ai-sdk-ollama 0.3.0 → 0.5.0

package/CHANGELOG.md

@@ -1,5 +1,46 @@
 # Changelog
 
+## 0.5.0
+
+### Minor Changes
+
+- Add browser support with automatic environment detection
+  - Add browser-specific provider using ollama/browser package
+  - Implement dual package exports for Node.js and browser environments
+  - Add comprehensive browser example with Vite and interactive UI
+  - Update build configuration to generate separate browser and Node.js bundles
+  - Add browser compatibility tests and examples
+
+- e02f8af: Feature: Browser support and streaming improvements. Closes https://github.com/jagreehal/ai-sdk-ollama/issues/14
+  - feat(browser): Automatic browser support via `ollama/browser` with dedicated `index.browser` export. Works seamlessly with bundlers and `ai` v5 in browser contexts.
+  - fix(streaming): Emit trailing `text-delta` on the final `done` chunk to avoid empty streams for models that only flush content at the end. Note: streams may include one additional text chunk now.
+  - tests: Add `gpt-oss:20b` integration coverage and make prompts/token limits more robust; update unit tests to reflect final text emission on `done`.
+  - docs/examples: Switch Node examples to per-file `tsx` execution and update READMEs; clarify how to run browser and node examples.
+  - chore(repo): Monorepo migration (no user-facing API changes), Dependabot config for package folder, and CI refinements.
+
+  No breaking changes to the public API.
+
+## 0.4.0
+
+### Minor Changes
+
+- **Reasoning Support**: Added support for reasoning (chain-of-thought) output
+  - 🧠 **Reasoning Content**: Models that support reasoning can now output their thinking process
+  - 📝 **Content Types**: Support for `LanguageModelV2Reasoning` content type in both non-streaming and streaming responses
+  - 🔄 **Streaming Support**: Full streaming support with `reasoning-start`, `reasoning-delta`, and `reasoning-end` events
+  - ⚙️ **Configurable**: Enable reasoning with `{ reasoning: true }` setting
+  - 🧪 **Comprehensive Testing**: Added unit tests for reasoning functionality
+  - 📚 **Documentation**: Updated README and examples with reasoning usage
+  - 🎯 **Backward Compatible**: Reasoning is disabled by default, existing code continues to work
+
+### Technical Improvements
+
+- Added `reasoning` setting to `OllamaChatSettings` interface
+- Enhanced `doGenerate` method to handle `thinking` field from Ollama responses
+- Enhanced `doStream` method to emit reasoning stream parts
+- Added reasoning support to content conversion logic
+- Updated type definitions to include reasoning content types
+
 ## 0.3.0
 
 ### Minor Changes

package/README.md

@@ -15,6 +15,10 @@ A Vercel AI SDK v5+ provider for Ollama built on the official `ollama` package.
   - [Quick Start](#quick-start)
     - [Installation](#installation)
     - [Basic Usage](#basic-usage)
+  - [Browser Support](#browser-support)
+    - [Browser Usage](#browser-usage)
+    - [Explicit Browser Import](#explicit-browser-import)
+    - [CORS Configuration](#cors-configuration)
   - [Key Features](#key-features)
     - [Cross Provider Compatibility](#cross-provider-compatibility)
     - [Native Ollama Power](#native-ollama-power)
@@ -24,6 +28,7 @@ A Vercel AI SDK v5+ provider for Ollama built on the official `ollama` package.
     - [Custom Ollama Instance](#custom-ollama-instance)
     - [Structured Output](#structured-output)
     - [Auto-Detection of Structured Outputs](#auto-detection-of-structured-outputs)
+    - [Reasoning Support](#reasoning-support)
   - [Common Issues](#common-issues)
   - [Supported Models](#supported-models)
   - [Testing](#testing)
@@ -114,6 +119,52 @@ const { embedding } = await embed({
 console.log(embedding.length); // 768 dimensions
 ```
 
+## Browser Support
+
+See the [browser example](../../examples/browser/).
+
+This provider works in both Node.js and browser environments. The library automatically selects the correct Ollama client based on the environment.
+
+### Browser Usage
+
+The same API works in browsers with automatic environment detection:
+
+```typescript
+import { ollama } from 'ai-sdk-ollama'; // Automatically uses browser version
+import { generateText } from 'ai';
+
+const { text } = await generateText({
+  model: ollama('llama3.2'),
+  prompt: 'Write a haiku about coding',
+});
+```
+
+### Explicit Browser Import
+
+You can also explicitly import the browser version:
+
+```typescript
+import { ollama } from 'ai-sdk-ollama/browser';
+```
+
+### CORS Configuration
+
+For browser usage, you have several options to handle CORS:
+
+```bash
+# Option 1: Use a proxy (recommended for development)
+# Configure your bundler (Vite, Webpack, etc.) to proxy /api/* to Ollama
+# See browser example for Vite proxy configuration
+
+# Option 2: Allow all origins (development only)
+OLLAMA_ORIGINS=* ollama serve
+
+# Option 3: Allow specific origins
+OLLAMA_ORIGINS="http://localhost:3000,https://myapp.com" ollama serve
+```
+
+**Recommended**: Use a development proxy (like Vite proxy) to avoid CORS issues entirely. See the browser example for a complete working setup.
+
 ## Key Features
 
 ### Cross Provider Compatibility
@@ -274,6 +325,48 @@ const { text } = await generateText({
 });
 ```
 
+### Reasoning Support
+
+Some models like DeepSeek-R1 support reasoning (chain-of-thought) output. Enable this feature to see the model's thinking process:
+
+```typescript
+// Enable reasoning for models that support it (e.g., deepseek-r1:7b)
+const model = ollama('deepseek-r1:7b', { reasoning: true });
+
+// Generate text with reasoning
+const { text } = await generateText({
+  model,
+  prompt:
+    'Solve: If I have 3 boxes, each with 4 smaller boxes, and each smaller box has 5 items, how many items total?',
+});
+
+console.log('Answer:', text);
+// DeepSeek-R1 includes reasoning in the output with <think> tags:
+// <think>
+// First, I'll calculate the number of smaller boxes: 3 × 4 = 12
+// Then, the total items: 12 × 5 = 60
+// </think>
+// You have 60 items in total.
+
+// Compare with reasoning disabled
+const modelNoReasoning = ollama('deepseek-r1:7b', { reasoning: false });
+const { text: noReasoningText } = await generateText({
+  model: modelNoReasoning,
+  prompt: 'Calculate 3 × 4 × 5',
+});
+// Output: 60 (without showing the thinking process)
+```
+
+**Recommended Reasoning Models**:
+
+- `deepseek-r1:7b` - Balanced performance and reasoning capability (5GB)
+- `deepseek-r1:1.5b` - Lightweight option (2.5GB)
+- `deepseek-r1:8b` - Llama-based distilled version (5.5GB)
+
+Install with: `ollama pull deepseek-r1:7b`
+
+**Note**: The reasoning feature is model-dependent. Models without reasoning support will work normally without showing thinking process.
+
 ## Common Issues
 
 - **Make sure Ollama is running** - Run `ollama serve` before using the provider
@@ -287,9 +380,10 @@ const { text } = await generateText({
 
 Works with any model in your Ollama installation:
 
-- **Chat**: `llama3.2`, `mistral`, `phi4-mini`, `qwen2.5`, `codellama`
+- **Chat**: `llama3.2`, `mistral`, `phi4-mini`, `qwen2.5`, `codellama`, `gpt-oss:20b`
 - **Vision**: `llava`, `bakllava`, `llama3.2-vision`, `minicpm-v`
 - **Embeddings**: `nomic-embed-text`, `all-minilm`, `mxbai-embed-large`
+- **Reasoning**: `deepseek-r1:7b`, `deepseek-r1:1.5b`, `deepseek-r1:8b`
 
 ## Testing