docs: refresh runtime/version/sample claims to 4.0.43 ground truth

Author: jfarcand

docs/src/content/docs/index.mdx

@@ -32,6 +32,6 @@ import { Card, CardGrid } from '@astrojs/starlight/components';
     Build a complete real-time application from scratch. Covers transports, rooms, AI streaming, MCP, clustering, channels, and production deployment.
   </Card>
   <Card title="Spring Boot & Quarkus" icon="setting">
-    First-class starters with auto-configuration, native image support, Micrometer metrics, and OpenTelemetry tracing. Tracks Spring Boot 4.0.5 and Quarkus 3.31.
+    First-class starters with auto-configuration, native image support, Micrometer metrics, and OpenTelemetry tracing. Tracks Spring Boot 4.0.5 and Quarkus 3.35.
   </Card>
 </CardGrid>

docs/src/content/docs/integrations/koog.md

@@ -87,18 +87,20 @@ aspiration.
 | `SYSTEM_PROMPT` | ✅ | Honored by the Koog prompt builder |
 | `CONVERSATION_MEMORY` | ✅ | Per-session memory threaded through `AgentExecutionContext` |
 | `AGENT_ORCHESTRATION` | ✅ | Works with `@Coordinator` and `@Fleet` |
-| `TOKEN_USAGE` | — | Not declared. Koog reports usage on its response objects but the bridge does not yet thread it into `StreamingSession.usage(TokenUsage)`. Tier 1 follow-up. |
-| `VISION` | — | Koog 0.7.x does not expose a stable multi-modal input API on the bridge path |
-| `MULTI_MODAL` | — | Same limitation |
-| `PROMPT_CACHING` | — | Koog 0.7.3 only ships Bedrock-specific cache variants; no OpenAI-compatible passthrough |
-| `PER_REQUEST_RETRY` | — | Inherits Koog's native retry layer (not per-request overridable from the bridge) |
+| `TOKEN_USAGE` | ✅ | The `StreamFrame.End` handler reads Koog's usage totals and emits a typed `TokenUsage` record via `session.usage()` after drain (`executeWithAgent` lines 223-232). |
+| `VISION` | ✅ | Koog 0.8.0 accepts `ContentPart.Image` natively via `buildPrompt`'s `user(String, List<ContentPart>)` overload (no-tools path) |
+| `AUDIO` | ✅ | Same path as `VISION` — `ContentPart.Audio` attached via `AttachmentContent.Binary.Base64` |
+| `MULTI_MODAL` | ✅ | Combined image + audio + text inputs work on the no-tools path; tools + multi-modal degrade gracefully (the tool path wins with a WARN — `AIAgent.run(String)` only accepts plain text) |
+| `PROMPT_CACHING` | ✅ | Koog 0.8.0 honors Bedrock-specific `CacheControl.Bedrock.{FiveMinutes, OneHour}` mapped from Atmosphere's portable `CacheHint`. Non-Bedrock providers silently drop the cache control — same shape Spring AI / LC4j take for OpenAI `prompt_cache_key`. |
+| `CANCELLATION` | ✅ | `executeWithHandle` returns an `ExecutionHandle` whose `cancel()` calls `Job.cancel()` + `Thread.interrupt()` + resolves the done-future with synthetic completion (terminal-path closure per Correctness Invariant #2) |
+| `PER_REQUEST_RETRY` | ✅ | Honored via `executeWithOuterRetry` which wraps `executeInternal` in a retry loop respecting `context.retryPolicy()`. Pre-stream transient failures retry on top of Koog's native HTTP retry. |
 
 Exclusions are **honest** — Koog declares them as absent in its `capabilities()` set so runtime-truth advertising is accurate (Correctness Invariant #5). When Koog upstream adds these surfaces in a future release, the bridge will honor them without a breaking change. No `KoogEmbeddingRuntime` ships today — if you need Koog-backed embeddings, wire a Spring AI or LangChain4j `EmbeddingModel` alongside the Koog agent runtime.
 
 ## Samples
 
-- [spring-boot-koog-chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-koog-chat) — `@AiEndpoint` chat sample routing through Koog's `executeStreaming`
-- [spring-boot-ai-classroom](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-classroom) — swap the runtime to Koog by changing one dependency, same `@Agent` code
+- [spring-boot-ai-chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-chat) — swap to Koog by adding `atmosphere-koog` to `pom.xml`; the same `@AiEndpoint` code routes through `KoogAgentRuntime` (Atmosphere's SPI promise)
+- [spring-boot-ai-classroom](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-classroom) — multi-room shared AI; same one-Maven-dep swap to route through Koog
 
 ## See Also
 

docs/src/content/docs/integrations/quarkus.md

@@ -1,11 +1,11 @@
 ---
 title: "Quarkus"
-description: "Build-time processing for Quarkus 3.31.3+"
+description: "Build-time processing for Quarkus 3.35.2+"
 ---
 
 # Quarkus Integration
 
-A Quarkus extension that integrates Atmosphere with Quarkus 3.31.3+. Provides build-time annotation scanning via Jandex, Arc CDI integration, and GraalVM native image support.
+A Quarkus extension that integrates Atmosphere with Quarkus 3.35.2+. Provides build-time annotation scanning via Jandex, Arc CDI integration, and GraalVM native image support.
 
 ## Maven Coordinates
 
@@ -113,11 +113,12 @@ See [Spring Boot → `@AiEndpoint` annotation surfaces](spring-boot/#aiendpoint-
 
 - **`load-on-startup`** must be > 0 for the endpoint to register at boot time (Quarkus's `UndertowDeploymentRecorder` skips on `<= 0`, unlike the Servlet spec).
 
-**Runtime coverage:** per-request retry is **Built-in only** in 4.0.36. Framework runtimes inherit their native retry layers. See the [per-runtime capability matrix](../../tutorial/11-ai-adapters/#per-runtime-capability-matrix).
+**Runtime coverage:** all eight framework adapters declare `PER_REQUEST_RETRY` honestly as of 4.0.43 (commit `374631e7`) — they all inherit `AbstractAgentRuntime.executeWithOuterRetry`. Each adapter stacks this on top of its own native retry layer (Spring Retry, LC4j `RetryUtils`, ADK `HttpClient`, Koog `CallRetryPolicy`, SK `OpenAIAsyncClient`). See the [per-runtime capability matrix](../../tutorial/11-ai-adapters/#per-runtime-capability-matrix).
 
 ## Samples
 
 - [Quarkus Chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/quarkus-chat) -- real-time chat with WebSocket and long-polling fallback
+- [Quarkus AI Chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/quarkus-ai-chat) -- five `@AiEndpoint` demos (basic streaming, retry, multi-modal, prompt caching, structured output) on Quarkus + Quarkus LangChain4j bridge, port `18810`
 
 ## See Also
 

docs/src/content/docs/integrations/semantic-kernel.md

@@ -7,7 +7,7 @@ description: "AgentRuntime backed by Microsoft Semantic Kernel — ChatCompletio
 
 `AgentRuntime` implementation backed by [Microsoft Semantic Kernel](https://learn.microsoft.com/en-us/semantic-kernel/) for Java. Semantic Kernel is Microsoft's enterprise-grade AI orchestration SDK with plugins, memory, and planners. Adding `atmosphere-semantic-kernel` as a dependency makes `@AiEndpoint` route streaming chat through an SK `ChatCompletionService` and makes `EmbeddingRuntime` use SK's `TextEmbeddingGenerationService`.
 
-Semantic Kernel is the **7th runtime** added to Atmosphere in 4.0.36.
+Semantic Kernel landed as the 7th `AgentRuntime` in Atmosphere 4.0.36; the 8th and 9th (AgentScope, Spring AI Alibaba) followed in 4.0.42.
 
 ## Maven Coordinates
 
@@ -126,7 +126,7 @@ See the `modules/semantic-kernel/README.md` exclusion note for the full trade-of
 
 ## Samples
 
-Semantic Kernel is wired into [spring-boot-ai-classroom](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-classroom) as one of the seven swappable runtimes. Drop the `atmosphere-semantic-kernel` JAR alongside `atmosphere-ai` and the same `@AiEndpoint` code routes through SK.
+Semantic Kernel is wired into [spring-boot-ai-classroom](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-classroom) as one of the nine swappable runtimes. Drop the `atmosphere-semantic-kernel` JAR alongside `atmosphere-ai` and the same `@AiEndpoint` code routes through SK.
 
 ## See Also
 

docs/src/content/docs/integrations/spring-boot.md

@@ -244,7 +244,7 @@ public class BackgroundChat {
 - `maxDelayMs` — cap on exponential backoff (default `30000`)
 - `backoffMultiplier` — exponential factor (default `2.0`)
 
-**Runtime coverage:** per-request retry is **Built-in only** in 4.0.36. Framework runtimes (Spring AI, LangChain4j, ADK, Koog, Embabel, Semantic Kernel) inherit their native retry layers and ignore the per-request override. The Built-in runtime threads `context.retryPolicy()` into `OpenAiCompatibleClient.sendWithRetry` as a real override. See the [per-runtime capability matrix](../../tutorial/11-ai-adapters/#per-runtime-capability-matrix) for the full breakdown.
+**Runtime coverage:** all eight framework adapters declare `PER_REQUEST_RETRY` honestly as of 4.0.43 (commit `374631e7`) — `AbstractAgentRuntime.executeWithOuterRetry` wraps each adapter's dispatch in a retry loop respecting `context.retryPolicy(...)`, on top of each runtime's own native retry layer (Spring Retry, LC4j `RetryUtils`, ADK `HttpClient`, Koog `CallRetryPolicy`, SK `OpenAIAsyncClient`). The Built-in runtime additionally threads `context.retryPolicy()` into `OpenAiCompatibleClient.sendWithRetry` as a native override. See the [per-runtime capability matrix](../../tutorial/11-ai-adapters/#per-runtime-capability-matrix) for the full breakdown.
 
 ## Samples
 

docs/src/content/docs/reference/ai.md

@@ -81,7 +81,7 @@ public interface AgentRuntime {
 
 To switch runtimes, change a single Maven dependency — no code changes needed.
 
-> **Spring AI Alibaba runtime — Spring Boot 3 only today.** Spring AI Alibaba `1.1.2.0` is compiled against Spring AI `1.1.2`, and `spring-ai-alibaba-graph-core-1.1.2.0` hardcodes Spring AI 1.1.2-only types like `DeepSeekAssistantMessage`, so the runtime requires Spring AI 1.1.2. Spring AI 1.1.2 itself requires Spring Boot 3 — it pins the SB3-era FQN of `RestClientAutoConfiguration`, which Spring Boot 4 ships at a renamed FQN. Drop `atmosphere-spring-ai-alibaba` into a Spring Boot 3 sample (e.g. `samples/spring-boot-ai-chat -Pspring-boot3`) and it round-trips end-to-end (verified via chrome-devtools against Ollama). A Spring Boot 4 path will become possible once Alibaba publishes a Spring AI 2.x-aligned `spring-ai-alibaba-agent-framework`. `atmosphere-agentscope` is unaffected and works on Spring Boot 4.
+> **Spring AI Alibaba runtime — Spring Boot 3 only today.** Spring AI Alibaba `1.1.2.2` is compiled against Spring AI `1.1.2`, and `spring-ai-alibaba-graph-core-1.1.2.x` hardcodes Spring AI 1.1.2-only types like `DeepSeekAssistantMessage`, so the runtime requires Spring AI 1.1.2. Spring AI 1.1.2 itself requires Spring Boot 3 — it pins the SB3-era FQN of `RestClientAutoConfiguration`, which Spring Boot 4 ships at a renamed FQN. Drop `atmosphere-spring-ai-alibaba` into a Spring Boot 3 sample (e.g. `samples/spring-boot-ai-chat -Pspring-boot3`) and it round-trips end-to-end (verified via chrome-devtools against Ollama). A Spring Boot 4 path will become possible once Alibaba publishes a Spring AI 2.x-aligned `spring-ai-alibaba-agent-framework`. `atmosphere-agentscope` is unaffected and works on Spring Boot 4.
 
 ### Per-Request Runtime Extensions
 
@@ -90,16 +90,22 @@ that need framework-native composition (Spring AI advisor chain, LangChain4j
 `AiServices`, Koog graph DSL, ADK multi-agent topology), a small per-request
 helper attaches the framework-native object to `AgentExecutionContext.metadata()`
 and the runtime applies it for that one call — no `AgentRuntime` SPI growth, no
-mutation of shared beans. All five helpers follow the `CacheHint` pattern:
-`from(context)` and `attach(context, ...)` static methods.
+mutation of shared beans. Every helper follows the `CacheHint` pattern:
+`from(context)` and `attach(context, ...)` static methods, with strict type
+checking that throws `IllegalArgumentException` on a wrong-type slot (silent
+drops would mask the override never firing).
 
 | Helper | Runtime | Slot it drives |
 |--------|---------|----------------|
-| `SpringAiAdvisors` | Spring AI | `ChatClient.prompt().advisors(...)` — RAG, memory, guardrails, observability |
-| `LangChain4jAiServices` | LangChain4j | Routes through caller's `AiServices`-backed interface (`TokenStream` callbacks bridged to session) |
+| `SpringAiAdvisors` | Spring AI | `ChatClient.prompt().advisors(...)` — RAG, memory, guardrails, observability (additive — multiple advisors compose into a chain) |
+| `LangChain4jAiServices` | LangChain4j | Routes through caller's `AiServices`-backed interface (`TokenStream` callbacks bridged to session) — gives access to `maxSequentialToolsInvocations`, custom system message providers, etc. |
 | `KoogStrategy` | Koog | Swaps default `chatAgentStrategy()` with a custom `AIAgentGraphStrategy<String, String>` from the `strategy {}` DSL |
 | `AdkRootAgent` | ADK | Replaces the runtime's default `LlmAgent` with `SequentialAgent` / `ParallelAgent` / `LoopAgent` / any `BaseAgent` subclass |
-| `ToolLoopPolicies` | Built-in | Per-request `ToolLoopPolicy(maxIterations, OnMaxIterations)` for the OpenAI-compatible tool loop |
+| `SemanticKernelInvocation` | Semantic Kernel | Per-request `InvocationContext` — unlocks `KernelHooks`, `withMaxAutoInvokeAttempts`, custom `PromptExecutionSettings` |
+| `EmbabelPromptRunner` | Embabel | `UnaryOperator<PromptRunner>` customizer applied AFTER the runtime's default wiring — stack `withTemperature` / `withModel` / `withGuardrails` on top. Atmosphere-native dispatch path only |
+| `AgentScopeAgent` | AgentScope | Per-request `ReActAgent` — useful when different prompts route through different agent topologies (planner vs. quick lookup) without re-installing the runtime client |
+| `SpringAiAlibabaRunnableConfig` | Spring AI Alibaba | Per-request `RunnableConfig` — Alibaba's natural per-invocation handle for `threadId` (memory thread continuation), `checkPointId` (resume), `streamMode`, metadata, store |
+| `ToolLoopPolicies` | Built-in, Koog | Per-request `ToolLoopPolicy(maxIterations, OnMaxIterations)` — Built-in honors via OpenAI-compatible tool loop, Koog via `AIAgent.maxIterations` |
 
 Example — Spring AI advisor scoped to one request:
 
@@ -123,11 +129,14 @@ returns non-empty). See the per-module READMEs for full DSL examples:
 [`modules/adk`](https://github.com/Atmosphere/atmosphere/blob/main/modules/adk/README.md#multi-agent-composition-adkrootagent),
 and the [`ToolLoopPolicy`](https://github.com/Atmosphere/atmosphere/blob/main/modules/ai/README.md#tool-loop-policy) section.
 
-Other runtimes (`AgentScope`, `Embabel`, `SemanticKernel`, `SpringAiAlibaba`) do
-not yet ship a per-request bridge. `Embabel` did get **native streaming** in the
-same merge: when `StreamingPromptRunnerBuilder.streaming().generateStream()` is
-available the runtime emits `Flux<String>` chunks directly to the session, with
-graceful fallback to `runner.generateText(...)` when the streaming API is absent.
+All eight framework runtimes now ship a per-request sidecar (the four added
+most recently — `SemanticKernelInvocation`, `EmbabelPromptRunner`,
+`AgentScopeAgent`, `SpringAiAlibabaRunnableConfig` — close the matrix that
+previously left these adapters without a per-request escape hatch). `Embabel`
+also has **native streaming**: when
+`StreamingPromptRunnerBuilder.streaming().generateStream()` is available the
+runtime emits `Flux<String>` chunks directly to the session, with graceful
+fallback to `runner.generateText(...)` when the streaming API is absent.
 
 ### Model-Lifecycle Observability
 
@@ -682,7 +691,8 @@ The `RecordingSession` test double captures all events, text chunks, metadata, a
 - [Spring Boot AI Chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-chat) -- built-in client with Gemini/OpenAI/Ollama
 - [Spring Boot AI Tools](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-tools) -- framework-agnostic `@AiTool` pipeline
 - [Spring Boot AI Classroom](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-ai-classroom) -- rooms-based multi-room AI with an Expo client
-- [Spring Boot Koog Chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-koog-chat) -- JetBrains Koog adapter
+- [Spring Boot RAG Chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-rag-chat) -- Spring AI VectorStore-backed RAG agent
+- [Quarkus AI Chat](https://github.com/Atmosphere/atmosphere/tree/main/samples/quarkus-ai-chat) -- five `@AiEndpoint` demos on Quarkus + LangChain4j bridge
 - [Spring Boot Dentist Agent](https://github.com/Atmosphere/atmosphere/tree/main/samples/spring-boot-dentist-agent) -- `@Agent` with tools, memory, and approval gates
 
 ## See Also

docs/src/content/docs/reference/foundation-primitives.md

@@ -110,7 +110,7 @@ broadcasters on reconnect but not in-flight agent runs.
 - **Shipped**: SPI + default implementation + unit tests + zero-warning
   compile. All eight primitives are shipped in the current branch.
 - **Complete**: shipped + wired into live paths + cross-runtime contract
-  tests green on all seven runtimes + security invariants verified.
+  tests green on all nine runtimes + security invariants verified.
 
 The completion bar is raised by three follow-up phases: gateway wire-in,
 cross-runtime contract test matrix, and security default hardening.

docs/src/content/docs/tutorial/00-cli.md

@@ -80,7 +80,7 @@ The interactive picker shows every sample grouped by category. Pick one, then ch
     4) spring-boot-ai-chat             AI streaming with conversation memory and structured events
     5) spring-boot-ai-classroom        Multiple clients share streaming AI responses
     6) spring-boot-ai-tools            Framework-agnostic @AiTool tool calling
-    7) spring-boot-koog-chat           Kotlin-based Koog agent chat
+    7) spring-boot-rag-chat            RAG agent over Spring AI VectorStore
     ...
 
   Pick a sample:
@@ -124,9 +124,12 @@ plain `mvn compile`.
 | `mcp-server` | `spring-boot-mcp-server` | MCP tools, resources, and prompts |
 | `rag` | `spring-boot-rag-chat` | RAG chat with Spring AI vector store + embeddings |
 | `agent` | `spring-boot-dentist-agent` | Single agent with `@Agent` / `@Command` + skill file |
-| `koog` | `spring-boot-koog-chat` | JetBrains Koog runtime (Kotlin) |
 | `multi-agent` | `spring-boot-multi-agent-startup-team` | 5-agent `@Coordinator` fleet with A2A specialists |
 | `classroom` | `spring-boot-ai-classroom` | Multi-room shared AI (Spring Boot + Expo React Native client) |
+| `ms-governance` | `spring-boot-ms-governance-chat` | Microsoft Agent Governance Toolkit demo (`@AgentScope` policy chain) |
+| `coding-agent` | `spring-boot-coding-agent` | Sandboxed coding agent — clones a repo, reads files, proposes a patch |
+| `guarded-agent` | `spring-boot-guarded-email-agent` | Plan-and-Verify (`atmosphere-verifier`) inbox-exfiltration demo |
+| `assistant` | `spring-boot-personal-assistant` | Long-lived `@Coordinator` assistant with memory + delegation |
 
 Pass `--skill-file <path>` to auto-select the `agent` template with
 your skill file wired in.
@@ -146,9 +149,9 @@ atmosphere new my-kotlin-ai --template ai-chat --runtime embabel
 
 Available runtimes: `builtin` (default — no extra deps),
 `spring-ai`, `langchain4j`, `adk`, `koog`, `embabel`,
-`semantic-kernel`. The mapping lives in `cli/runtime-overlays.json`;
-each entry lists the dep set to inject (and any extra repository, e.g.
-Embabel's release repo).
+`semantic-kernel`, `agentscope`, `spring-ai-alibaba`. The mapping
+lives in `cli/runtime-overlays.json`; each entry lists the dep set to
+inject (and any extra repository, e.g. Embabel's release repo).
 
 #### `--force`: Strip Pre-Pinned Adapters
 
@@ -201,7 +204,8 @@ always in sync.
 | `spring-boot-ai-chat` | Conversation memory, structured events, capability validation | 8080 |
 | `spring-boot-ai-classroom` | Multiple clients share streaming AI responses | 8080 |
 | `spring-boot-ai-tools` | Framework-agnostic `@AiTool` tool calling | 8080 |
-| `spring-boot-koog-chat` | Kotlin-based Koog agent chat | 8080 |
+| `spring-boot-rag-chat` | RAG agent over Spring AI VectorStore | 8080 |
+| `quarkus-ai-chat` | Same `@AiEndpoint` shape as `spring-boot-ai-chat` on Quarkus + LangChain4j bridge | 18810 |
 | `spring-boot-rag-chat` | RAG with Spring AI vector store | 8080 |
 
 ### Agents & Coordination