docs: fix mid-identifier wrap in tables, drop hardcoded deps in AI Adapters

custom.css adds word-break:keep-all + overflow-wrap:normal + white-space:nowrap on inline code inside tables so identifiers like skillFile, description, responseAs, journalFormat, Class<?> stop splitting mid-character under narrow columns. Verified in a browser against /docs/agents/coordinator/ and /docs/tutorial/11-ai-adapters/. Tutorial 11 removes the Built-in properties+dependency block and four per-runtime Dependency subsections — the Runtimes table at the top already carries the Maven coords, and the duplicated snippets held a wrong Embabel version (4.0.38) that would have resolved to nothing.

Author: jfarcand

docs/src/content/docs/tutorial/11-ai-adapters.md

@@ -9,19 +9,7 @@ Atmosphere's AI layer follows the same adapter pattern as its transport layer. J
 
 ## Built-in LLM Client (zero extra dependencies)
 
-Before reaching for an adapter module, know that `atmosphere-ai` itself includes a built-in `OpenAiCompatibleClient`. This client works with **OpenAI**, **Google Gemini**, **Ollama**, **Azure OpenAI**, and any OpenAI-compatible endpoint -- with zero additional dependencies beyond `atmosphere-ai`.
-
-```xml
-<properties>
-    <atmosphere.version>4.0.38</atmosphere.version>
-</properties>
-
-<dependency>
-    <groupId>org.atmosphere</groupId>
-    <artifactId>atmosphere-ai</artifactId>
-    <version>${atmosphere.version}</version>
-</dependency>
-```
+Before reaching for an adapter module, know that `atmosphere-ai` itself includes a built-in `OpenAiCompatibleClient`. This client works with **OpenAI**, **Google Gemini**, **Ollama**, **Azure OpenAI**, and any OpenAI-compatible endpoint — with zero additional dependencies beyond `atmosphere-ai`. The Maven coordinates for every module in this chapter are listed in the [Runtimes table](#runtimes) below and in the [Spring Boot integration guide](/docs/tutorial/14-spring-boot/).
 
 Use it directly via `AiConfig`:
 
@@ -133,16 +121,6 @@ When multiple `AgentRuntime` implementations are on the classpath, the one with
 | `SpringAiToolBridge` | Converts Atmosphere `ToolDefinition` to Spring AI `ToolCallback`. Spring AI handles the tool call loop automatically. |
 | `AtmosphereSpringAiAutoConfiguration` | Spring Boot `@AutoConfiguration`. Activates when `ChatClient` is on the classpath. |
 
-### Dependency
-
-```xml
-<dependency>
-    <groupId>org.atmosphere</groupId>
-    <artifactId>atmosphere-spring-ai</artifactId>
-    <version>${atmosphere.version}</version>
-</dependency>
-```
-
 ### Auto-configuration
 
 The auto-configuration creates a `SpringAiStreamingAdapter` bean and, if a `ChatClient` bean exists, wires it into `SpringAiAgentRuntime`:
@@ -210,16 +188,6 @@ When an `@AiEndpoint` receives a message, `SpringAiAgentRuntime.execute()` runs
 | `LangChain4jToolBridge` | Converts `ToolDefinition` to `ToolSpecification` and handles tool execution. |
 | `AtmosphereLangChain4jAutoConfiguration` | Activates when `StreamingChatLanguageModel` is on the classpath. |
 
-### Dependency
-
-```xml
-<dependency>
-    <groupId>org.atmosphere</groupId>
-    <artifactId>atmosphere-langchain4j</artifactId>
-    <version>${atmosphere.version}</version>
-</dependency>
-```
-
 ### Configuration example
 
 A typical LangChain4j wiring (the same pattern used internally by `atmosphere-ai` when the LangChain4j adapter is on the classpath):
@@ -289,16 +257,6 @@ Unlike Spring AI, LangChain4j does not execute tool callbacks automatically. Whe
 | `AdkEventAdapter` | Subscribes to a `Flowable<Event>` and forwards partial streaming texts, turn completions, and errors to a `StreamingSession`. |
 | `AtmosphereAdkAutoConfiguration` | Activates when `com.google.adk.runner.Runner` is on the classpath. |
 
-### Dependency
-
-```xml
-<dependency>
-    <groupId>org.atmosphere</groupId>
-    <artifactId>atmosphere-adk</artifactId>
-    <version>${atmosphere.version}</version>
-</dependency>
-```
-
 ### ADK-specific details
 
 ADK requires tools to be registered at agent construction time. You cannot add tools dynamically per-request. Use `AdkAgentRuntime.configureWithTools()`:
@@ -352,21 +310,6 @@ The adapter handles three event types:
 
 Embabel is a Kotlin-based agent framework with built-in planning, tool calling, and orchestration. The `atmosphere-embabel` adapter bridges Embabel's `OutputChannel` pattern to `StreamingSession`, streaming agent events (thinking, tool calls, results) to the browser.
 
-### Dependency
-
-```xml
-<dependency>
-    <groupId>org.atmosphere</groupId>
-    <artifactId>atmosphere-embabel</artifactId>
-    <version>${atmosphere.version}</version>
-</dependency>
-<dependency>
-    <groupId>com.embabel.agent</groupId>
-    <artifactId>embabel-agent-platform-autoconfigure</artifactId>
-    <version>4.0.38</version>
-</dependency>
-```
-
 ### Usage
 
 Define an Embabel agent:

docs/src/styles/custom.css

@@ -10,3 +10,39 @@
   --sl-color-accent: #4338ca;
   --sl-color-accent-high: #1e1b4b;
 }
+
+/* --------------------------------------------------------------------
+ * Tables with inline `code` identifiers.
+ *
+ * Starlight's default `code` styling inherits `overflow-wrap:
+ * break-word` from the container, which causes identifiers like
+ * `skillFile`, `description`, `responseAs`, `journalFormat`,
+ * `Class<?>` to split mid-character (e.g. `skillFil` / `e`) when a
+ * narrow table column is too tight for the full token. That is
+ * unreadable. Keep identifiers intact — a narrow cell would rather
+ * widen or let the token sit on its own line than break it in half.
+ * ------------------------------------------------------------------ */
+.sl-markdown-content table code,
+.sl-markdown-content table :not(pre) > code {
+  word-break: keep-all;
+  overflow-wrap: normal;
+  white-space: nowrap;
+}
+
+/* Allow the first column (typically `Attribute` / `Type`) to grow as
+ * needed so the identifier fits. Starlight's default table CSS uses
+ * `table-layout: auto`, which already respects this — the explicit
+ * `white-space: nowrap` above + this `min-width: 0` ensures the
+ * browser doesn't artificially constrain the column. */
+.sl-markdown-content table th,
+.sl-markdown-content table td {
+  min-width: 0;
+}
+
+/* On very narrow viewports the table may need horizontal scroll
+ * rather than broken identifiers. Starlight already wraps tables in
+ * an overflow-x container; make sure it still works with nowrap. */
+.sl-markdown-content .expressive-code + table,
+.sl-markdown-content table {
+  max-width: 100%;
+}