves the stateless MCP 2026-07-28 revision alongside every 2025 revision on one endpoint, with per-server protocol pinning and a one-call-one-execution guarantee.

Compare open source AI Gateway options and learn how to evaluate AI traffic control, security, cost governance, and enterprise readiness.

AISIX model ensembles fan one chat request to a panel of LLMs, then a judge synthesizes one answer — so several low-cost models can rival a frontier model.

Learn why API7 built AISIX from scratch as an AI-native gateway for LLM workloads, with unified model access, observability, and fine-grained traffic control.