Skip to content

v0.13.5

Choose a tag to compare

View all tags
@github-actions github-actions released this 06 Jul 02:11
· 959 commits to main since this release
v0.13.5
0c7bdab

[0.13.5] — 2026-07-06

Groq and xAI providers

Two new first-class providers, both wired through a shared makeOpenAICompatProvider factory so they inherit the full OpenAI-compatible stack — streaming, native function calling, and structured output — without a bespoke adapter each.

ReactiveAgents.create().withProvider("groq").withModel("llama-3.3-70b-versatile").build();
ReactiveAgents.create().withProvider("xai").withModel("grok-4").build();
  • Live-verified end to end on both providers: plain completion plus native tool-call round-trips.
  • Provider-aware native-function-call fallback; logprobs and embeddings are capability-gated (Groq/xAI expose neither).
  • Structured output on Groq is model-dependent — json_schema strict works on gpt-oss and some models, others accept only json_object; the parse-retry loop covers the gap.
  • Provider count is now 8 (Anthropic, OpenAI, Gemini, LiteLLM, Ollama, test, Groq, xAI).

Agentic UI Kit — @reactive-agents/ui-core

New framework-agnostic headless package (@reactive-agents/ui-core) holding the shared controllers behind the UI bindings: a progressive UI-tree reconciler, a task-inbox fetch controller, and interaction + approval POST controllers. The React, Svelte, and Vue packages now delegate to these controllers instead of each re-implementing the wire protocol.

  • React rewired onto ui-core with a core useRun hook and the full v1 family surface — Interact (AgentPrompt/ChoiceCard), Inbox (useTaskInbox/TaskInbox), Observe (useRunCost/useRunSteps + CostMeter/StepTimeline), Render (AgentSurface registry + UI-tree schema), plus a useResumableRun hook and an AgentDevtools overlay with testing/styles subpaths.
  • Svelte and Vue rewired onto the same controllers (createRun, createInteractions, createResumableRun, run cost/steps), restoring requestInit/header pass-through on structured streams.

Cortex request_user_input interaction rail

Cortex gains a durable request-for-input rail: runner methods plus a .withInteraction(...) surface, an interaction-watcher, and a real end-to-end pause → register → respond → resume flow. The Cortex UI renders a live Interact panel and streaming structured previews.

Surfaced run errors instead of a generic message

Reasoning failures now propagate the real error string to result.error end to end. Previously the kernel captured the full message in state.error but normalizeReasoningResult dropped it during its whitelist rebuild, so callers only ever saw a generic "Reasoning failed". An error field now rides ReasoningResultSchema / ExecutionReasoningResult through normalization; e.g. a bad model id now surfaces "…404 The model … does not exist" on result.error.

Assets 2
Loading