Skip to content

AlphaOne HyperRouter v1.6.02

Latest
Latest

Choose a tag to compare

View all tags
@arjuna325 arjuna325 released this 24 Aug 11:33
v1.6.02
297dc6a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Changelog

All notable changes to AlphaOne HyperRouter Engine are documented here.

[1.6.02] — 2026-08-24

Added

  • Added Local Cloud (Modal) as an alternative Local AI source alongside Local Device GGUF.
  • Added support for user-owned Modal Managed Inference Endpoints through an OpenAI-compatible interface.
  • Added bounded handling for Modal scale-to-zero / cold-start conditions before provider rollover.
  • Added Cloudflare Workers AI as a Direct Provider with support for user-configured Workers AI Model IDs.
  • Cloudflare Workers AI now participates in Direct Provider priority selection and normal pre-commit Auto RollOver.

Changed

  • Local Cloud request preparation was improved for compatibility with OpenAI-compatible inference servers, including Qwen/SGLang deployments.
  • Local Cloud non-stream and streaming responses now preserve compatible tool-calling and usage information more consistently.
  • Responses API egress now consistently preserves the public client-facing model identity:
    alphaone-hyperrouter.
  • Physical upstream model IDs remain internal and are used only where required for routing and diagnostics.
  • Responses tool-choice compatibility was improved for function tools and namespaced agent workflows.
  • Pre-commit streaming behavior was hardened so failed or cold-starting Local Cloud attempts cannot incorrectly claim the downstream response lifecycle.
  • Cloudflare model selection is intentionally independent of the High / Medium / Low / Custom HyperRouter profile. The configured Cloudflare model is used directly.
  • Cloudflare Workers AI uses HyperRouter's existing OpenAI-compatible Chat and Responses compatibility pipeline.

Fixed

  • Fixed Local Cloud compatibility failures where upstream inference servers required the system message to appear first.
  • Fixed Responses function tool-choice translation compatibility with OpenAI-compatible Local Cloud servers.
  • Fixed physical upstream model IDs appearing in client-facing Responses output.
  • Improved provider rollover behavior for Local Cloud failures before downstream output is committed.

Security

  • Local Cloud authentication credentials are protected at rest using Windows DPAPI with DataProtectionScope.CurrentUser.
  • Cloudflare API credentials are also protected using the same local DPAPI security boundary.
  • Existing secrets are not rendered back into the Settings page.
  • Leaving an existing secret field blank preserves the previously configured credential.
  • Provider credentials remain inside HyperRouter and are not exposed through client model discovery.
  • The stable client-facing model remains alphaone-hyperrouter.

Verification

Validated Local Cloud workflows include:

  • Cold-start recovery.
  • Chat Completions non-stream and streaming.
  • Tool calls and tool-result continuation.
  • Responses API non-stream and streaming.
  • Responses model-identity preservation.
  • Function-call continuation.
  • Clean pre-commit rollover from Local Cloud to another provider.

Validated Cloudflare Workers AI workflows include:

  • Chat Completions non-stream and streaming.
  • Reasoning and visible-content streaming.
  • Tool calls and tool-result continuation.
  • User-selected Workers AI Model IDs.
  • Rate-limit classification and Direct Provider Auto RollOver behavior.

Full release regression testing completed successfully before release lock.

Assets 3
Loading