fix(sdk): stop four public surfaces accepting input and dropping it

docs/api/README.md

@@ -2746,6 +2746,7 @@ console.log(result.content);
 - [DataStreamWriter](type-aliases/DataStreamWriter.md)
 - [WebSocketMessageType](type-aliases/WebSocketMessageType.md)
 - [WebSocketMessage](type-aliases/WebSocketMessage.md)
+- [WebSocketRequestId](type-aliases/WebSocketRequestId.md)
 - [AuthenticatedUser](type-aliases/AuthenticatedUser.md)
 - [WebSocketConnection](type-aliases/WebSocketConnection.md)
 - [AuthStrategy](type-aliases/AuthStrategy.md)

docs/api/classes/WebSocketMessageRouter.md

@@ -34,7 +34,7 @@ Register a message route
 
 ##### handler
 
-(`connection`, `payload`) => `Promise`\<`unknown`\>
+(`connection`, `payload`, `requestId?`) => `Promise`\<`unknown`\>
 
 #### Returns
 

docs/api/functions/createAgentWebSocketHandler.md

@@ -6,16 +6,32 @@
 
 # Function: createAgentWebSocketHandler()
 
-> **createAgentWebSocketHandler**(`_neurolink`): [`WebSocketHandler`](../type-aliases/WebSocketHandler.md)
+> **createAgentWebSocketHandler**(`neurolink`): [`WebSocketHandler`](../type-aliases/WebSocketHandler.md)
 
-Create a WebSocket handler for AI agent interactions
+Server Adapters for exposing NeuroLink as HTTP APIs
+
+Supports multiple frameworks: Hono, Express, Fastify, Koa
 
 ## Parameters
 
-### \_neurolink
+### neurolink
 
-`unknown`
+[`NeuroLink`](../classes/NeuroLink.md)
 
 ## Returns
 
 [`WebSocketHandler`](../type-aliases/WebSocketHandler.md)
+
+## Example
+
+```typescript
+import { NeuroLink } from "@juspay/neurolink";
+import { createServer } from "@juspay/neurolink/server";
+
+const neurolink = new NeuroLink({ provider: "openai" });
+const server = await createServer(neurolink, {
+  framework: "hono",
+  config: { port: 3000 },
+});
+await server.start();
+```

docs/api/type-aliases/StreamLoopArgs.md

@@ -163,3 +163,13 @@ Signals the channel that no further chunks will arrive (success or error path al
 #### Returns
 
 `void`
+
+---
+
+### responseFormat?
+
+> `optional` **responseFormat?**: [`OpenAICompatResponseFormat`](OpenAICompatResponseFormat.md)
+
+`response_format` for this turn's requests, computed once up front from
+`options.schema` (see `suppressResponseFormatWithTools`). Absent when no
+schema was requested, or when tools suppress it.

docs/api/type-aliases/StreamResult.md

@@ -177,6 +177,45 @@ Verbatim provider finish/stop reason for the turn's terminal model call.
 
 > `optional` **stepsUsed?**: `number`
 
+#### structuredData?
+
+> `optional` **structuredData?**: `unknown`
+
+Parsed structured output for a `stream({ schema })` turn, available
+AFTER the stream has been drained.
+
+Lives on metadata for the same reason `finishReason` above does: it is a
+mutable reference the loop fills in, so result-object spreads in stream
+wrappers cannot snapshot it before it resolves. A plain field rather than
+a promise, deliberately — when a middleware short-circuits the request
+and the loop never runs, an unresolved promise would hang every reader,
+whereas an absent field is simply absent. Readers must tolerate that:
+absence means the model never produced the object.
+
+#### structuredDataUsage?
+
+> `optional` **structuredDataUsage?**: `object`
+
+Tokens spent by the tool-free re-ask that produced `structuredData`,
+when one was needed.
+
+That re-ask is a second, separately-billed model call, and the stream's
+own `usage` has already resolved by the time it runs — so folding these
+tokens into it would mutate a value a caller may have read, and leaving
+them out entirely would under-report what the turn cost. Reported here
+instead: same delivery as `structuredData`, filled at the same moment,
+read at the same moment. Absent when no re-ask was needed, which is the
+common case. The generate path accounts for its equivalent re-ask
+inline, since there the usage has not been handed out yet.
+
+##### structuredDataUsage.inputTokens
+
+> **inputTokens**: `number`
+
+##### structuredDataUsage.outputTokens
+
+> **outputTokens**: `number`
+
 #### thoughtSignature?
 
 > `optional` **thoughtSignature?**: `string`

docs/api/type-aliases/TextGenerationOptions.md

@@ -656,6 +656,40 @@ module's own default, unchanged unless a caller opts in here.
 
 ---
 
+### videoOptions?
+
+> `optional` **videoOptions?**: `object`
+
+Video processing options (mirrors `GenerateOptions.videoOptions`). Never
+declared here before, so `buildGenerateTextOptions` had nowhere to
+forward the caller's setting even once the allowlist itself named it.
+
+#### frames?
+
+> `optional` **frames?**: `number`
+
+Frames to extract. Unset lets VideoProcessor pick from the clip's duration; clamped to 100.
+
+#### quality?
+
+> `optional` **quality?**: `number`
+
+Frame encoder quality, clamped to 1-100. Default 80.
+
+#### format?
+
+> `optional` **format?**: `"jpeg"` \| `"png"`
+
+Frame encoding. Default jpeg.
+
+#### transcribeAudio?
+
+> `optional` **transcribeAudio?**: `boolean`
+
+Not implemented yet (#433) — warns rather than silently doing nothing.
+
+---
+
 ### enableSummarization?
 
 > `optional` **enableSummarization?**: `boolean`

docs/api/type-aliases/WebSocketRequestId.md

@@ -0,0 +1,15 @@
+[**NeuroLink API Reference**](../README.md)
+
+---
+
+[NeuroLink API Reference](../README.md) / WebSocketRequestId
+
+# Type Alias: WebSocketRequestId
+
+> **WebSocketRequestId** = `string` \| `number`
+
+Client-supplied correlation id on an agent-protocol WebSocket message
+(`{ type, payload, id }`). Echoed back on the matching response frame so a
+caller with more than one in-flight message on a connection can tell which
+frame answers which request; optional and ignored entirely by a client
+that never sends one.

docs/guides/server-adapters/websocket.md

@@ -634,7 +634,7 @@ ws.onmessage = (event) => {
       console.log("Connected:", message.connectionId);
       break;
     case "response":
-      console.log("Response:", message.data);
+      console.log("Response:", message.content);
       break;
     case "stream_start":
       console.log("Stream starting...");

package.json

@@ -111,6 +111,7 @@
     "test:middleware": "pnpm exec tsx test/continuous-test-suite-middleware.ts",
     "test:stream-middleware": "pnpm exec tsx test/continuous-test-suite-stream-middleware.ts",
     "test:stream-tool-telemetry": "pnpm exec tsx test/continuous-test-suite-stream-tool-telemetry.ts",
+    "test:stream-teardown-abort": "pnpm exec tsx test/continuous-test-suite-stream-teardown-abort.ts",
     "test:observability": "pnpm exec tsx test/continuous-test-suite-observability.ts",
     "test:ppt": "pnpm exec tsx test/continuous-test-suite-ppt.ts",
     "test:vertex-loop-characterization": "tsx test/continuous-test-suite-vertex-loop-characterization.ts",
@@ -127,6 +128,7 @@
     "test:mcp:infra": "pnpm exec tsx test/continuous-test-suite-mcp-infra.ts",
     "test:vendor-recovery": "pnpm exec tsx test/continuous-test-suite-native-vendor-recovery.ts",
     "test:anthropic-generate-guard": "pnpm exec tsx test/continuous-test-suite-anthropic-generate-guard.ts",
+    "test:stream-structured-output": "pnpm exec tsx test/continuous-test-suite-stream-structured-output.ts",
     "test:browser-bundle": "pnpm exec tsx test/continuous-test-suite-browser-bundle.ts",
     "test:providers-mocked": "pnpm exec tsx test/continuous-test-suite-providers-mocked.ts",
     "scaffold:provider": "pnpm exec tsx tools/scaffold-provider.ts",
@@ -144,8 +146,10 @@
     "test:decide": "pnpm exec tsx test/continuous-test-suite-decide.ts",
     "test:agent-mode": "pnpm exec tsx test/continuous-test-suite-agent-mode.ts",
     "test:servers": "pnpm exec tsx test/continuous-test-suite-servers.ts",
+    "test:websocket-agent-handler": "pnpm exec tsx test/continuous-test-suite-websocket-agent-handler.ts",
     "test:tool-reliability": "pnpm exec tsx test/continuous-test-suite-tool-reliability.ts",
     "test:video-abort": "pnpm exec tsx test/continuous-test-suite-video-abort.ts",
+    "test:video-options-allowlist": "pnpm exec tsx test/continuous-test-suite-video-options-allowlist.ts",
     "test:tts": "pnpm exec tsx test/continuous-test-suite-tts.ts",
     "test:tts:unit": "pnpm exec tsx test/continuous-test-suite-tts-unit.ts",
     "test:stt:unit": "pnpm exec tsx test/continuous-test-suite-stt-unit.ts",

src/cli/loop/optionsSchema.ts

@@ -28,6 +28,7 @@ export const textGenerationOptionsSchema: Record<
     | "csvOptions"
     | "pdfOptions"
     | "imageOptions"
+    | "videoOptions" // Complex object, not set via simple CLI commands
     | "tts"
     | "stt" // Complex object, set via --stt* flags
     | "thinkingConfig" // Complex object, use thinking/thinkingBudget instead