Skip to content

v0.4.0 — Declarative standalone config

Choose a tag to compare

View all tags
@github-actions github-actions released this 22 Jul 03:31
v0.4.0
1cb66cb
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

🚀 Get started

New to AISIX? Get the gateway running and route your first LLM request in minutes:

📦 Download

Pull the container image from the GitHub Container Registry:

docker pull ghcr.io/api7/aisix:0.4.0

Rolling tags (0.4, latest) and the docker.io/api7/aisix mirror for private/offline deployments are listed on the package page.

The theme of this release: run AISIX from files. A new file-based resource source lets standalone deployments define providers, keys, MCP/A2A agents, and guardrails as canonical YAML documents — the same shape the control plane writes to etcd — with no Admin API calls. Plus deeper latency observability and guardrail streaming fixes.

📄 Declarative standalone mode

  • File-based resource source (resources.yaml) — standalone deployments load their full resource set from a file, no Admin API required (#759)
  • Canonical resource model — provider_key, MCP, and A2A field names converge on one canonical schema shared across control plane, data plane, and files (#757), declared the single source of truth (#753, #754)
  • e2e now seeds cases by writing canonical documents straight to etcd, matching how real deployments load (#750, #756)

🔭 Observability

  • TTFT + end-to-end latency histograms — bucketed metrics for SLO quantiles (e.g. p95/p99) on time-to-first-token and total request latency (#751)
  • Failed-request capture — the request body is recorded on failed chat/messages/responses requests (#749), with structure-preserving truncation so large payloads stay valid (#747)
  • /status/config — load-observability endpoint plus config metrics (#774)
  • Deployment state now derives from the target's serving state (#743)

🧭 Routing

  • fallback_on_statuses — opt selected upstream status codes into retry/failover (#752)

🛡️ Guardrails

  • Monitor-mode output guardrails no longer hold back or fail streaming closed on /v1/responses (#760)
  • Provider error-response body is logged on a config-error 4xx (#773); monitor mode pinned against provider failures in tests (#746)

⚠️ Breaking changes

  • A2A auth — a2a_agent.auth_type converges on the control-plane enum; the oauth2 value is dropped (#755)

📚 Docs

  • Admin API + guardrail schema descriptions polished (#741, #742); self-hosted vs managed API scope clarified (#762); 409 on A2A agent update documented (#758)
  • CONTRIBUTING + RELEASING guides added (#745)

Full changelog: v0.3.1...v0.4.0

Assets 2
Loading