genui_x 0.0.13

Connect any LLM — Claude, OpenAI, Gemini, Ollama, or a custom proxy — to Google's genui (Generative UI) framework with one Transport.

0.0.13

• Upgrade dependency to genui: ^0.9.2 in both the package and example app.
• Upgrade transport dependency to http: ^1.6.0.
• Improve SSE parser compatibility for OpenAI-compatible proxies by accepting both data: {...} and data:{...} line formats.
• Harden transport error handling by skipping empty outbound messages and surfacing API errors as user-safe text chunks.
• Harden travel example rendering against model output variance: register both travel-catalog and legacy travel IDs, accept flat or nested (data) props, and add safe fallbacks for missing fields.
• Smooth travel example waiting UX by keeping progress visible until a real UI/text result is rendered.
• Add troubleshooting notes in README for 200-with-delayed-UI behavior, proxy SSE formatting differences, catalog ID mismatches, and prop shape compatibility.

0.0.12

• Add GenuiXTransport.ollama() factory — thin wrapper around GenuiXTransport.openai() configured for Ollama's OpenAI-compatible endpoint. Defaults to http://localhost:11434 and llama3.2. No API key required (a placeholder bearer token is sent so the OpenAI-compat layer accepts the request). Disables retries by default (maxRetries: 0) since local servers don't rate-limit.
• Add example/lib/ollama_main.dart — runnable local-Ollama chat demo with enforceJsonMode: true for stricter A2UI compliance from local models. Supports OLLAMA_BASE_URL and OLLAMA_MODEL --dart-defines.
• README overhaul for discoverability — new tagline, provider matrix table (Anthropic / OpenAI / Gemini / Ollama / custom proxy at a glance), CI badge, demo-GIF placeholder block, Ollama backend section, and Ollama example run instructions.
• pubspec.yaml — add homepage and documentation URLs, refresh topics to [genui, llm, claude, gemini, openai] (drop redundant flutter and broad ai), and stage a commented-out screenshots: block for the demo GIF.
• No public API removals or signature changes; existing call sites for .anthropic(), .openai(), and .gemini() are byte-identical.

0.0.11

• Add GenuiXTransport.gemini() factory — pre-configures Google's Generative Language API endpoint (/v1beta/models/{model}:streamGenerateContent?alt=sse), the x-goog-api-key header, and the Gemini SSE format. Works with gemini-2.5-flash, gemini-2.5-pro, and Vertex AI / proxy endpoints that mirror the Generative Language API surface.
• Add GenuiXStreamFormat.gemini and a Gemini SSE parser (GeminiSseParser) that extracts text from candidates[*].content.parts[*].text.
• Add enforceJsonMode option on GenuiXTransport.openai() and GenuiXConfig — when true, injects response_format: {"type": "json_object"} for tighter A2UI compliance on OpenAI and OpenAI-compatible backends. Respects user-supplied requestBodyOverrides['response_format'].
• Add example/lib/gemini_main.dart — runnable Gemini chat demo.
• Add GitHub Actions CI (.github/workflows/ci.yml) running flutter analyze and flutter test on push and pull request.
• Refactor request building: extract _buildUri() and _buildPayload() in GenuiXTransport to keep provider-specific logic isolated. The Anthropic and OpenAI request shapes are unchanged; this is a non-breaking internal cleanup that supports the new Gemini path.
• Internal rename pass — no public API change. lib/src/claude_transport.dart → genui_x_transport.dart, lib/src/claude_config.dart → genui_x_config.dart, lib/src/sse_parser.dart → anthropic_sse_parser.dart (class ClaudeSseParser → AnthropicSseParser). Internal _streamClaude() / _toClaudeMessage() / _sseParser are now _streamLlm() / _toMessage() / _anthropicSseParser. Test files renamed to match. AGENTS.md docs updated. The public surface (GenuiXTransport, GenuiXConfig, GenuiXStreamFormat, errors) is unchanged.

0.0.10

• Add GenuiXTransport.anthropic() factory constructor — mirrors .openai() with explicit Anthropic defaults (x-api-key header, /v1/messages endpoint, Anthropic SSE format).
• Add automatic retry on 429 responses with exponential backoff — configurable via maxRetries (default 3). Respects Retry-After header when present.

0.0.9

• Add GenuiXTransport.openai() factory constructor — pre-configures Authorization: Bearer header, /v1/chat/completions endpoint, and OpenAI SSE format. Works with OpenAI, OpenRouter, LiteLLM, and any OpenAI-compatible proxy.
• Add surfaceOperations (SurfaceOperations?) to GenuiXTransport and GenuiXConfig — use PromptBuilder.custom() to control which A2UI operations (create, update, delete) the AI may perform. Defaults to PromptBuilder.chat() (create-only) when not set.
• Add clientDataModel (Map<String, Object?>?) to GenuiXTransport and GenuiXConfig — inject app-state context (user profile, active session, domain data) into the system prompt.
• Add GenuiXRateLimitError — thrown on 429 responses. Exposes retryAfter: int? parsed from the Retry-After header.
• Re-export PromptFragments and SurfaceOperations from genui_x.dart — users no longer need a separate import 'package:genui/genui.dart' to configure the transport.
• Remove Flutter upper-bound constraint (<4.0.0) per pub.dev deprecation guidance.
• Overhaul README — add genui/genui_x architecture diagram, backend examples, surfaceOperations and clientDataModel usage.

0.0.8

• Improve doc comments on all public APIs — GenuiXTransport, GenuiXConfig, GenuiXStreamFormat, GenuiXAuthError, GenuiXApiError.
• Add example/lib/proxy_main.dart — runnable example for OpenAI-compatible backends (LiteLLM, OpenRouter, custom proxy).
• Update example/README.md to document all three entry points.

0.0.7

• Rename ClaudeTransport → GenuiXTransport, ClaudeConfig → GenuiXConfig, ClaudeStreamFormat → GenuiXStreamFormat — names no longer imply Claude-only.
• Rename ClaudeAuthException → GenuiXAuthError and ClaudeApiException → GenuiXApiError for provider-neutral naming.
• Add cancel() to ClaudeTransport — abort an in-flight request and reset isLoading.
• Add clearHistory() to ClaudeTransport — reset conversation history without creating a new transport.
• Add debug: bool to ClaudeConfig and ClaudeTransport — prints request URL, model, status code, and errors via debugPrint.
• Add transport-level tests: error handling, streaming (Anthropic + OpenAI), isLoading state, cancel(), clearHistory().

0.0.5

• Add systemPromptFragments to ClaudeTransport and ClaudeConfig — inject persona, domain restrictions, or date context into the system prompt.
• Add isLoading (ValueNotifier<bool>) to ClaudeTransport — drive loading indicators without manually tracking conversation state.
• Tighten SDK constraint to >=3.10.0 <4.0.0 and Flutter to >=3.35.7 <4.0.0 to match genui 0.8.0 requirements.

0.0.4

• Rename main library export to genui_x.dart — import with package:genui_x/genui_x.dart.
• Add requestBodyOverrides to ClaudeTransport and ClaudeConfig for provider-specific JSON-mode support.
• Add travel demo example app and minimal example app.
• Fix baseUrl resolution in example apps.

0.0.3

• Add OpenAI-style stream format support and parser.
• Allow selecting stream format via ClaudeConfig.
• Document OpenAI-style proxy configuration.
• Add request body overrides for JSON-mode responses.

0.0.2

• Add configurable endpoint path and request headers for proxies/LiteLLM.
• Document proxy usage and authorization header support.

0.0.1

• Initial release of the Claude transport adapter for genui.
• Includes ClaudeTransport, configuration, and SSE parsing.
• Adds a Flutter example app and basic tests.