feat(ai): debug logging system across all activities and adapters

.changeset/debug-logging-adapters.md

@@ -0,0 +1,13 @@
+---
+'@tanstack/ai-openai': patch
+'@tanstack/ai-anthropic': patch
+'@tanstack/ai-gemini': patch
+'@tanstack/ai-ollama': patch
+'@tanstack/ai-openrouter': patch
+'@tanstack/ai-grok': patch
+'@tanstack/ai-groq': patch
+'@tanstack/ai-elevenlabs': patch
+'@tanstack/ai-fal': patch
+---
+
+Wire each adapter's text, summarize, image, speech, transcription, and video paths through the new `InternalLogger` from `@tanstack/ai/adapter-internals`: `logger.request(...)` before each SDK call, `logger.provider(...)` for every chunk received, and `logger.errors(...)` in catch blocks. Migrates all pre-existing ad-hoc `console.*` calls in adapter catch blocks (including the OpenAI and ElevenLabs realtime adapters) onto the structured logger. No adapter factory or config-shape changes.

.changeset/debug-logging-ai.md

@@ -0,0 +1,20 @@
+---
+'@tanstack/ai': minor
+---
+
+**Pluggable debug logging across every activity.** `chat`, `summarize`, `generateImage`, `generateVideo`, `generateSpeech`, and `generateTranscription` now accept a `debug?: DebugOption` that turns on structured per-category logs (`request`, `provider`, `output`, `middleware`, `tools`, `agentLoop`, `config`, `errors`).
+
+```ts
+chat({ adapter, messages, debug: true }) // all categories on
+chat({ adapter, messages, debug: false }) // silent (incl. errors)
+chat({ adapter, messages, debug: { middleware: false } }) // all except middleware
+chat({ adapter, messages, debug: { logger: pino } }) // route to a custom logger
+```
+
+Additions:
+
+- New `Logger` interface (`debug` / `info` / `warn` / `error`) and default `ConsoleLogger` that routes to matching `console.*` methods and prints nested `meta` via `console.dir(meta, { depth: null, colors: true })` so streamed provider payloads render in full.
+- New `DebugCategories` / `DebugConfig` / `DebugOption` public types.
+- New internal `@tanstack/ai/adapter-internals` subpath export exposing `InternalLogger` + `resolveDebugOption` so provider adapters can thread logging without leaking internals on the public surface.
+- Each log line is prefixed with an emoji + `[tanstack-ai:<category>]` tag so categories are visually distinguishable in dense streams. Errors log unconditionally unless explicitly silenced.
+- `TextEngine`, `MiddlewareRunner`, and every activity entry point thread a resolved `InternalLogger` through the pipeline — no globals, concurrent calls stay independent.

docs/advanced/debug-logging.md

@@ -0,0 +1,143 @@
+---
+title: Debug Logging
+id: debug-logging
+order: 3
+description: "Turn on structured, category-toggleable debug logging to see every chunk, middleware transform, and tool call inside TanStack AI."
+keywords:
+  - tanstack ai
+  - debug
+  - logging
+  - logger
+  - pino
+  - troubleshooting
+  - chunks
+  - middleware debugging
+---
+
+# Debug Logging
+
+You have a `chat()` that isn't behaving as expected — a missing chunk, a middleware that doesn't seem to fire, a tool call with wrong args. By the end of this guide, you'll have turned on debug logging and will see every chunk, middleware transform, and tool call flowing through your call.
+
+## Turn it on
+
+Add `debug: true` to any activity call:
+
+```typescript
+import { chat } from "@tanstack/ai";
+import { openaiText } from "@tanstack/ai-openai";
+
+const stream = chat({
+  adapter: openaiText("gpt-4o"),
+  messages: [{ role: "user", content: "Hello" }],
+  debug: true,
+});
+```
+
+Every internal event now prints to the console with a `[tanstack-ai:<category>]` prefix:
+
+```
+[tanstack-ai:request] activity=chat provider=openai model=gpt-4o messages=1 tools=0 stream=true
+[tanstack-ai:agentLoop] run started
+[tanstack-ai:provider] provider=openai type=response.output_text.delta
+[tanstack-ai:output] type=TEXT_MESSAGE_CONTENT
+...
+```
+
+## Narrow what's printed
+
+Pass a `DebugConfig` object instead of `true`. Every unspecified category defaults to `true`, so toggle by setting specific flags to `false`:
+
+```typescript
+chat({
+  adapter: openaiText("gpt-4o"),
+  messages,
+  debug: { middleware: false }, // everything except middleware
+});
+```
+
+If you want to see ONLY a specific set of categories, set the rest to `false` explicitly. Errors default to `true` — keep them on unless you really want total silence:
+
+```typescript
+chat({
+  adapter: openaiText("gpt-4o"),
+  messages,
+  debug: {
+    provider: true,
+    output: true,
+    middleware: false,
+    tools: false,
+    agentLoop: false,
+    config: false,
+    errors: true,         // keep errors on — they're cheap and important
+    request: false,
+  },
+});
+```
+
+## Pipe into your own logger
+
+Pass a `Logger` implementation and all debug output flows through it instead of `console`:
+
+```typescript
+import type { Logger } from "@tanstack/ai";
+import pino from "pino";
+
+const pinoLogger = pino();
+const logger: Logger = {
+  debug: (msg, meta) => pinoLogger.debug(meta, msg),
+  info:  (msg, meta) => pinoLogger.info(meta, msg),
+  warn:  (msg, meta) => pinoLogger.warn(meta, msg),
+  error: (msg, meta) => pinoLogger.error(meta, msg),
+};
+
+chat({
+  adapter: openaiText("gpt-4o"),
+  messages,
+  debug: { logger }, // all categories on, piped to pino
+});
+```
+
+The default logger is exported as `ConsoleLogger` if you want to wrap it:
+
+```typescript
+import { ConsoleLogger } from "@tanstack/ai";
+```
+
+## Categories reference
+
+| Category | Logs | Applies to |
+|----------|------|------------|
+| `request` | Outgoing call to a provider (model, message count, tool count) | All activities |
+| `provider` | Every raw chunk/frame received from a provider SDK | Streaming activities (chat, realtime) |
+| `output` | Every chunk or result yielded to the caller | All activities |
+| `middleware` | Inputs and outputs around every middleware hook | `chat()` only |
+| `tools` | Before/after tool call execution | `chat()` only |
+| `agentLoop` | Agent-loop iterations and phase transitions | `chat()` only |
+| `config` | Config transforms returned by middleware `onConfig` hooks | `chat()` only |
+| `errors` | Every caught error anywhere in the pipeline | All activities |
+
+## Errors are always logged
+
+Errors flow through the logger unconditionally — even when you omit `debug`:
+
+```typescript
+chat({ adapter, messages }); // still prints [tanstack-ai:errors] ... on failure
+```
+
+To fully silence (including errors), set `debug: false` or `debug: { errors: false }`. Errors also always reach the caller via thrown exceptions or `RUN_ERROR` stream chunks — the logger is additive, not the only surface.
+
+## Non-chat activities
+
+The same `debug` option works on every activity:
+
+```typescript
+summarize({ adapter, text, debug: true });
+generateImage({ adapter, prompt: "a cat", debug: { logger } });
+generateSpeech({ adapter, text, debug: { request: true } });
+```
+
+The chat-only categories (`middleware`, `tools`, `agentLoop`, `config`) simply never fire for these activities because those concepts don't exist in their pipelines.
+
+## Related
+
+If you're building middleware and want to see chunks flow through it, `debug: { middleware: true }` is faster than writing a logging middleware. See [Middleware](./middleware) for writing your own middleware, or [Observability](./observability) for the programmatic event client.

docs/advanced/extend-adapter.md

@@ -1,7 +1,7 @@
 ---
 title: Extend Adapter
 id: extend-adapter
-order: 7
+order: 8
 description: "Extend TanStack AI adapter factories with custom model IDs and fine-tuned models while keeping full type safety for input modalities and provider options."
 keywords:
   - tanstack ai

docs/advanced/middleware.md

@@ -49,6 +49,9 @@ const stream = chat({
 });
 ```
 
+> **Just want to see chunks flowing through your middleware during development?**
+> Use `debug: { middleware: true }` on your `chat()` call — no custom middleware required. See [Debug Logging](./debug-logging).
+
 ## Lifecycle Overview
 
 Every `chat()` invocation follows a predictable lifecycle. Middleware hooks fire at specific phases:

docs/advanced/multimodal-content.md

@@ -1,7 +1,7 @@
 ---
 title: Multimodal Content
 id: multimodal-content
-order: 3
+order: 4
 description: "Send images, audio, video, and documents alongside text in TanStack AI messages with typed ContentPart primitives for multimodal models."
 keywords:
   - tanstack ai

docs/advanced/observability.md

@@ -19,6 +19,9 @@ The `@tanstack/ai` package offers you an event client for observability and debu
 It's a fully type-safe decoupled event-driven system that emits events whenever they are internally
 triggered and you can subscribe to those events for observability.
 
+> **Looking for quick diagnostic console output instead of a programmatic event stream?**
+> See [Debug Logging](./debug-logging) for turning on category-toggleable logging across every adapter and middleware hook.
+
 Because the same event client is used for both the TanStack Devtools system and observability locally it will work
 by subscribing to the event bus and emitting events to/from the event bus into the listeners by default. If you 
 want to subscribe to events in production as well you need to pass in a third argument to the `on` function,

docs/advanced/per-model-type-safety.md

@@ -1,7 +1,7 @@
 ---
 title: Per-Model Type Safety
 id: per-model-type-safety
-order: 4
+order: 5
 description: "TanStack AI narrows modelOptions and content types to the specific model you select, enforcing capabilities at compile time."
 keywords:
   - tanstack ai

docs/advanced/runtime-adapter-switching.md

@@ -1,7 +1,7 @@
 ---
 title: Runtime Adapter Switching
 id: runtime-adapter-switching
-order: 5
+order: 6
 description: "Let users switch between LLM providers at runtime in TanStack AI while keeping full TypeScript type safety for each adapter's model options."
 keywords:
   - tanstack ai

docs/advanced/tree-shaking.md

@@ -1,7 +1,7 @@
 ---
 title: Tree-Shaking
 id: tree-shaking
-order: 6
+order: 7
 description: "TanStack AI's tree-shakeable architecture — import only the activities and adapters you use for minimal bundle size across chat, image, and speech."
 keywords:
   - tanstack ai

docs/config.json

@@ -163,6 +163,10 @@
           "label": "Middleware",
           "to": "advanced/middleware"
         },
+        {
+          "label": "Debug Logging",
+          "to": "advanced/debug-logging"
+        },
         {
           "label": "Observability",
           "to": "advanced/observability"