thurakhant · thurakhant · Jun 7, 2026 · Apr 19, 2026 · Apr 22, 2026 · Apr 22, 2026
diff --git a/.pubignore b/.pubignore
@@ -20,6 +20,8 @@ AGENTS.md
 
 # Dev-only docs
 doc/RELEASE_CHECKLIST.md
+doc/superpowers/
+docs/
 
 # Environment and secrets
 .env

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,3 +1,32 @@
+## 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-define`s.
+* 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.
+* Upgrade dependency to `genui: ^0.9.2` in both the package and example app.
+* Upgrade transport dependency to `http: ^1.6.0`.
+* Refresh docs and agent notes to reference `genui ^0.9.2` instead of `^0.8.0`.
+* Exclude dev-only docs (`doc/superpowers/`, `docs/`) from pub publish via
+  `.pubignore` so package publish surface stays clean.
+* Validate compatibility with `flutter analyze --fatal-infos` and `flutter test`
+  in both the root package and `example/` app.
+* 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

diff --git a/README.md b/README.md
@@ -3,10 +3,38 @@
 [![pub package](https://img.shields.io/pub/v/genui_x.svg)](https://pub.dev/packages/genui_x)
 [![pub points](https://img.shields.io/pub/points/genui_x.svg)](https://pub.dev/packages/genui_x/score)
 [![likes](https://img.shields.io/pub/likes/genui_x.svg)](https://pub.dev/packages/genui_x/score)
+[![CI](https://github.com/thurakhant/genui_x/actions/workflows/ci.yml/badge.svg)](https://github.com/thurakhant/genui_x/actions/workflows/ci.yml)
 
-A lightweight AI backend adapter for Google's [genui](https://pub.dev/packages/genui) (Generative UI) framework.
+> **The missing backend for [Google's genui](https://pub.dev/packages/genui) framework.** One `Transport`, every major LLM — Anthropic Claude, OpenAI, Google Gemini, local Ollama, or any OpenAI-compatible proxy.
 
-Connect any AI backend — Anthropic Claude, OpenAI, OpenRouter, LiteLLM, or your own proxy — to genui with a single class.
+<!--
+  Demo GIF placeholder — record ~15 s of `example/lib/travel_main.dart`
+  (prompt → streamed widget) and drop the file at `screenshots/demo.gif`,
+  then uncomment the line below.
+
+![genui_x demo](screenshots/demo.gif)
+-->
+
+```dart
+final transport = GenuiXTransport.anthropic(  // or .openai() / .gemini() / .ollama()
+  apiKey: const String.fromEnvironment('CLAUDE_API_KEY'),
+  catalog: myCatalog,
+);
+```
+
+`genui` defines *what* the AI can render. `genui_x` connects it to *who* generates it.
+
+## Provider matrix
+
+| Provider | Factory | Streaming | JSON mode | Auth header | Default model |
+|---|---|:-:|:-:|---|---|
+| Anthropic Claude | `GenuiXTransport.anthropic()` | ✅ | n/a (Claude is JSON-friendly) | `x-api-key` | `claude-haiku-4-5-20251001` |
+| OpenAI / OpenRouter / LiteLLM | `GenuiXTransport.openai()` | ✅ | ✅ via `enforceJsonMode` | `Authorization: Bearer` | `gpt-4o-mini` |
+| Google Gemini | `GenuiXTransport.gemini()` | ✅ | n/a (uses `systemInstruction`) | `x-goog-api-key` | `gemini-2.5-flash` |
+| Ollama (local) | `GenuiXTransport.ollama()` | ✅ | ✅ via `enforceJsonMode` | `Authorization: Bearer` (placeholder) | `llama3.2` |
+| Custom proxy | `GenuiXTransport(...)` | ✅ | via `requestBodyOverrides` | configurable | configurable |
+
+All providers share the same retry-on-429, cancel, `clearHistory`, `isLoading`, and `debug` controls.
 
 ---
 
@@ -37,8 +65,8 @@ Your App
 ```yaml
 # pubspec.yaml
 dependencies:
-  genui: ^0.8.0
-  genui_x: ^0.0.11
+  genui: ^0.9.2
+  genui_x: ^0.0.12
 ```
 
 ### 2. Create your catalog
@@ -152,6 +180,22 @@ Sends `x-goog-api-key`, posts to
 `candidates[*].content.parts[*].text` SSE stream. Override `baseUrl` to
 point at a Vertex AI gateway or your own proxy.
 
+### Ollama (local models)
+
+```dart
+final transport = GenuiXTransport.ollama(
+  catalog: myCatalog,
+  // model: 'llama3.2',                  // optional — default
+  // baseUrl: 'http://localhost:11434',  // optional — default
+  // enforceJsonMode: true,              // recommended for stricter A2UI output
+);
+```
+
+Wraps `GenuiXTransport.openai()` against Ollama's OpenAI-compatible
+endpoint. No API key is required by Ollama, but a placeholder bearer token
+(`ollama`) is sent so the OpenAI-compat layer accepts the request. Pull
+the model first with `ollama pull llama3.2`.
+
 ### OpenRouter / LiteLLM / custom proxy
 
 ```dart
@@ -264,6 +308,8 @@ final transport = GenuiXTransport(
 | OpenAI | `gpt-4o` | Higher quality |
 | Gemini | `gemini-2.5-flash` | Default for `.gemini()` — fast, low cost |
 | Gemini | `gemini-2.5-pro` | Higher quality |
+| Ollama | `llama3.2` | Default for `.ollama()` — local |
+| Ollama | any local model slug | e.g. `qwen2.5-coder`, `mistral` |
 | OpenRouter | any model slug | via `GenuiXTransport.openai(baseUrl: ...)` |
 
 ---
@@ -289,6 +335,13 @@ flutter run -t lib/proxy_main.dart \
 # Google Gemini
 flutter run -t lib/gemini_main.dart \
   --dart-define=GEMINI_API_KEY=your-google-api-key
+
+# Ollama (local — no API key)
+flutter run -t lib/ollama_main.dart
+# or with overrides:
+flutter run -t lib/ollama_main.dart \
+  --dart-define=OLLAMA_BASE_URL=http://192.168.1.50:11434 \
+  --dart-define=OLLAMA_MODEL=qwen2.5-coder
 ```
 
 ---

diff --git a/example/README.md b/example/README.md
@@ -1,6 +1,6 @@
 # genui_x examples
 
-Three runnable entry points — pick the one that matches your setup.
+Five runnable entry points — pick the one that matches your setup.
 
 ---
 
@@ -36,3 +36,35 @@ flutter run -t lib/proxy_main.dart \
 ```
 
 Shows `GenuiXStreamFormat.openai`, `authorization` header, and `Bearer ` prefix.
+
+---
+
+## Gemini — Google's Generative Language API
+
+```bash
+flutter run -t lib/gemini_main.dart \
+  --dart-define=GEMINI_API_KEY=your-google-api-key
+```
+
+Uses `GenuiXTransport.gemini()`, the `x-goog-api-key` header, and the
+`streamGenerateContent?alt=sse` endpoint.
+
+---
+
+## Ollama — local model, no API key
+
+```bash
+# 1. Pull a model first (one-time):
+ollama pull llama3.2
+
+# 2. Run the example:
+flutter run -t lib/ollama_main.dart
+
+# Optional overrides:
+flutter run -t lib/ollama_main.dart \
+  --dart-define=OLLAMA_BASE_URL=http://192.168.1.50:11434 \
+  --dart-define=OLLAMA_MODEL=qwen2.5-coder
+```
+
+Uses `GenuiXTransport.ollama()` against Ollama's OpenAI-compatible
+endpoint with `enforceJsonMode: true` for stricter A2UI output.
diff --git a/example/lib/AGENTS.md b/example/lib/AGENTS.md
@@ -4,7 +4,7 @@
 # example/lib/
 
 ## Purpose
-Four Flutter entry points demonstrating different complexity levels and backend choices for `genui_x` integration. Each file is a complete runnable app including its own `Catalog`, `CatalogItem`, `widgetBuilder`, and chat UI.
+Five Flutter entry points demonstrating different complexity levels and backend choices for `genui_x` integration. Each file is a complete runnable app including its own `Catalog`, `CatalogItem`, `widgetBuilder`, and chat UI.
 
 ## Key Files
 
@@ -14,6 +14,8 @@ Four Flutter entry points demonstrating different complexity levels and backend
 | `main.dart` | Full-featured demo — `WeatherWidget` catalog with proxy/model overrides via `--dart-define`, text input chat |
 | `travel_main.dart` | Travel demo — custom `TravelPlanWidget` catalog, shows `data`-wrapped component properties and list props |
 | `gemini_main.dart` | Gemini-backend demo — uses `GenuiXTransport.gemini()` against the Generative Language API |
+| `ollama_main.dart` | Local Ollama demo — uses `GenuiXTransport.ollama()`, no API key required, `enforceJsonMode: true` |
+| `proxy_main.dart` | Generic OpenAI-compatible proxy demo (LiteLLM, OpenRouter, etc.) configured via `--dart-define` |
 
 ## For AI Agents
 
@@ -60,7 +62,7 @@ cd example && flutter test      # widget smoke tests
 - `../AGENTS.md` — example-wide conventions
 
 ### External
-- `package:genui_x` — `GenuiXTransport` (with `.openai()`, `.anthropic()`, `.gemini()` factories), `GenuiXConfig`, `GenuiXStreamFormat`
+- `package:genui_x` — `GenuiXTransport` (with `.openai()`, `.anthropic()`, `.gemini()`, `.ollama()` factories), `GenuiXConfig`, `GenuiXStreamFormat`
 - `package:genui` — `Catalog`, `CatalogItem`, `SurfaceController`, `Conversation`, `Surface`, `ChatMessage`, `PromptBuilder`
 - `package:json_schema_builder` — `S.object()`, `S.string()`, etc.