Skip to content

The Open Standard for Generative UI

5.1 KB
Raw Blame

@openuidev/server

0.3.0

Minor Changes

  • #1348 9e76b1d Thanks @Aditya-thesys! - OpenUI Lang 1.0.

    • Upgrade @openuidev/lang-core and @openuidev/react-lang together, and together with the server that generates the system prompt. A 1.0 prompt teaches bare action steps (@ToAssistant("Hi") without Action([...])), which a 0.x client does not run.
    • react-ui 0.18 requires react-lang 1.0 (peer >=1.0.0 <2.0.0); upgrade them together.
    • browser-bundle 0.2 bundles react-lang 1.0 and react-ui 0.18.
    • server 0.3 uses lang-core 1.0 for autofix (parser and system prompt); stay on server 0.2 with a 0.x client. Autofix ignores warnings and accepts a ternary entry, so a program that renders does not start a fix round.
    • vue-lang, svelte-lang and angular-lang stay on 0.4 (OpenUI Lang 0.x). Keep @openuidev/lang-core ^0.4 for the prompts they render.
    • A program whose root statement is written last shows nothing while it streams, until the root line arrives. Write root first, as the system prompt asks.
    • A cut-off output that never reached its root statement, or any statement calling the root component, now renders blank instead of a partial tree.

Patch Changes

0.2.0

Minor Changes

  • #1268 f1f66c4 Thanks @AbhinRustagi! - Add optional Renderer slots for content, query loading, errors, and retry controls. Keep queryLoader supported on Renderer and A2UIRenderer. Preserve the default loading spinner; error/retry views render nothing unless supplied by the application.

    Preserve the existing query lifecycle while waiting for generation to finish. Keep mounted input state through retries and refresh query results when generated scripts change.

Patch Changes

0.1.1

Patch Changes

0.1.0

Minor Changes

  • #1194 68c9280 Thanks @AbhinRustagi! - Add createAutofix from @openuidev/server/openai for Chat Completions and @openuidev/server/vercel for Vercel AI SDK UI message streams. Both expose fix({ generation, messages, signal }) for completed text. Streaming uses chat.completions({ stream, messages, signal }) for Chat Completions, responses() for OpenAI Responses (not yet supported), and ai({ stream, messages, signal }) for Vercel AI SDK UI message streams. Validate locally and call Autofix only for invalid UI. Streaming appends the complete corrected program as native text deltas before completion, preserving tool calls and metadata. A closed ```openui-lang fence is held back so the repair is inserted inside that same fence. The Vercel AI SDK 7 helper accepts toUIMessageStream({ stream: result.stream }) output and returns AI SDK UI message SSE. @openuidev/server/openai also exports storeChatCompletionHistory and chatCompletionMessagesToItems (moved off the package root).