@tanstack/ai 0.52.3 → 0.54.0

package/dist/esm/types.d.ts

@@ -1865,6 +1865,118 @@ export interface VideoUrlResult {
     /** Persisted artifact references for generated assets, when available */
     artifacts?: Array<PersistedArtifactRef>;
 }
+/**
+ * Options for world generation (live, prompt-steerable sessions).
+ *
+ * @experimental World generation is an experimental feature and may change.
+ */
+export interface WorldGenerationOptions<TProviderOptions extends object = object> {
+    /** The model to use for world generation */
+    model: string;
+    /** Natural-language description of the world or scene */
+    prompt: string;
+    /**
+     * Provider mint options. Reactor resolution/seed/audio are browser
+     * `sendCommand` fields, not token-mint fields.
+     */
+    modelOptions?: TProviderOptions;
+    /**
+     * Internal logger threaded from the generateWorld() entry point. Adapters
+     * must call logger.request() before the SDK call and logger.errors() in
+     * catch blocks.
+     */
+    logger: InternalLogger;
+    /**
+     * Effective abort signal composed by the activity from caller `abortSignal`
+     * and/or `timeout`. Adapters should forward this to the provider SDK when
+     * supported. Request-specific — never store on a global client config.
+     */
+    abortSignal?: AbortSignal;
+}
+/**
+ * Result of world generation. JSON-serializable so a server route can return
+ * it to a browser. The browser uses `token` + `model` to open the live
+ * session (set the prompt, start streaming, steer mid-run).
+ *
+ * @experimental World generation is an experimental feature and may change.
+ */
+export interface WorldGenerationResult {
+    /** Unique identifier for this generation */
+    id: string;
+    /** Model used for generation (provider connect slug) */
+    model: string;
+    /** Short-lived session token for the client connection */
+    token: string;
+    /** Token expiry as milliseconds since epoch */
+    expiresAt: number;
+    /** Prompt the client should send when it starts the session */
+    prompt: string;
+    /** Session status after the server half finishes */
+    status: 'ready' | 'waiting';
+    /** Provider session id, when the adapter created one */
+    sessionId?: string;
+    /** Token usage / billing, when the adapter can report it */
+    usage?: TokenUsage;
+}
+/**
+ * Options for live generation (prompt-steerable video sessions).
+ *
+ * @experimental Live generation is an experimental feature and may change.
+ */
+export interface LiveVideoGenerationOptions<TProviderOptions extends object = object> {
+    /** The model to use for live generation */
+    model: string;
+    /** Natural-language description of the shot or scene */
+    prompt: string;
+    /**
+     * Provider mint options. For fal live this is `tokenDuration`. Reactor
+     * resolution/seed/audio are browser `sendCommand` fields.
+     */
+    modelOptions?: TProviderOptions;
+    /**
+     * Internal logger threaded from the generateLiveVideo() entry point. Adapters
+     * must call logger.request() before the SDK call and logger.errors() in
+     * catch blocks.
+     */
+    logger: InternalLogger;
+    /**
+     * Effective abort signal composed by the activity from caller `abortSignal`
+     * and/or `timeout`. Adapters should forward this to the provider SDK when
+     * supported. Request-specific — never store on a global client config.
+     */
+    abortSignal?: AbortSignal;
+}
+/**
+ * Result of live generation. JSON-serializable so a server route can return
+ * it to a browser.
+ *
+ * Reactor: connect with `token` and `model` (the connect slug).
+ * fal: `model` is the WMA app id. Open `wma(model)` through a server proxy
+ * that attaches `FAL_KEY`. Do not send `token` as Key credentials.
+ *
+ * @experimental Live generation is an experimental feature and may change.
+ */
+export interface LiveVideoGenerationResult {
+    /** Unique identifier for this generation */
+    id: string;
+    /**
+     * Connect id for the browser client. Reactor: `reactor/helios`.
+     * fal: WMA app id `fal-ai/minimax-h3-max-director`.
+     */
+    model: string;
+    /** Short-lived session token. Reactor uses this to connect. fal does not. */
+    token: string;
+    /** Token expiry as milliseconds since epoch */
+    expiresAt: number;
+    /** Prompt the client should send when it starts the session */
+    prompt: string;
+    /** Session status after the server half finishes */
+    status: 'ready' | 'waiting';
+    /** Provider session id, when the adapter created one */
+    sessionId?: string;
+    /** Token usage / billing, when the adapter can report it */
+    usage?: TokenUsage;
+}
 /**
  * Options for text-to-speech generation.
  * These are the common options supported across providers.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tanstack/ai",
-  "version": "0.52.3",
+  "version": "0.54.0",
   "description": "Type-safe TypeScript AI SDK for streaming chat, tool calling, agents, structured outputs, and multimodal generation.",
   "author": "Tanner Linsley",
   "license": "MIT",
@@ -88,7 +88,7 @@
     "@ag-ui/core": "0.1.1-canary.beta.0",
     "@standard-schema/spec": "^1.1.0",
     "partial-json": "^0.1.7",
-    "@tanstack/ai-event-client": "^0.11.2",
+    "@tanstack/ai-event-client": "^0.11.3",
     "@tanstack/ai-utils": "^0.4.0"
   },
   "peerDependencies": {

package/skills/ai-core/adapter-configuration/SKILL.md

@@ -45,16 +45,20 @@ Create an adapter and use it with `chat()`:
 import { chat, toServerSentEventsResponse } from '@tanstack/ai'
 import { openaiText } from '@tanstack/ai-openai'
 
-const stream = chat({
-  adapter: openaiText('gpt-5.2'),
-  messages,
-  modelOptions: {
-    temperature: 0.7,
-    max_output_tokens: 1000,
-  },
-})
+export async function POST(request: Request) {
+  const { messages } = await request.json()
+
+  const stream = chat({
+    adapter: openaiText('gpt-5.2'),
+    messages,
+    modelOptions: {
+      temperature: 0.7,
+      max_output_tokens: 1000,
+    },
+  })
 
-return toServerSentEventsResponse(stream)
+  return toServerSentEventsResponse(stream)
+}
 ```
 
 The adapter factory function takes the model name as a string literal and an
@@ -73,18 +77,19 @@ top-level options on `chat()`. See the per-provider table in
 Each provider has a dedicated package with tree-shakeable adapter factories.
 The text adapter is the primary one for chat/completions:
 
-| Provider          | Package                          | Factory                                     | Env Var                                           |
-| ----------------- | -------------------------------- | ------------------------------------------- | ------------------------------------------------- |
-| OpenAI            | `@tanstack/ai-openai`            | `openaiText`                                | `OPENAI_API_KEY`                                  |
-| Anthropic         | `@tanstack/ai-anthropic`         | `anthropicText`                             | `ANTHROPIC_API_KEY`                               |
-| Gemini            | `@tanstack/ai-gemini`            | `geminiText`                                | `GOOGLE_API_KEY` or `GEMINI_API_KEY`              |
-| Grok (xAI)        | `@tanstack/ai-grok`              | `grokText`                                  | `XAI_API_KEY`                                     |
-| Groq              | `@tanstack/ai-groq`              | `groqText`                                  | `GROQ_API_KEY`                                    |
-| OpenRouter        | `@tanstack/ai-openrouter`        | `openRouterText`                            | `OPENROUTER_API_KEY`                              |
-| Ollama            | `@tanstack/ai-ollama`            | `ollamaText`                                | `OLLAMA_HOST` (default: `http://localhost:11434`) |
-| Bedrock           | `@tanstack/ai-bedrock`           | `bedrockText`                               | `BEDROCK_API_KEY` or `AWS_BEARER_TOKEN_BEDROCK`   |
-| BytePlus          | `@tanstack/ai-byteplus`          | `byteplusText`                              | `ARK_API_KEY` (falls back to `BYTEPLUS_API_KEY`)  |
-| OpenAI-compatible | `@tanstack/ai-openai/compatible` | `openaiCompatible` / `openaiCompatibleText` | provider-specific (passed via `apiKey`)           |
+| Provider          | Package                          | Factory                                     | Env Var                                                                                |
+| ----------------- | -------------------------------- | ------------------------------------------- | -------------------------------------------------------------------------------------- |
+| OpenAI            | `@tanstack/ai-openai`            | `openaiText`                                | `OPENAI_API_KEY`                                                                       |
+| Anthropic         | `@tanstack/ai-anthropic`         | `anthropicText`                             | `ANTHROPIC_API_KEY`                                                                    |
+| Gemini            | `@tanstack/ai-gemini`            | `geminiText`                                | `GOOGLE_API_KEY` or `GEMINI_API_KEY`                                                   |
+| Grok (xAI)        | `@tanstack/ai-grok`              | `grokText`                                  | `XAI_API_KEY`                                                                          |
+| Groq              | `@tanstack/ai-groq`              | `groqText`                                  | `GROQ_API_KEY`                                                                         |
+| OpenRouter        | `@tanstack/ai-openrouter`        | `openRouterText`                            | `OPENROUTER_API_KEY`                                                                   |
+| Ollama            | `@tanstack/ai-ollama`            | `ollamaText`                                | `OLLAMA_HOST` (default: `http://localhost:11434`)                                      |
+| Bedrock           | `@tanstack/ai-bedrock`           | `bedrockText`                               | `BEDROCK_API_KEY` or `AWS_BEARER_TOKEN_BEDROCK`                                        |
+| BytePlus          | `@tanstack/ai-byteplus`          | `byteplusText`                              | `ARK_API_KEY` (falls back to `BYTEPLUS_API_KEY`)                                       |
+| OpenAI-compatible | `@tanstack/ai-openai/compatible` | `openaiCompatible` / `openaiCompatibleText` | provider-specific (passed via `apiKey`)                                                |
+| Cloudflare        | `@tanstack/ai-cloudflare`        | `cloudflareText`                            | `CLOUDFLARE_ACCOUNT_ID` + `CLOUDFLARE_API_TOKEN`, or `{ binding: env.AI }` in a Worker |
 
 > **BytePlus uses two keys.** `byteplusText` / `byteplusVideo` /
 > `byteplusImage` read `ARK_API_KEY` (ModelArk, `Authorization: Bearer`), but
@@ -94,7 +99,7 @@ The text adapter is the primary one for chat/completions:
 
 ```typescript
 // Each factory takes model as first arg, optional config as second
-import { openaiText } from '@tanstack/ai-openai'
+import { openaiText, createOpenaiChat } from '@tanstack/ai-openai'
 import { anthropicText } from '@tanstack/ai-anthropic'
 import { geminiText } from '@tanstack/ai-gemini'
 import { grokText } from '@tanstack/ai-grok'
@@ -108,17 +113,16 @@ import { byteplusText } from '@tanstack/ai-byteplus'
 const adapter = openaiText('gpt-5.2')
 const adapter2 = anthropicText('claude-sonnet-4-6')
 const adapter3 = geminiText('gemini-2.5-pro')
-const adapter4 = grokText('grok-4')
+const adapter4 = grokText('grok-4.6')
 const adapter5 = groqText('llama-3.3-70b-versatile')
 const adapter6 = openRouterText('anthropic/claude-sonnet-4')
-const adapter7 = ollamaText('llama3.3')
+const adapter7 = ollamaText('llama3.3:latest')
 const adapter8 = bedrockText('us.anthropic.claude-3-7-sonnet-20250219-v1:0')
 const adapter9 = byteplusText('seed-2-0-lite-260428')
 
-// Optional: pass explicit API key
-const adapterWithKey = openaiText('gpt-5.2', {
-  apiKey: 'sk-...',
-})
+// Optional: pass an explicit API key via the create* sibling
+// (the plain factory reads it from the environment)
+const adapterWithKey = createOpenaiChat('gpt-5.2', 'sk-...')
 ```
 
 `@tanstack/ai-bedrock` (Amazon Bedrock) branches on `config.api`:
@@ -136,26 +140,32 @@ input or configuration:
 
 ```typescript
 import { chat, toServerSentEventsResponse } from '@tanstack/ai'
-import type { TextAdapter } from '@tanstack/ai/adapters'
+import type { ModelMessage } from '@tanstack/ai'
 import { openaiText } from '@tanstack/ai-openai'
 import { anthropicText } from '@tanstack/ai-anthropic'
 import { geminiText } from '@tanstack/ai-gemini'
 
 // Define a map of provider+model to adapter factory calls
-const adapters: Record<string, () => TextAdapter> = {
+const adapters = {
   'openai/gpt-5.2': () => openaiText('gpt-5.2'),
   'anthropic/claude-sonnet-4-6': () => anthropicText('claude-sonnet-4-6'),
   'gemini/gemini-2.5-pro': () => geminiText('gemini-2.5-pro'),
 }
 
-export function handleChat(providerModel: string, messages: Array<any>) {
-  const createAdapter = adapters[providerModel]
-  if (!createAdapter) {
+function isKnownProviderModel(key: string): key is keyof typeof adapters {
+  return key in adapters
+}
+
+export function handleChat(
+  providerModel: string,
+  messages: Array<ModelMessage>,
+) {
+  if (!isKnownProviderModel(providerModel)) {
     throw new Error(`Unknown provider/model: ${providerModel}`)
   }
 
   const stream = chat({
-    adapter: createAdapter(),
+    adapter: adapters[providerModel](),
     messages,
   })
 
@@ -173,6 +183,10 @@ import { openaiText } from '@tanstack/ai-openai'
 import { anthropicText } from '@tanstack/ai-anthropic'
 import { geminiText } from '@tanstack/ai-gemini'
 
+const messages = [
+  { role: 'user' as const, content: 'Plan a database migration.' },
+]
+
 // OpenAI: reasoning with effort and summary
 const openaiStream = chat({
   adapter: openaiText('gpt-5.2'),
@@ -198,16 +212,18 @@ const anthropicStream = chat({
   },
 })
 
-// Anthropic: adaptive thinking (claude-sonnet-4-6 and newer)
+// Anthropic: adaptive thinking (Sonnet 5, Fable 5, Opus 4.7+) — depth is
+// tuned with output_config.effort instead of a token budget
 const adaptiveStream = chat({
-  adapter: anthropicText('claude-sonnet-4-6'),
+  adapter: anthropicText('claude-sonnet-5'),
   messages,
   modelOptions: {
     max_tokens: 16000,
     thinking: {
       type: 'adaptive',
+      display: 'summarized', // stream the reasoning text (default 'omitted')
     },
-    effort: 'high', // 'max' | 'high' | 'medium' | 'low'
+    output_config: { effort: 'high' }, // 'low' | 'medium' | 'high' | 'xhigh' | 'max'
   },
 })
 
@@ -262,6 +278,14 @@ inside `modelOptions` using each provider's **native** key. They are not
 top-level fields on `chat()`/`ai()`/`generate()`.
 
 ```typescript
+import { chat } from '@tanstack/ai'
+import { openaiText } from '@tanstack/ai-openai'
+import { anthropicText } from '@tanstack/ai-anthropic'
+import { geminiText } from '@tanstack/ai-gemini'
+import { ollamaText } from '@tanstack/ai-ollama'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
+
 // OpenAI — native keys
 chat({
   adapter: openaiText('gpt-5.2'),
@@ -284,8 +308,9 @@ chat({
 })
 
 // Ollama — NESTED under modelOptions.options
+// (use the `family:tag` id — a bare `llama3.3` falls back to untyped options)
 chat({
-  adapter: ollamaText('llama3.3'),
+  adapter: ollamaText('llama3.3:latest'),
   messages,
   modelOptions: {
     options: { temperature: 0.7, top_p: 0.9, num_predict: 1000 },
@@ -300,7 +325,7 @@ Per-provider sampling keys (all live inside `modelOptions`):
 | OpenAI            | `temperature` | `top_p` | `max_output_tokens`                       |
 | Anthropic         | `temperature` | `top_p` | `max_tokens`                              |
 | Gemini            | `temperature` | `topP`  | `maxOutputTokens`                         |
-| Grok (xAI)        | `temperature` | `top_p` | `max_tokens`                              |
+| Grok (xAI)        | `temperature` | `top_p` | `max_output_tokens`                       |
 | Groq              | `temperature` | `top_p` | `max_completion_tokens`                   |
 | OpenRouter (chat) | `temperature` | `topP`  | `maxCompletionTokens`                     |
 | Ollama            | `temperature` | `top_p` | `num_predict` (nested in `options`)       |
@@ -325,7 +350,16 @@ some sampling options use provider-native names. Ollama nests all sampling under
 Adapters can declare an optional capability method:
 
 ```ts
-supportsCombinedToolsAndSchema?(modelOptions?: TProviderOptions): boolean
+import { AnthropicTextAdapter } from '@tanstack/ai-anthropic'
+
+// The TextAdapter contract:
+//   supportsCombinedToolsAndSchema?: (modelOptions?: TProviderOptions) => boolean
+// Subclasses override it to narrow the capability:
+class LegacyPathAnthropic extends AnthropicTextAdapter<'claude-sonnet-4-6'> {
+  override supportsCombinedToolsAndSchema(): boolean {
+    return false
+  }
+}
 ```
 
 When `true`, the engine wires `outputSchema` into the regular
@@ -337,16 +371,16 @@ runs.
 
 Current per-adapter status (#605):
 
-| Adapter                                      | Returns                                                                                               |
-| -------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
-| `openaiText` / `openaiChatCompletions`       | `true` (all supported models)                                                                         |
-| `anthropicText`                              | `true` for Claude 4.5+ (gated by `ANTHROPIC_COMBINED_TOOLS_AND_SCHEMA_MODELS`), `false` otherwise     |
-| `geminiText`                                 | `true` for Gemini 3.x (gated by `GEMINI_COMBINED_TOOLS_AND_SCHEMA_MODELS`), `false` otherwise         |
-| `grokText`                                   | `true` for Grok 4 family (gated by `GROK_COMBINED_TOOLS_AND_SCHEMA_MODELS`), `false` otherwise        |
-| `groqText`                                   | `false` (Groq API rejects schema + tools + stream)                                                    |
-| `openRouterText` / `openRouterResponsesText` | `false` (per-call resolution is a follow-up)                                                          |
-| `ollamaText`                                 | `false` (constrained-decoding vs tool-call grammar conflict)                                          |
-| `byteplusText`                               | Per model — `true` only for the 10 ids in `BYTEPLUS_STRUCTURED_OUTPUT_CHAT_MODELS`, `false` otherwise |
+| Adapter                                      | Returns                                                                                                                              |
+| -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------ |
+| `openaiText` / `openaiChatCompletions`       | `true` (all supported models)                                                                                                        |
+| `anthropicText`                              | `true` for Claude 4.5+ (gated by `ANTHROPIC_COMBINED_TOOLS_AND_SCHEMA_MODELS`), `false` otherwise                                    |
+| `geminiText`                                 | `true` for Gemini 3.x (gated by `GEMINI_COMBINED_TOOLS_AND_SCHEMA_MODELS`), `false` otherwise                                        |
+| `grokText`                                   | `true` (all chat models — inherits the OpenAI Responses base; no per-model gate)                                                     |
+| `groqText`                                   | `false` (Groq API rejects schema + tools + stream)                                                                                   |
+| `openRouterText` / `openRouterResponsesText` | Per model — `true` only when the model and every `modelOptions.models` fallback are in `OPENROUTER_COMBINED_TOOLS_AND_SCHEMA_MODELS` |
+| `ollamaText`                                 | `false` (constrained-decoding vs tool-call grammar conflict)                                                                         |
+| `byteplusText`                               | Per model — `true` only for the 10 ids in `BYTEPLUS_STRUCTURED_OUTPUT_CHAT_MODELS`, `false` otherwise                                |
 
 Subclasses can override to narrow the capability. When extending an
 adapter for a custom model that doesn't support the combination, return
@@ -370,7 +404,9 @@ dedicated package required.
 
 ```typescript
 import { openaiCompatible } from '@tanstack/ai-openai/compatible'
-import { createModel } from '@tanstack/ai'
+import { chat, createModel } from '@tanstack/ai'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
 
 // Provider-factory: configure baseURL + apiKey + models ONCE,
 // then select a model per call (the model arg is a type-safe union).
@@ -399,6 +435,9 @@ For a single model, use the one-shot helper:
 
 ```typescript
 import { openaiCompatibleText } from '@tanstack/ai-openai/compatible'
+import { chat } from '@tanstack/ai'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
 
 chat({
   adapter: openaiCompatibleText('deepseek-chat', {
@@ -427,13 +466,17 @@ ElevenLabs `baseUrl`/`headers`). The vendor names still work; when both are
 set, `baseURL` and `defaultHeaders` win.
 
 ```typescript
+import { createGeminiChat } from '@tanstack/ai-gemini'
+
 const gateway = {
   baseURL: 'https://gateway.example.com/google-ai-studio',
   defaultHeaders: {
     'cf-aig-authorization': `Bearer ${process.env.GATEWAY_TOKEN}`,
   },
 }
-createGeminiChat('gemini-3.8-flash', apiKey, { ...gateway })
+createGeminiChat('gemini-3.8-flash', process.env.GOOGLE_API_KEY!, {
+  ...gateway,
+})
 ```
 
 ## Common Mistakes
@@ -443,13 +486,19 @@ createGeminiChat('gemini-3.8-flash', apiKey, { ...gateway })
 The legacy `openai()` (and `anthropic()`, etc.) monolithic adapters are
 deprecated. They take the model in `chat()`, not in the factory.
 
-```typescript
-// WRONG: Legacy monolithic adapter pattern
+```typescript ignore
+// WRONG: Legacy monolithic adapter pattern (no longer exported)
 import { openai } from '@tanstack/ai-openai'
 chat({ adapter: openai(), model: 'gpt-5.2', messages })
+```
 
+```typescript
 // CORRECT: Tree-shakeable adapter, model in factory
+import { chat } from '@tanstack/ai'
 import { openaiText } from '@tanstack/ai-openai'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
+
 chat({ adapter: openaiText('gpt-5.2'), messages })
 ```
 

package/skills/ai-core/adapter-configuration/references/anthropic-adapter.md

@@ -21,45 +21,58 @@ import { anthropicText } from '@tanstack/ai-anthropic'
 
 ## Key Chat Models
 
-| Model               | Context Window | Max Output | Notes                                       |
-| ------------------- | -------------- | ---------- | ------------------------------------------- |
-| `claude-fable-5`    | 1M             | 128K       | Most capable; thinking always on (adaptive) |
-| `claude-sonnet-5`   | 1M             | 128K       | Best balance; adaptive thinking by default  |
-| `claude-opus-4-8`   | 1M             | 128K       | Opus tier; adaptive thinking, no sampling   |
-| `claude-opus-4-7`   | 1M             | 128K       | Older Opus; adaptive thinking, no sampling  |
-| `claude-opus-4-6`   | 200K           | 128K       | Older Opus, adaptive + budget thinking      |
-| `claude-sonnet-4-6` | 1M             | 64K        | Previous gen balanced, adaptive + budget    |
-| `claude-sonnet-4-5` | 200K           | 64K        | Previous gen balanced                       |
-| `claude-opus-4-5`   | 200K           | 32K        | Previous gen most capable                   |
-| `claude-opus-4-1`   | 200K           | 64K        | Deprecated (retires 2026-08-05)             |
-| `claude-haiku-4-5`  | 200K           | 64K        | Fast and affordable                         |
+| Model                | Context Window | Max Output | Notes                                       |
+| -------------------- | -------------- | ---------- | ------------------------------------------- |
+| `claude-fable-5-1`   | 1M             | 128K       | Newest; thinking always on (adaptive)       |
+| `claude-fable-5`     | 1M             | 128K       | Most capable; thinking always on (adaptive) |
+| `claude-opus-5`      | 1M             | 128K       | Opus tier; budget thinking + sampling       |
+| `claude-opus-5-fast` | 1M             | 128K       | Fast-mode Opus 5; same options as opus-5    |
+| `claude-sonnet-5`    | 1M             | 128K       | Best balance; adaptive thinking by default  |
+| `claude-opus-4-8`    | 1M             | 128K       | Opus tier; adaptive thinking, no sampling   |
+| `claude-opus-4-7`    | 1M             | 128K       | Older Opus; adaptive thinking, no sampling  |
+| `claude-opus-4-6`    | 200K           | 128K       | Older Opus, adaptive + budget thinking      |
+| `claude-sonnet-4-6`  | 1M             | 64K        | Previous gen balanced, adaptive + budget    |
+| `claude-sonnet-4-5`  | 200K           | 64K        | Previous gen balanced                       |
+| `claude-opus-4-5`    | 200K           | 32K        | Previous gen most capable                   |
+| `claude-opus-4-1`    | 200K           | 64K        | Deprecated (retires 2026-08-05)             |
+| `claude-haiku-4-5`   | 200K           | 64K        | Fast and affordable                         |
 
 Note: Model IDs use the format `claude-sonnet-5`, `claude-opus-4-8`, etc.
-Retired models (Claude 3.x, Sonnet 3.7, Opus 4 / Sonnet 4) and the `-fast`
-variant ids were removed — every registered id resolves against the
-first-party Anthropic API.
+Retired models (Claude 3.x, Sonnet 3.7, Opus 4 / Sonnet 4) were removed —
+every registered id resolves against the first-party Anthropic API.
+`claude-opus-5-fast` is the only `-fast` id that remains.
+
+`output_config.effort` is typed only on the adaptive-era models
+(`claude-opus-4-7`, `claude-opus-4-8`, `claude-sonnet-5`, `claude-fable-5`,
+`claude-fable-5-1`). There is no top-level `effort` option on any model;
+`claude-opus-4-6` / `claude-sonnet-4-6` accept `thinking: { type: 'adaptive' }`
+but no effort knob.
 
 ## Provider-Specific modelOptions
 
 ```typescript
+import { chat } from '@tanstack/ai'
+import { anthropicText } from '@tanstack/ai-anthropic'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
+
 chat({
   adapter: anthropicText('claude-sonnet-4-6'),
   messages,
   modelOptions: {
     // Sampling
     temperature: 0.7,
-    top_p: 0.9, // cannot be combined with temperature
+    // top_p: 0.9, // cannot be combined with temperature
     max_tokens: 16000,
     // Extended thinking (budget-based)
     thinking: {
       type: 'enabled',
       budget_tokens: 8000, // must be >= 1024 and < max_tokens
     },
-    // Adaptive thinking (claude-sonnet-4-6, claude-opus-4-6+)
-    thinking: {
-      type: 'adaptive',
-    },
-    effort: 'high', // 'max' | 'high' | 'medium' | 'low'
+    // Adaptive thinking (claude-sonnet-4-6, claude-opus-4-6+) — the
+    // alternative to the budget shape above; effort is tuned via
+    // output_config.effort on the adaptive-era models (see below)
+    // thinking: { type: 'adaptive' },
     // Service tier
     service_tier: 'auto', // 'auto' | 'standard_only'
     // Stop sequences
@@ -99,6 +112,11 @@ ANTHROPIC_API_KEY
 The per-model types restrict `modelOptions` on the newest models:
 
 ```typescript
+import { chat } from '@tanstack/ai'
+import { anthropicText } from '@tanstack/ai-anthropic'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
+
 chat({
   adapter: anthropicText('claude-sonnet-5'), // or 'claude-fable-5', 'claude-opus-4-8'
   messages,

package/skills/ai-core/adapter-configuration/references/byteplus-adapter.md

@@ -62,6 +62,11 @@ Media models: `BYTEPLUS_VIDEO_MODELS` (Seedance —
 ## Provider-Specific modelOptions
 
 ```typescript
+import { chat } from '@tanstack/ai'
+import { byteplusText } from '@tanstack/ai-byteplus'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
+
 chat({
   adapter: byteplusText('dola-seed-2-1-turbo-260628'),
   messages,

package/skills/ai-core/adapter-configuration/references/gemini-adapter.md

@@ -38,6 +38,15 @@ Most Gemini text models accept `text`, `image`, `audio`, `video`, and `document`
 ## Provider-Specific modelOptions
 
 ```typescript
+import { chat } from '@tanstack/ai'
+import {
+  geminiText,
+  HarmBlockThreshold,
+  HarmCategory,
+} from '@tanstack/ai-gemini'
+
+const messages = [{ role: 'user' as const, content: 'Hello' }]
+
 chat({
   adapter: geminiText('gemini-2.5-pro'),
   messages,
@@ -47,15 +56,14 @@ chat({
       includeThoughts: true,
       thinkingBudget: 4096,
     },
-    // Thinking (level-based, advanced models)
-    thinkingConfig: {
-      thinkingLevel: 'THINKING_LEVEL_HIGH',
-    },
+    // Thinking (level-based, advanced models) — the alternative to the
+    // budget shape above:
+    // thinkingConfig: { thinkingLevel: 'THINKING_LEVEL_HIGH' },
     // Safety settings
     safetySettings: [
       {
-        category: 'HARM_CATEGORY_HATE_SPEECH',
-        threshold: 'BLOCK_MEDIUM_AND_ABOVE',
+        category: HarmCategory.HARM_CATEGORY_HATE_SPEECH,
+        threshold: HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
       },
     ],
     // Tool config