@athenaintel/react 0.23.3 → 0.24.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +6 -0
- package/dist/assets/AssetPanel.cjs +1 -1
- package/dist/assets/AssetPanel.cjs.map +1 -1
- package/dist/assets/AssetPanel.js +1 -1
- package/dist/assets/AssetPanel.js.map +1 -1
- package/dist/assets/AthenaAssetEmbed.cjs +1 -1
- package/dist/assets/AthenaAssetEmbed.cjs.map +1 -1
- package/dist/assets/AthenaAssetEmbed.js +1 -1
- package/dist/assets/AthenaAssetEmbed.js.map +1 -1
- package/dist/chat/AthenaChat.cjs +15 -23
- package/dist/chat/AthenaChat.cjs.map +1 -1
- package/dist/chat/AthenaChat.d.ts +9 -4
- package/dist/chat/AthenaChat.js +15 -23
- package/dist/chat/AthenaChat.js.map +1 -1
- package/dist/chat/StatewireApprovalCard.cjs +3 -18
- package/dist/chat/StatewireApprovalCard.cjs.map +1 -1
- package/dist/chat/StatewireApprovalCard.js +3 -18
- package/dist/chat/StatewireApprovalCard.js.map +1 -1
- package/dist/chat/StatewireClientToolBridge.cjs +2 -16
- package/dist/chat/StatewireClientToolBridge.cjs.map +1 -1
- package/dist/chat/StatewireClientToolBridge.js +2 -16
- package/dist/chat/StatewireClientToolBridge.js.map +1 -1
- package/dist/chat/StatewireQueuedMessages.cjs +5 -16
- package/dist/chat/StatewireQueuedMessages.cjs.map +1 -1
- package/dist/chat/StatewireQueuedMessages.js +5 -16
- package/dist/chat/StatewireQueuedMessages.js.map +1 -1
- package/dist/chat/WaitingOnYouBanner.cjs +3 -18
- package/dist/chat/WaitingOnYouBanner.cjs.map +1 -1
- package/dist/chat/WaitingOnYouBanner.js +3 -18
- package/dist/chat/WaitingOnYouBanner.js.map +1 -1
- package/dist/chat/run-activity.cjs +0 -16
- package/dist/chat/run-activity.cjs.map +1 -1
- package/dist/chat/run-activity.js +0 -16
- package/dist/chat/run-activity.js.map +1 -1
- package/dist/chat/statewire-client-tools.cjs +1 -17
- package/dist/chat/statewire-client-tools.cjs.map +1 -1
- package/dist/chat/statewire-client-tools.js +1 -17
- package/dist/chat/statewire-client-tools.js.map +1 -1
- package/dist/chat/super-grouping/activity-timeline.cjs +0 -16
- package/dist/chat/super-grouping/activity-timeline.cjs.map +1 -1
- package/dist/chat/super-grouping/activity-timeline.js +0 -16
- package/dist/chat/super-grouping/activity-timeline.js.map +1 -1
- package/dist/chat/tiptap-text.cjs +1 -17
- package/dist/chat/tiptap-text.cjs.map +1 -1
- package/dist/chat/tiptap-text.js +1 -17
- package/dist/chat/tiptap-text.js.map +1 -1
- package/dist/chat/use-athena-run-live.cjs +0 -15
- package/dist/chat/use-athena-run-live.cjs.map +1 -1
- package/dist/chat/use-athena-run-live.js +0 -15
- package/dist/chat/use-athena-run-live.js.map +1 -1
- package/dist/chat/use-statewire-approval-request.cjs +4 -16
- package/dist/chat/use-statewire-approval-request.cjs.map +1 -1
- package/dist/chat/use-statewire-approval-request.d.ts +2 -0
- package/dist/chat/use-statewire-approval-request.js +4 -16
- package/dist/chat/use-statewire-approval-request.js.map +1 -1
- package/dist/chat-ui/composer/tiptap-composer.cjs +1 -16
- package/dist/chat-ui/composer/tiptap-composer.cjs.map +1 -1
- package/dist/chat-ui/composer/tiptap-composer.js +1 -16
- package/dist/chat-ui/composer/tiptap-composer.js.map +1 -1
- package/dist/chat-ui/index.cjs +0 -16
- package/dist/chat-ui/index.cjs.map +1 -1
- package/dist/chat-ui/index.js +0 -16
- package/dist/chat-ui/index.js.map +1 -1
- package/dist/chat-ui/mentions/mention-chip.cjs +1 -16
- package/dist/chat-ui/mentions/mention-chip.cjs.map +1 -1
- package/dist/chat-ui/mentions/mention-chip.js +1 -16
- package/dist/chat-ui/mentions/mention-chip.js.map +1 -1
- package/dist/chat-ui/mentions/suggestions/data/sources/asset-type-icons.cjs +4 -6
- package/dist/chat-ui/mentions/suggestions/data/sources/asset-type-icons.cjs.map +1 -1
- package/dist/chat-ui/mentions/suggestions/data/sources/asset-type-icons.js +4 -6
- package/dist/chat-ui/mentions/suggestions/data/sources/asset-type-icons.js.map +1 -1
- package/dist/chat-ui/mentions/suggestions/data/sources/index.cjs +2 -9
- package/dist/chat-ui/mentions/suggestions/data/sources/index.cjs.map +1 -1
- package/dist/chat-ui/mentions/suggestions/data/sources/index.js +2 -9
- package/dist/chat-ui/mentions/suggestions/data/sources/index.js.map +1 -1
- package/dist/chat-ui/messages/message-components.cjs +5 -19
- package/dist/chat-ui/messages/message-components.cjs.map +1 -1
- package/dist/chat-ui/messages/message-components.js +5 -19
- package/dist/chat-ui/messages/message-components.js.map +1 -1
- package/dist/chat-ui/messages/runtime-message.cjs +1 -16
- package/dist/chat-ui/messages/runtime-message.cjs.map +1 -1
- package/dist/chat-ui/messages/runtime-message.js +1 -16
- package/dist/chat-ui/messages/runtime-message.js.map +1 -1
- package/dist/chat-ui/messages/statewire-continue-run.cjs +3 -15
- package/dist/chat-ui/messages/statewire-continue-run.cjs.map +1 -1
- package/dist/chat-ui/messages/statewire-continue-run.js +3 -15
- package/dist/chat-ui/messages/statewire-continue-run.js.map +1 -1
- package/dist/chat-ui/messages/user-edit-composer.cjs +0 -16
- package/dist/chat-ui/messages/user-edit-composer.cjs.map +1 -1
- package/dist/chat-ui/messages/user-edit-composer.js +0 -16
- package/dist/chat-ui/messages/user-edit-composer.js.map +1 -1
- package/dist/chat-ui/messages/wake-marker.cjs +1 -16
- package/dist/chat-ui/messages/wake-marker.cjs.map +1 -1
- package/dist/chat-ui/messages/wake-marker.js +1 -16
- package/dist/chat-ui/messages/wake-marker.js.map +1 -1
- package/dist/chat-ui/statewire.cjs +0 -16
- package/dist/chat-ui/statewire.cjs.map +1 -1
- package/dist/chat-ui/statewire.js +0 -16
- package/dist/chat-ui/statewire.js.map +1 -1
- package/dist/chat-ui/thread/statewire-history-edge.cjs +2 -17
- package/dist/chat-ui/thread/statewire-history-edge.cjs.map +1 -1
- package/dist/chat-ui/thread/statewire-history-edge.js +2 -17
- package/dist/chat-ui/thread/statewire-history-edge.js.map +1 -1
- package/dist/dictation/DictationStrip.cjs +2 -1
- package/dist/dictation/DictationStrip.cjs.map +1 -1
- package/dist/dictation/DictationStrip.d.ts +2 -0
- package/dist/dictation/DictationStrip.js +2 -1
- package/dist/dictation/DictationStrip.js.map +1 -1
- package/dist/dictation/dictation-engine.cjs.map +1 -1
- package/dist/dictation/dictation-engine.d.ts +8 -0
- package/dist/dictation/dictation-engine.js.map +1 -1
- package/dist/dictation/recorded-dictation-engine.cjs +153 -0
- package/dist/dictation/recorded-dictation-engine.cjs.map +1 -0
- package/dist/dictation/recorded-dictation-engine.d.ts +66 -0
- package/dist/dictation/recorded-dictation-engine.js +153 -0
- package/dist/dictation/recorded-dictation-engine.js.map +1 -0
- package/dist/dictation/use-athena-dictation-engine.cjs +64 -0
- package/dist/dictation/use-athena-dictation-engine.cjs.map +1 -0
- package/dist/dictation/use-athena-dictation-engine.d.ts +25 -0
- package/dist/dictation/use-athena-dictation-engine.js +64 -0
- package/dist/dictation/use-athena-dictation-engine.js.map +1 -0
- package/dist/dictation/use-dictation.cjs +17 -6
- package/dist/dictation/use-dictation.cjs.map +1 -1
- package/dist/dictation/use-dictation.d.ts +20 -4
- package/dist/dictation/use-dictation.js +17 -6
- package/dist/dictation/use-dictation.js.map +1 -1
- package/dist/index.cjs +28 -39
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.ts +4 -1
- package/dist/index.js +10 -21
- package/dist/index.js.map +1 -1
- package/dist/packages/agent-runtime/src/edit-command.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/edit-command.js.map +1 -1
- package/dist/packages/agent-runtime/src/errors.cjs +18 -0
- package/dist/packages/agent-runtime/src/errors.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/errors.js +19 -1
- package/dist/packages/agent-runtime/src/errors.js.map +1 -1
- package/dist/packages/agent-runtime/src/generated/runconfig-contract.cjs +1 -0
- package/dist/packages/agent-runtime/src/generated/runconfig-contract.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/generated/runconfig-contract.js +1 -0
- package/dist/packages/agent-runtime/src/generated/runconfig-contract.js.map +1 -1
- package/dist/packages/agent-runtime/src/interrupts.cjs +61 -12
- package/dist/packages/agent-runtime/src/interrupts.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/interrupts.js +61 -12
- package/dist/packages/agent-runtime/src/interrupts.js.map +1 -1
- package/dist/packages/agent-runtime/src/quota-safe-local-storage-session.cjs +49 -0
- package/dist/packages/agent-runtime/src/quota-safe-local-storage-session.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/quota-safe-local-storage-session.js +49 -0
- package/dist/packages/agent-runtime/src/quota-safe-local-storage-session.js.map +1 -1
- package/dist/packages/agent-runtime/src/thread-status.js +1 -1
- package/dist/packages/agent-runtime/src/tool-metadata.cjs +1 -37
- package/dist/packages/agent-runtime/src/tool-metadata.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/tool-metadata.js +1 -37
- package/dist/packages/agent-runtime/src/tool-metadata.js.map +1 -1
- package/dist/packages/agent-runtime/src/use-deep-agent-config.cjs +0 -1
- package/dist/packages/agent-runtime/src/use-deep-agent-config.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/use-deep-agent-config.js +0 -1
- package/dist/packages/agent-runtime/src/use-deep-agent-config.js.map +1 -1
- package/dist/packages/agent-runtime/src/use-deep-agent-thread.cjs.map +1 -1
- package/dist/packages/agent-runtime/src/use-deep-agent-thread.js.map +1 -1
- package/dist/packages/react/package.json.cjs +1 -1
- package/dist/packages/react/package.json.js +1 -1
- package/dist/provider/AthenaProvider.cjs +3 -17
- package/dist/provider/AthenaProvider.cjs.map +1 -1
- package/dist/provider/AthenaProvider.js +3 -17
- package/dist/provider/AthenaProvider.js.map +1 -1
- package/dist/runtime/statewire-lifecycle.cjs +0 -16
- package/dist/runtime/statewire-lifecycle.cjs.map +1 -1
- package/dist/runtime/statewire-lifecycle.js +0 -16
- package/dist/runtime/statewire-lifecycle.js.map +1 -1
- package/dist/runtime/useAthenaStatewireRuntime.cjs +3 -17
- package/dist/runtime/useAthenaStatewireRuntime.cjs.map +1 -1
- package/dist/runtime/useAthenaStatewireRuntime.js +3 -17
- package/dist/runtime/useAthenaStatewireRuntime.js.map +1 -1
- package/dist/threads/statewire-thread-list.cjs +4 -17
- package/dist/threads/statewire-thread-list.cjs.map +1 -1
- package/dist/threads/statewire-thread-list.js +4 -17
- package/dist/threads/statewire-thread-list.js.map +1 -1
- package/dist/threads/use-rest-conversation-history.cjs +2 -16
- package/dist/threads/use-rest-conversation-history.cjs.map +1 -1
- package/dist/threads/use-rest-conversation-history.js +2 -16
- package/dist/threads/use-rest-conversation-history.js.map +1 -1
- package/dist/threads/use-statewire-conversation-history.cjs +2 -16
- package/dist/threads/use-statewire-conversation-history.cjs.map +1 -1
- package/dist/threads/use-statewire-conversation-history.js +2 -16
- package/dist/threads/use-statewire-conversation-history.js.map +1 -1
- package/dist/toolkits/index.cjs +0 -7
- package/dist/toolkits/index.cjs.map +1 -1
- package/dist/toolkits/index.d.ts +0 -2
- package/dist/toolkits/index.js +0 -7
- package/dist/toolkits/index.js.map +1 -1
- package/dist/tools/index.cjs +4 -20
- package/dist/tools/index.cjs.map +1 -1
- package/dist/tools/index.js +1 -17
- package/dist/tools/index.js.map +1 -1
- package/dist/tools/misc-tool-uis.cjs +2 -17
- package/dist/tools/misc-tool-uis.cjs.map +1 -1
- package/dist/tools/misc-tool-uis.js +2 -17
- package/dist/tools/misc-tool-uis.js.map +1 -1
- package/dist/tools/primitives.cjs +1 -16
- package/dist/tools/primitives.cjs.map +1 -1
- package/dist/tools/primitives.js +1 -16
- package/dist/tools/primitives.js.map +1 -1
- package/dist/tools/tool-fallback.cjs +2 -17
- package/dist/tools/tool-fallback.cjs.map +1 -1
- package/dist/tools/tool-fallback.js +2 -17
- package/dist/tools/tool-fallback.js.map +1 -1
- package/dist/tools/voice-update-tool-ui.cjs +1 -17
- package/dist/tools/voice-update-tool-ui.cjs.map +1 -1
- package/dist/tools/voice-update-tool-ui.js +1 -17
- package/dist/tools/voice-update-tool-ui.js.map +1 -1
- package/package.json +1 -1
- package/dist/node_modules/qrcode-generator/dist/qrcode.cjs +0 -32
- package/dist/node_modules/qrcode-generator/dist/qrcode.cjs.map +0 -1
- package/dist/node_modules/qrcode-generator/dist/qrcode.js +0 -31
- package/dist/node_modules/qrcode-generator/dist/qrcode.js.map +0 -1
- package/dist/node_modules/zod/v4/classic/errors.cjs +0 -44
- package/dist/node_modules/zod/v4/classic/errors.cjs.map +0 -1
- package/dist/node_modules/zod/v4/classic/errors.js +0 -44
- package/dist/node_modules/zod/v4/classic/errors.js.map +0 -1
- package/dist/node_modules/zod/v4/classic/iso.cjs +0 -43
- package/dist/node_modules/zod/v4/classic/iso.cjs.map +0 -1
- package/dist/node_modules/zod/v4/classic/iso.js +0 -43
- package/dist/node_modules/zod/v4/classic/iso.js.map +0 -1
- package/dist/node_modules/zod/v4/classic/parse.cjs +0 -29
- package/dist/node_modules/zod/v4/classic/parse.cjs.map +0 -1
- package/dist/node_modules/zod/v4/classic/parse.js +0 -29
- package/dist/node_modules/zod/v4/classic/parse.js.map +0 -1
- package/dist/node_modules/zod/v4/classic/schemas.cjs +0 -871
- package/dist/node_modules/zod/v4/classic/schemas.cjs.map +0 -1
- package/dist/node_modules/zod/v4/classic/schemas.js +0 -871
- package/dist/node_modules/zod/v4/classic/schemas.js.map +0 -1
- package/dist/node_modules/zod/v4/core/api.cjs +0 -570
- package/dist/node_modules/zod/v4/core/api.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/api.js +0 -570
- package/dist/node_modules/zod/v4/core/api.js.map +0 -1
- package/dist/node_modules/zod/v4/core/checks.cjs +0 -411
- package/dist/node_modules/zod/v4/core/checks.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/checks.js +0 -411
- package/dist/node_modules/zod/v4/core/checks.js.map +0 -1
- package/dist/node_modules/zod/v4/core/core.cjs +0 -77
- package/dist/node_modules/zod/v4/core/core.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/core.js +0 -77
- package/dist/node_modules/zod/v4/core/core.js.map +0 -1
- package/dist/node_modules/zod/v4/core/doc.cjs +0 -38
- package/dist/node_modules/zod/v4/core/doc.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/doc.js +0 -38
- package/dist/node_modules/zod/v4/core/doc.js.map +0 -1
- package/dist/node_modules/zod/v4/core/errors.cjs +0 -76
- package/dist/node_modules/zod/v4/core/errors.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/errors.js +0 -76
- package/dist/node_modules/zod/v4/core/errors.js.map +0 -1
- package/dist/node_modules/zod/v4/core/json-schema-processors.cjs +0 -312
- package/dist/node_modules/zod/v4/core/json-schema-processors.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/json-schema-processors.js +0 -312
- package/dist/node_modules/zod/v4/core/json-schema-processors.js.map +0 -1
- package/dist/node_modules/zod/v4/core/parse.cjs +0 -96
- package/dist/node_modules/zod/v4/core/parse.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/parse.js +0 -96
- package/dist/node_modules/zod/v4/core/parse.js.map +0 -1
- package/dist/node_modules/zod/v4/core/regexes.cjs +0 -86
- package/dist/node_modules/zod/v4/core/regexes.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/regexes.js +0 -86
- package/dist/node_modules/zod/v4/core/regexes.js.map +0 -1
- package/dist/node_modules/zod/v4/core/registries.cjs +0 -52
- package/dist/node_modules/zod/v4/core/registries.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/registries.js +0 -52
- package/dist/node_modules/zod/v4/core/registries.js.map +0 -1
- package/dist/node_modules/zod/v4/core/schemas.cjs +0 -1412
- package/dist/node_modules/zod/v4/core/schemas.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/schemas.js +0 -1413
- package/dist/node_modules/zod/v4/core/schemas.js.map +0 -1
- package/dist/node_modules/zod/v4/core/to-json-schema.cjs +0 -370
- package/dist/node_modules/zod/v4/core/to-json-schema.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/to-json-schema.js +0 -370
- package/dist/node_modules/zod/v4/core/to-json-schema.js.map +0 -1
- package/dist/node_modules/zod/v4/core/util.cjs +0 -444
- package/dist/node_modules/zod/v4/core/util.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/util.js +0 -444
- package/dist/node_modules/zod/v4/core/util.js.map +0 -1
- package/dist/node_modules/zod/v4/core/versions.cjs +0 -9
- package/dist/node_modules/zod/v4/core/versions.cjs.map +0 -1
- package/dist/node_modules/zod/v4/core/versions.js +0 -9
- package/dist/node_modules/zod/v4/core/versions.js.map +0 -1
- package/dist/packages/agent-runtime/src/thread-handoff.cjs +0 -85
- package/dist/packages/agent-runtime/src/thread-handoff.cjs.map +0 -1
- package/dist/packages/agent-runtime/src/thread-handoff.js +0 -85
- package/dist/packages/agent-runtime/src/thread-handoff.js.map +0 -1
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"statewire-history-edge.js","sources":["../../../src/chat-ui/thread/statewire-history-edge.tsx"],"sourcesContent":["import { useAuiState } from \"@assistant-ui/react\";\nimport {\n type DeepAgentHistoryLoadEvent,\n type DeepAgentToolOutputStore,\n isStatewireChannelTeardown,\n useDeepAgentWindowSync,\n} from \"@athenaintel/agent-runtime\";\nimport type { FC } from \"react\";\nimport { useEffect, useLayoutEffect, useRef, useState } from \"react\";\nimport { cn } from \"../../lib/utils\";\nimport { useStatewireToolOutputs } from \"../../provider/statewire-window-context\";\n\n/** The attribute `ChatThreadShell` stamps on the scroll viewport. */\nexport const THREAD_VIEWPORT_ATTR = \"data-aui-thread-viewport\";\n\n/**\n * How far above the viewport the edge row counts as near: the row sits above\n * the first loaded message, so reaching it means the user has scrolled past\n * everything loaded and the next page should already be on its way.\n */\nexport const HISTORY_EDGE_NEAR_MARGIN_PX = 1500;\n\nexport interface HistoryEdgeRowProps {\n /** Whether the server holds messages older than the loaded window. */\n hasOlder: boolean;\n /** Whether the next page can be asked for; `false` with `hasOlder` is the connection's loading depth spent. Defaults to `hasOlder`. */\n canLoadOlder?: boolean;\n /** A page of older history is in flight. */\n loading: boolean;\n onLoadOlder: () => void;\n className?: string;\n /** The row element, for the viewport observer. */\n rowRef?: (element: HTMLElement | null) => void;\n}\n\n/**\n * The top edge of the LOADED history: a manual way to load the next page of\n * older messages, and the in-flight state. Renders nothing when the whole\n * transcript is loaded (or the head interest has not resolved).\n */\nexport const HistoryEdgeRow: FC<HistoryEdgeRowProps> = ({\n hasOlder,\n canLoadOlder = hasOlder,\n loading,\n onLoadOlder,\n className,\n rowRef,\n}) => {\n if (!hasOlder) return null;\n return (\n <div\n ref={rowRef}\n data-testid=\"thread-history-edge\"\n className={cn(\n \"flex w-full max-w-(--thread-max-width) items-center justify-center py-1 text-xs text-muted-foreground\",\n className,\n )}\n >\n {loading ? (\n <span data-testid=\"thread-history-edge-loading\">Loading earlier messages…</span>\n ) : canLoadOlder ? (\n <button\n type=\"button\"\n data-testid=\"thread-history-edge-load\"\n onClick={onLoadOlder}\n className=\"rounded-lg px-2.5 py-1 text-xs font-medium text-muted-foreground transition-colors hover:bg-accent hover:text-foreground\"\n >\n Load earlier messages\n </button>\n ) : (\n <span data-testid=\"thread-history-edge-limit\">Earlier messages can't be loaded here</span>\n )}\n </div>\n );\n};\n\nexport interface StatewireHistoryEdgeProps {\n /**\n * The thread's tool-output store (the one its converter substitutes from).\n * Defaults to `AthenaProvider`'s; hosts mounting the runtime themselves pass\n * the store they gave `useDeepAgentRuntime({ toolOutputs })`.\n */\n toolOutputs?: DeepAgentToolOutputStore;\n className?: string;\n /** A page interest resolved (telemetry). */\n onHistoryLoad?: (event: DeepAgentHistoryLoadEvent) => void;\n}\n\n/**\n * The client half of the interest window for a DOM thread — mounted once per\n * thread, inside the scroll viewport above the messages (`ChatThreadShell`'s\n * `beforeMessages` slot):\n *\n * - renders the \"Load earlier messages\" edge row and loads the next page when\n * the row comes near the viewport (scrolling up) or on its button, keeping\n * the reading position when the page lands above the view;\n * - fetches cut tool outputs whole for the rendered cards, so they re-render\n * complete through the converter (`toolOutputs`).\n *\n * Inert until the head interest resolves: no row, no reads.\n */\nexport const StatewireHistoryEdge: FC<StatewireHistoryEdgeProps> = ({\n toolOutputs: toolOutputsProp,\n className,\n onHistoryLoad,\n}) => {\n const providerToolOutputs = useStatewireToolOutputs();\n const toolOutputs = toolOutputsProp ?? providerToolOutputs ?? undefined;\n const messages = useAuiState((s) => s.thread.messages);\n const [nearTop, setNearTop] = useState(false);\n const [nearTopEpoch, setNearTopEpoch] = useState(0);\n const nearTopRef = useRef(false);\n const [rowEl, setRowEl] = useState<HTMLElement | null>(null);\n const { hasOlder, canLoadOlder, loading, loadOlder } = useDeepAgentWindowSync({\n messages,\n toolOutputs,\n nearTop,\n nearTopEpoch,\n onHistoryLoad,\n onError: (error, command) => {\n // Leaving the thread settles an in-flight read as lost: lifecycle, not a failure.\n if (error instanceof Error && isStatewireChannelTeardown(error)) return;\n console.warn(`[StatewireHistoryEdge] ${command} failed:`, error);\n },\n });\n\n // The row's own visibility is the near-top level: a level, not an edge, so\n // pages chain while the user stays at the top and stop once the prepended\n // content has pushed the row away again.\n useEffect(() => {\n if (!rowEl || !hasOlder || typeof IntersectionObserver === \"undefined\") return;\n const observer = new IntersectionObserver(\n (entries) => {\n for (const entry of entries) {\n const near = entry.isIntersecting;\n if (nearTopRef.current === near) continue;\n nearTopRef.current = near;\n setNearTop(near);\n if (near) setNearTopEpoch((epoch) => epoch + 1);\n }\n },\n {\n root: rowEl.closest<HTMLElement>(`[${THREAD_VIEWPORT_ATTR}]`),\n rootMargin: `${HISTORY_EDGE_NEAR_MARGIN_PX}px 0px`,\n },\n );\n observer.observe(rowEl);\n return () => {\n observer.disconnect();\n nearTopRef.current = false;\n setNearTop(false);\n };\n }, [rowEl, hasOlder]);\n\n // A prepend (older history landing) keeps the reading position: the\n // viewport scrolls by the height the page added above the view, so the\n // message the user was on stays put. The pre-prepend height is measured\n // only while a page is in flight (a layout read per render would force a\n // reflow on every streamed token otherwise).\n const shapeRef = useRef<{ firstId: string | undefined; length: number; scrollHeight: number } | null>(\n null,\n );\n const firstId = messages[0]?.id;\n const length = messages.length;\n useLayoutEffect(() => {\n const viewport = rowEl?.closest<HTMLElement>(`[${THREAD_VIEWPORT_ATTR}]`) ?? null;\n const prev = shapeRef.current;\n const prepended =\n viewport !== null &&\n prev !== null &&\n prev.firstId !== undefined &&\n firstId !== prev.firstId &&\n length > prev.length;\n if (prepended) viewport.scrollTop += viewport.scrollHeight - prev.scrollHeight;\n shapeRef.current =\n loading || prepended || prev === null\n ? { firstId, length, scrollHeight: viewport?.scrollHeight ?? 0 }\n : { firstId, length, scrollHeight: prev.scrollHeight };\n });\n\n return (\n <HistoryEdgeRow\n rowRef={setRowEl}\n hasOlder={hasOlder}\n canLoadOlder={canLoadOlder}\n loading={loading}\n onLoadOlder={loadOlder}\n className={className}\n />\n );\n};\n"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"statewire-history-edge.js","sources":["../../../src/chat-ui/thread/statewire-history-edge.tsx"],"sourcesContent":["import { useAuiState } from \"@assistant-ui/react\";\nimport {\n type DeepAgentHistoryLoadEvent,\n type DeepAgentToolOutputStore,\n isStatewireChannelTeardown,\n useDeepAgentWindowSync,\n} from \"@athenaintel/agent-runtime\";\nimport type { FC } from \"react\";\nimport { useEffect, useLayoutEffect, useRef, useState } from \"react\";\nimport { cn } from \"../../lib/utils\";\nimport { useStatewireToolOutputs } from \"../../provider/statewire-window-context\";\n\n/** The attribute `ChatThreadShell` stamps on the scroll viewport. */\nexport const THREAD_VIEWPORT_ATTR = \"data-aui-thread-viewport\";\n\n/**\n * How far above the viewport the edge row counts as near: the row sits above\n * the first loaded message, so reaching it means the user has scrolled past\n * everything loaded and the next page should already be on its way.\n */\nexport const HISTORY_EDGE_NEAR_MARGIN_PX = 1500;\n\nexport interface HistoryEdgeRowProps {\n /** Whether the server holds messages older than the loaded window. */\n hasOlder: boolean;\n /** Whether the next page can be asked for; `false` with `hasOlder` is the connection's loading depth spent. Defaults to `hasOlder`. */\n canLoadOlder?: boolean;\n /** A page of older history is in flight. */\n loading: boolean;\n onLoadOlder: () => void;\n className?: string;\n /** The row element, for the viewport observer. */\n rowRef?: (element: HTMLElement | null) => void;\n}\n\n/**\n * The top edge of the LOADED history: a manual way to load the next page of\n * older messages, and the in-flight state. Renders nothing when the whole\n * transcript is loaded (or the head interest has not resolved).\n */\nexport const HistoryEdgeRow: FC<HistoryEdgeRowProps> = ({\n hasOlder,\n canLoadOlder = hasOlder,\n loading,\n onLoadOlder,\n className,\n rowRef,\n}) => {\n if (!hasOlder) return null;\n return (\n <div\n ref={rowRef}\n data-testid=\"thread-history-edge\"\n className={cn(\n \"flex w-full max-w-(--thread-max-width) items-center justify-center py-1 text-xs text-muted-foreground\",\n className,\n )}\n >\n {loading ? (\n <span data-testid=\"thread-history-edge-loading\">Loading earlier messages…</span>\n ) : canLoadOlder ? (\n <button\n type=\"button\"\n data-testid=\"thread-history-edge-load\"\n onClick={onLoadOlder}\n className=\"rounded-lg px-2.5 py-1 text-xs font-medium text-muted-foreground transition-colors hover:bg-accent hover:text-foreground\"\n >\n Load earlier messages\n </button>\n ) : (\n <span data-testid=\"thread-history-edge-limit\">Earlier messages can't be loaded here</span>\n )}\n </div>\n );\n};\n\nexport interface StatewireHistoryEdgeProps {\n /**\n * The thread's tool-output store (the one its converter substitutes from).\n * Defaults to `AthenaProvider`'s; hosts mounting the runtime themselves pass\n * the store they gave `useDeepAgentRuntime({ toolOutputs })`.\n */\n toolOutputs?: DeepAgentToolOutputStore;\n className?: string;\n /** A page interest resolved (telemetry). */\n onHistoryLoad?: (event: DeepAgentHistoryLoadEvent) => void;\n}\n\n/**\n * The client half of the interest window for a DOM thread — mounted once per\n * thread, inside the scroll viewport above the messages (`ChatThreadShell`'s\n * `beforeMessages` slot):\n *\n * - renders the \"Load earlier messages\" edge row and loads the next page when\n * the row comes near the viewport (scrolling up) or on its button, keeping\n * the reading position when the page lands above the view;\n * - fetches cut tool outputs whole for the rendered cards, so they re-render\n * complete through the converter (`toolOutputs`).\n *\n * Inert until the head interest resolves: no row, no reads.\n */\nexport const StatewireHistoryEdge: FC<StatewireHistoryEdgeProps> = ({\n toolOutputs: toolOutputsProp,\n className,\n onHistoryLoad,\n}) => {\n const providerToolOutputs = useStatewireToolOutputs();\n const toolOutputs = toolOutputsProp ?? providerToolOutputs ?? undefined;\n const messages = useAuiState((s) => s.thread.messages);\n const [nearTop, setNearTop] = useState(false);\n const [nearTopEpoch, setNearTopEpoch] = useState(0);\n const nearTopRef = useRef(false);\n const [rowEl, setRowEl] = useState<HTMLElement | null>(null);\n const { hasOlder, canLoadOlder, loading, loadOlder } = useDeepAgentWindowSync({\n messages,\n toolOutputs,\n nearTop,\n nearTopEpoch,\n onHistoryLoad,\n onError: (error, command) => {\n // Leaving the thread settles an in-flight read as lost: lifecycle, not a failure.\n if (error instanceof Error && isStatewireChannelTeardown(error)) return;\n console.warn(`[StatewireHistoryEdge] ${command} failed:`, error);\n },\n });\n\n // The row's own visibility is the near-top level: a level, not an edge, so\n // pages chain while the user stays at the top and stop once the prepended\n // content has pushed the row away again.\n useEffect(() => {\n if (!rowEl || !hasOlder || typeof IntersectionObserver === \"undefined\") return;\n const observer = new IntersectionObserver(\n (entries) => {\n for (const entry of entries) {\n const near = entry.isIntersecting;\n if (nearTopRef.current === near) continue;\n nearTopRef.current = near;\n setNearTop(near);\n if (near) setNearTopEpoch((epoch) => epoch + 1);\n }\n },\n {\n root: rowEl.closest<HTMLElement>(`[${THREAD_VIEWPORT_ATTR}]`),\n rootMargin: `${HISTORY_EDGE_NEAR_MARGIN_PX}px 0px`,\n },\n );\n observer.observe(rowEl);\n return () => {\n observer.disconnect();\n nearTopRef.current = false;\n setNearTop(false);\n };\n }, [rowEl, hasOlder]);\n\n // A prepend (older history landing) keeps the reading position: the\n // viewport scrolls by the height the page added above the view, so the\n // message the user was on stays put. The pre-prepend height is measured\n // only while a page is in flight (a layout read per render would force a\n // reflow on every streamed token otherwise).\n const shapeRef = useRef<{ firstId: string | undefined; length: number; scrollHeight: number } | null>(\n null,\n );\n const firstId = messages[0]?.id;\n const length = messages.length;\n useLayoutEffect(() => {\n const viewport = rowEl?.closest<HTMLElement>(`[${THREAD_VIEWPORT_ATTR}]`) ?? null;\n const prev = shapeRef.current;\n const prepended =\n viewport !== null &&\n prev !== null &&\n prev.firstId !== undefined &&\n firstId !== prev.firstId &&\n length > prev.length;\n if (prepended) viewport.scrollTop += viewport.scrollHeight - prev.scrollHeight;\n shapeRef.current =\n loading || prepended || prev === null\n ? { firstId, length, scrollHeight: viewport?.scrollHeight ?? 0 }\n : { firstId, length, scrollHeight: prev.scrollHeight };\n });\n\n return (\n <HistoryEdgeRow\n rowRef={setRowEl}\n hasOlder={hasOlder}\n canLoadOlder={canLoadOlder}\n loading={loading}\n onLoadOlder={loadOlder}\n className={className}\n />\n );\n};\n"],"names":[],"mappings":";;;;;;;AAaO,MAAM,uBAAuB;AAO7B,MAAM,8BAA8B;AAoBpC,MAAM,iBAA0C,CAAC;AAAA,EACtD;AAAA,EACA,eAAe;AAAA,EACf;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,MAAM;AACJ,MAAI,CAAC,SAAU,QAAO;AACtB,SACE;AAAA,IAAC;AAAA,IAAA;AAAA,MACC,KAAK;AAAA,MACL,eAAY;AAAA,MACZ,WAAW;AAAA,QACT;AAAA,QACA;AAAA,MAAA;AAAA,MAGD,oBACC,oBAAC,QAAA,EAAK,eAAY,+BAA8B,UAAA,6BAAyB,IACvE,eACF;AAAA,QAAC;AAAA,QAAA;AAAA,UACC,MAAK;AAAA,UACL,eAAY;AAAA,UACZ,SAAS;AAAA,UACT,WAAU;AAAA,UACX,UAAA;AAAA,QAAA;AAAA,MAAA,IAID,oBAAC,QAAA,EAAK,eAAY,6BAA4B,UAAA,wCAAA,CAAqC;AAAA,IAAA;AAAA,EAAA;AAI3F;AA2BO,MAAM,uBAAsD,CAAC;AAAA,EAClE,aAAa;AAAA,EACb;AAAA,EACA;AACF,MAAM;;AACJ,QAAM,sBAAsB,wBAAA;AAC5B,QAAM,cAAc,mBAAmB,uBAAuB;AAC9D,QAAM,WAAW,YAAY,CAAC,MAAM,EAAE,OAAO,QAAQ;AACrD,QAAM,CAAC,SAAS,UAAU,IAAI,SAAS,KAAK;AAC5C,QAAM,CAAC,cAAc,eAAe,IAAI,SAAS,CAAC;AAClD,QAAM,aAAa,OAAO,KAAK;AAC/B,QAAM,CAAC,OAAO,QAAQ,IAAI,SAA6B,IAAI;AAC3D,QAAM,EAAE,UAAU,cAAc,SAAS,UAAA,IAAc,uBAAuB;AAAA,IAC5E;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,SAAS,CAAC,OAAO,YAAY;AAE3B,UAAI,iBAAiB,SAAS,2BAA2B,KAAK,EAAG;AACjE,cAAQ,KAAK,0BAA0B,OAAO,YAAY,KAAK;AAAA,IACjE;AAAA,EAAA,CACD;AAKD,YAAU,MAAM;AACd,QAAI,CAAC,SAAS,CAAC,YAAY,OAAO,yBAAyB,YAAa;AACxE,UAAM,WAAW,IAAI;AAAA,MACnB,CAAC,YAAY;AACX,mBAAW,SAAS,SAAS;AAC3B,gBAAM,OAAO,MAAM;AACnB,cAAI,WAAW,YAAY,KAAM;AACjC,qBAAW,UAAU;AACrB,qBAAW,IAAI;AACf,cAAI,KAAM,iBAAgB,CAAC,UAAU,QAAQ,CAAC;AAAA,QAChD;AAAA,MACF;AAAA,MACA;AAAA,QACE,MAAM,MAAM,QAAqB,IAAI,oBAAoB,GAAG;AAAA,QAC5D,YAAY,GAAG,2BAA2B;AAAA,MAAA;AAAA,IAC5C;AAEF,aAAS,QAAQ,KAAK;AACtB,WAAO,MAAM;AACX,eAAS,WAAA;AACT,iBAAW,UAAU;AACrB,iBAAW,KAAK;AAAA,IAClB;AAAA,EACF,GAAG,CAAC,OAAO,QAAQ,CAAC;AAOpB,QAAM,WAAW;AAAA,IACf;AAAA,EAAA;AAEF,QAAM,WAAU,cAAS,CAAC,MAAV,mBAAa;AAC7B,QAAM,SAAS,SAAS;AACxB,kBAAgB,MAAM;AACpB,UAAM,YAAW,+BAAO,QAAqB,IAAI,oBAAoB,SAAQ;AAC7E,UAAM,OAAO,SAAS;AACtB,UAAM,YACJ,aAAa,QACb,SAAS,QACT,KAAK,YAAY,UACjB,YAAY,KAAK,WACjB,SAAS,KAAK;AAChB,QAAI,UAAW,UAAS,aAAa,SAAS,eAAe,KAAK;AAClE,aAAS,UACP,WAAW,aAAa,SAAS,OAC7B,EAAE,SAAS,QAAQ,eAAc,qCAAU,iBAAgB,MAC3D,EAAE,SAAS,QAAQ,cAAc,KAAK,aAAA;AAAA,EAC9C,CAAC;AAED,SACE;AAAA,IAAC;AAAA,IAAA;AAAA,MACC,QAAQ;AAAA,MACR;AAAA,MACA;AAAA,MACA;AAAA,MACA,aAAa;AAAA,MACb;AAAA,IAAA;AAAA,EAAA;AAGN;"}
|
|
@@ -34,11 +34,12 @@ const DictationStrip = ({
|
|
|
34
34
|
transcript,
|
|
35
35
|
stream,
|
|
36
36
|
bars = 48,
|
|
37
|
+
stoppingLabel = "Finishing…",
|
|
37
38
|
className
|
|
38
39
|
}) => {
|
|
39
40
|
const history = audioLevelHistory.useAudioLevelHistory(stream, { size: bars });
|
|
40
41
|
const caption = [transcript.final, transcript.interim].filter(Boolean).join(" ");
|
|
41
|
-
const statusLabel = status === "starting" ? "Starting microphone…" : status === "stopping" ?
|
|
42
|
+
const statusLabel = status === "starting" ? "Starting microphone…" : status === "stopping" ? stoppingLabel : "Listening…";
|
|
42
43
|
return /* @__PURE__ */ jsxRuntime.jsxs(
|
|
43
44
|
"div",
|
|
44
45
|
{
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"DictationStrip.cjs","sources":["../../src/dictation/DictationStrip.tsx"],"sourcesContent":["import type { FC } from 'react';\nimport { cn } from '../lib/utils';\nimport { type AudioLevelHistory, useAudioLevelHistory } from './audio-level-history';\nimport type { DictationStatus, DictationTranscript } from './dictation-state';\n\nexport interface DictationStripProps {\n status: DictationStatus;\n transcript: DictationTranscript;\n /** Microphone stream to visualize; null draws the resting baseline. */\n stream: MediaStream | null;\n /** Number of bars across the strip. Defaults to 48. */\n bars?: number;\n className?: string;\n}\n\nconst MIN_BAR_HEIGHT_PX = 3;\nconst MAX_BAR_HEIGHT_PX = 24;\n\nfunction barHeight(level: number): number {\n return Math.round(MIN_BAR_HEIGHT_PX + level * (MAX_BAR_HEIGHT_PX - MIN_BAR_HEIGHT_PX));\n}\n\n/**\n * The bars. A rolling level history renders newest-right: unfilled slots are\n * the dotted baseline, filled slots grow with the microphone level.\n */\nexport const DictationWaveform: FC<{ history: AudioLevelHistory; className?: string }> = ({\n history,\n className,\n}) => (\n <div\n className={cn('aui-dictation-waveform flex h-6 items-center gap-[3px]', className)}\n aria-hidden=\"true\"\n >\n {history.map((level, index) => (\n <span\n // Slots are positional; the history is a fixed-size window.\n key={index}\n className={cn(\n 'aui-dictation-bar w-[3px] shrink-0 rounded-full transition-[height] duration-75',\n level === null ? 'bg-muted-foreground/30' : 'bg-foreground/80',\n )}\n style={{ height: level === null ? MIN_BAR_HEIGHT_PX : barHeight(level) }}\n />\n ))}\n </div>\n);\n\n/**\n * The composer's recording mode: a live waveform in place of the text editor\n * and the interim transcript as a caption under it.\n */\nexport const DictationStrip: FC<DictationStripProps> = ({\n status,\n transcript,\n stream,\n bars = 48,\n className,\n}) => {\n const history = useAudioLevelHistory(stream, { size: bars });\n const caption = [transcript.final, transcript.interim].filter(Boolean).join(' ');\n const statusLabel =\n status === 'starting' ? 'Starting microphone…' : status === 'stopping' ?
|
|
1
|
+
{"version":3,"file":"DictationStrip.cjs","sources":["../../src/dictation/DictationStrip.tsx"],"sourcesContent":["import type { FC } from 'react';\nimport { cn } from '../lib/utils';\nimport { type AudioLevelHistory, useAudioLevelHistory } from './audio-level-history';\nimport type { DictationStatus, DictationTranscript } from './dictation-state';\n\nexport interface DictationStripProps {\n status: DictationStatus;\n transcript: DictationTranscript;\n /** Microphone stream to visualize; null draws the resting baseline. */\n stream: MediaStream | null;\n /** Number of bars across the strip. Defaults to 48. */\n bars?: number;\n /** Status while the engine flushes. Defaults to `Finishing…`. */\n stoppingLabel?: string;\n className?: string;\n}\n\nconst MIN_BAR_HEIGHT_PX = 3;\nconst MAX_BAR_HEIGHT_PX = 24;\n\nfunction barHeight(level: number): number {\n return Math.round(MIN_BAR_HEIGHT_PX + level * (MAX_BAR_HEIGHT_PX - MIN_BAR_HEIGHT_PX));\n}\n\n/**\n * The bars. A rolling level history renders newest-right: unfilled slots are\n * the dotted baseline, filled slots grow with the microphone level.\n */\nexport const DictationWaveform: FC<{ history: AudioLevelHistory; className?: string }> = ({\n history,\n className,\n}) => (\n <div\n className={cn('aui-dictation-waveform flex h-6 items-center gap-[3px]', className)}\n aria-hidden=\"true\"\n >\n {history.map((level, index) => (\n <span\n // Slots are positional; the history is a fixed-size window.\n key={index}\n className={cn(\n 'aui-dictation-bar w-[3px] shrink-0 rounded-full transition-[height] duration-75',\n level === null ? 'bg-muted-foreground/30' : 'bg-foreground/80',\n )}\n style={{ height: level === null ? MIN_BAR_HEIGHT_PX : barHeight(level) }}\n />\n ))}\n </div>\n);\n\n/**\n * The composer's recording mode: a live waveform in place of the text editor\n * and the interim transcript as a caption under it.\n */\nexport const DictationStrip: FC<DictationStripProps> = ({\n status,\n transcript,\n stream,\n bars = 48,\n stoppingLabel = 'Finishing…',\n className,\n}) => {\n const history = useAudioLevelHistory(stream, { size: bars });\n const caption = [transcript.final, transcript.interim].filter(Boolean).join(' ');\n const statusLabel =\n status === 'starting' ? 'Starting microphone…' : status === 'stopping' ? stoppingLabel : 'Listening…';\n\n return (\n <div\n className={cn('aui-dictation-strip flex min-h-10 flex-col gap-1 px-1.5 py-2', className)}\n role=\"status\"\n aria-live=\"polite\"\n >\n <div className=\"flex items-center gap-3\">\n <span className=\"aui-dictation-status text-muted-foreground shrink-0 text-xs\">{statusLabel}</span>\n <DictationWaveform history={history} className=\"min-w-0 flex-1 justify-end overflow-hidden\" />\n </div>\n {caption && (\n <p className=\"aui-dictation-caption text-muted-foreground line-clamp-2 text-sm\">{caption}</p>\n )}\n </div>\n );\n};\n"],"names":["jsx","cn","useAudioLevelHistory","jsxs"],"mappings":";;;;;AAiBA,MAAM,oBAAoB;AAC1B,MAAM,oBAAoB;AAE1B,SAAS,UAAU,OAAuB;AACxC,SAAO,KAAK,MAAM,oBAAoB,SAAS,oBAAoB,kBAAkB;AACvF;AAMO,MAAM,oBAA4E,CAAC;AAAA,EACxF;AAAA,EACA;AACF,MACEA,2BAAAA;AAAAA,EAAC;AAAA,EAAA;AAAA,IACC,WAAWC,OAAAA,GAAG,0DAA0D,SAAS;AAAA,IACjF,eAAY;AAAA,IAEX,UAAA,QAAQ,IAAI,CAAC,OAAO,UACnBD,2BAAAA;AAAAA,MAAC;AAAA,MAAA;AAAA,QAGC,WAAWC,OAAAA;AAAAA,UACT;AAAA,UACA,UAAU,OAAO,2BAA2B;AAAA,QAAA;AAAA,QAE9C,OAAO,EAAE,QAAQ,UAAU,OAAO,oBAAoB,UAAU,KAAK,EAAA;AAAA,MAAE;AAAA,MALlE;AAAA,IAAA,CAOR;AAAA,EAAA;AACH;AAOK,MAAM,iBAA0C,CAAC;AAAA,EACtD;AAAA,EACA;AAAA,EACA;AAAA,EACA,OAAO;AAAA,EACP,gBAAgB;AAAA,EAChB;AACF,MAAM;AACJ,QAAM,UAAUC,kBAAAA,qBAAqB,QAAQ,EAAE,MAAM,MAAM;AAC3D,QAAM,UAAU,CAAC,WAAW,OAAO,WAAW,OAAO,EAAE,OAAO,OAAO,EAAE,KAAK,GAAG;AAC/E,QAAM,cACJ,WAAW,aAAa,yBAAyB,WAAW,aAAa,gBAAgB;AAE3F,SACEC,2BAAAA;AAAAA,IAAC;AAAA,IAAA;AAAA,MACC,WAAWF,OAAAA,GAAG,gEAAgE,SAAS;AAAA,MACvF,MAAK;AAAA,MACL,aAAU;AAAA,MAEV,UAAA;AAAA,QAAAE,2BAAAA,KAAC,OAAA,EAAI,WAAU,2BACb,UAAA;AAAA,UAAAH,2BAAAA,IAAC,QAAA,EAAK,WAAU,+DAA+D,UAAA,aAAY;AAAA,UAC3FA,2BAAAA,IAAC,mBAAA,EAAkB,SAAkB,WAAU,6CAAA,CAA6C;AAAA,QAAA,GAC9F;AAAA,QACC,WACCA,2BAAAA,IAAC,KAAA,EAAE,WAAU,oEAAoE,UAAA,QAAA,CAAQ;AAAA,MAAA;AAAA,IAAA;AAAA,EAAA;AAIjG;;;"}
|
|
@@ -8,6 +8,8 @@ export interface DictationStripProps {
|
|
|
8
8
|
stream: MediaStream | null;
|
|
9
9
|
/** Number of bars across the strip. Defaults to 48. */
|
|
10
10
|
bars?: number;
|
|
11
|
+
/** Status while the engine flushes. Defaults to `Finishing…`. */
|
|
12
|
+
stoppingLabel?: string;
|
|
11
13
|
className?: string;
|
|
12
14
|
}
|
|
13
15
|
/**
|
|
@@ -32,11 +32,12 @@ const DictationStrip = ({
|
|
|
32
32
|
transcript,
|
|
33
33
|
stream,
|
|
34
34
|
bars = 48,
|
|
35
|
+
stoppingLabel = "Finishing…",
|
|
35
36
|
className
|
|
36
37
|
}) => {
|
|
37
38
|
const history = useAudioLevelHistory(stream, { size: bars });
|
|
38
39
|
const caption = [transcript.final, transcript.interim].filter(Boolean).join(" ");
|
|
39
|
-
const statusLabel = status === "starting" ? "Starting microphone…" : status === "stopping" ?
|
|
40
|
+
const statusLabel = status === "starting" ? "Starting microphone…" : status === "stopping" ? stoppingLabel : "Listening…";
|
|
40
41
|
return /* @__PURE__ */ jsxs(
|
|
41
42
|
"div",
|
|
42
43
|
{
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"DictationStrip.js","sources":["../../src/dictation/DictationStrip.tsx"],"sourcesContent":["import type { FC } from 'react';\nimport { cn } from '../lib/utils';\nimport { type AudioLevelHistory, useAudioLevelHistory } from './audio-level-history';\nimport type { DictationStatus, DictationTranscript } from './dictation-state';\n\nexport interface DictationStripProps {\n status: DictationStatus;\n transcript: DictationTranscript;\n /** Microphone stream to visualize; null draws the resting baseline. */\n stream: MediaStream | null;\n /** Number of bars across the strip. Defaults to 48. */\n bars?: number;\n className?: string;\n}\n\nconst MIN_BAR_HEIGHT_PX = 3;\nconst MAX_BAR_HEIGHT_PX = 24;\n\nfunction barHeight(level: number): number {\n return Math.round(MIN_BAR_HEIGHT_PX + level * (MAX_BAR_HEIGHT_PX - MIN_BAR_HEIGHT_PX));\n}\n\n/**\n * The bars. A rolling level history renders newest-right: unfilled slots are\n * the dotted baseline, filled slots grow with the microphone level.\n */\nexport const DictationWaveform: FC<{ history: AudioLevelHistory; className?: string }> = ({\n history,\n className,\n}) => (\n <div\n className={cn('aui-dictation-waveform flex h-6 items-center gap-[3px]', className)}\n aria-hidden=\"true\"\n >\n {history.map((level, index) => (\n <span\n // Slots are positional; the history is a fixed-size window.\n key={index}\n className={cn(\n 'aui-dictation-bar w-[3px] shrink-0 rounded-full transition-[height] duration-75',\n level === null ? 'bg-muted-foreground/30' : 'bg-foreground/80',\n )}\n style={{ height: level === null ? MIN_BAR_HEIGHT_PX : barHeight(level) }}\n />\n ))}\n </div>\n);\n\n/**\n * The composer's recording mode: a live waveform in place of the text editor\n * and the interim transcript as a caption under it.\n */\nexport const DictationStrip: FC<DictationStripProps> = ({\n status,\n transcript,\n stream,\n bars = 48,\n className,\n}) => {\n const history = useAudioLevelHistory(stream, { size: bars });\n const caption = [transcript.final, transcript.interim].filter(Boolean).join(' ');\n const statusLabel =\n status === 'starting' ? 'Starting microphone…' : status === 'stopping' ?
|
|
1
|
+
{"version":3,"file":"DictationStrip.js","sources":["../../src/dictation/DictationStrip.tsx"],"sourcesContent":["import type { FC } from 'react';\nimport { cn } from '../lib/utils';\nimport { type AudioLevelHistory, useAudioLevelHistory } from './audio-level-history';\nimport type { DictationStatus, DictationTranscript } from './dictation-state';\n\nexport interface DictationStripProps {\n status: DictationStatus;\n transcript: DictationTranscript;\n /** Microphone stream to visualize; null draws the resting baseline. */\n stream: MediaStream | null;\n /** Number of bars across the strip. Defaults to 48. */\n bars?: number;\n /** Status while the engine flushes. Defaults to `Finishing…`. */\n stoppingLabel?: string;\n className?: string;\n}\n\nconst MIN_BAR_HEIGHT_PX = 3;\nconst MAX_BAR_HEIGHT_PX = 24;\n\nfunction barHeight(level: number): number {\n return Math.round(MIN_BAR_HEIGHT_PX + level * (MAX_BAR_HEIGHT_PX - MIN_BAR_HEIGHT_PX));\n}\n\n/**\n * The bars. A rolling level history renders newest-right: unfilled slots are\n * the dotted baseline, filled slots grow with the microphone level.\n */\nexport const DictationWaveform: FC<{ history: AudioLevelHistory; className?: string }> = ({\n history,\n className,\n}) => (\n <div\n className={cn('aui-dictation-waveform flex h-6 items-center gap-[3px]', className)}\n aria-hidden=\"true\"\n >\n {history.map((level, index) => (\n <span\n // Slots are positional; the history is a fixed-size window.\n key={index}\n className={cn(\n 'aui-dictation-bar w-[3px] shrink-0 rounded-full transition-[height] duration-75',\n level === null ? 'bg-muted-foreground/30' : 'bg-foreground/80',\n )}\n style={{ height: level === null ? MIN_BAR_HEIGHT_PX : barHeight(level) }}\n />\n ))}\n </div>\n);\n\n/**\n * The composer's recording mode: a live waveform in place of the text editor\n * and the interim transcript as a caption under it.\n */\nexport const DictationStrip: FC<DictationStripProps> = ({\n status,\n transcript,\n stream,\n bars = 48,\n stoppingLabel = 'Finishing…',\n className,\n}) => {\n const history = useAudioLevelHistory(stream, { size: bars });\n const caption = [transcript.final, transcript.interim].filter(Boolean).join(' ');\n const statusLabel =\n status === 'starting' ? 'Starting microphone…' : status === 'stopping' ? stoppingLabel : 'Listening…';\n\n return (\n <div\n className={cn('aui-dictation-strip flex min-h-10 flex-col gap-1 px-1.5 py-2', className)}\n role=\"status\"\n aria-live=\"polite\"\n >\n <div className=\"flex items-center gap-3\">\n <span className=\"aui-dictation-status text-muted-foreground shrink-0 text-xs\">{statusLabel}</span>\n <DictationWaveform history={history} className=\"min-w-0 flex-1 justify-end overflow-hidden\" />\n </div>\n {caption && (\n <p className=\"aui-dictation-caption text-muted-foreground line-clamp-2 text-sm\">{caption}</p>\n )}\n </div>\n );\n};\n"],"names":[],"mappings":";;;AAiBA,MAAM,oBAAoB;AAC1B,MAAM,oBAAoB;AAE1B,SAAS,UAAU,OAAuB;AACxC,SAAO,KAAK,MAAM,oBAAoB,SAAS,oBAAoB,kBAAkB;AACvF;AAMO,MAAM,oBAA4E,CAAC;AAAA,EACxF;AAAA,EACA;AACF,MACE;AAAA,EAAC;AAAA,EAAA;AAAA,IACC,WAAW,GAAG,0DAA0D,SAAS;AAAA,IACjF,eAAY;AAAA,IAEX,UAAA,QAAQ,IAAI,CAAC,OAAO,UACnB;AAAA,MAAC;AAAA,MAAA;AAAA,QAGC,WAAW;AAAA,UACT;AAAA,UACA,UAAU,OAAO,2BAA2B;AAAA,QAAA;AAAA,QAE9C,OAAO,EAAE,QAAQ,UAAU,OAAO,oBAAoB,UAAU,KAAK,EAAA;AAAA,MAAE;AAAA,MALlE;AAAA,IAAA,CAOR;AAAA,EAAA;AACH;AAOK,MAAM,iBAA0C,CAAC;AAAA,EACtD;AAAA,EACA;AAAA,EACA;AAAA,EACA,OAAO;AAAA,EACP,gBAAgB;AAAA,EAChB;AACF,MAAM;AACJ,QAAM,UAAU,qBAAqB,QAAQ,EAAE,MAAM,MAAM;AAC3D,QAAM,UAAU,CAAC,WAAW,OAAO,WAAW,OAAO,EAAE,OAAO,OAAO,EAAE,KAAK,GAAG;AAC/E,QAAM,cACJ,WAAW,aAAa,yBAAyB,WAAW,aAAa,gBAAgB;AAE3F,SACE;AAAA,IAAC;AAAA,IAAA;AAAA,MACC,WAAW,GAAG,gEAAgE,SAAS;AAAA,MACvF,MAAK;AAAA,MACL,aAAU;AAAA,MAEV,UAAA;AAAA,QAAA,qBAAC,OAAA,EAAI,WAAU,2BACb,UAAA;AAAA,UAAA,oBAAC,QAAA,EAAK,WAAU,+DAA+D,UAAA,aAAY;AAAA,UAC3F,oBAAC,mBAAA,EAAkB,SAAkB,WAAU,6CAAA,CAA6C;AAAA,QAAA,GAC9F;AAAA,QACC,WACC,oBAAC,KAAA,EAAE,WAAU,oEAAoE,UAAA,QAAA,CAAQ;AAAA,MAAA;AAAA,IAAA;AAAA,EAAA;AAIjG;"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"dictation-engine.cjs","sources":["../../src/dictation/dictation-engine.ts"],"sourcesContent":["/**\n * Speech-to-text engines behind the composer's dictation mode.\n *\n * An engine turns microphone audio into a stream of transcript updates. The\n * SDK ships the browser's own `SpeechRecognition` engine (Chrome, Edge,\n * Safari), which needs no backend. A host can plug in another engine (for\n * example a relay to a server-side recognizer) through the `engine` option\n * of `useDictation` / `<AthenaChat dictation>`; the composer UI is engine\n * agnostic.\n */\n\nexport interface DictationTranscriptUpdate {\n /** Text the engine has committed. Appended to earlier committed text. */\n final: string;\n /** Text still being recognized; replaced on the next update. */\n interim: string;\n}\n\nexport interface DictationEngineCallbacks {\n onTranscript: (update: DictationTranscriptUpdate) => void;\n /** The engine stopped on its own (silence timeout, network, permission). */\n onEnd: () => void;\n onError: (error: DictationEngineError) => void;\n}\n\nexport interface DictationEngineSession {\n /** Ask the engine to finish. `onEnd` fires once it has flushed. */\n stop: () => void;\n /** Tear the engine down without waiting for a flush. */\n abort: () => void;\n}\n\nexport interface DictationEngine {\n /** Stable identifier surfaced in diagnostics (e.g. `web-speech`). */\n readonly id: string;\n /**\n * Whether this engine can run in the current browser. When false the\n * composer hides the mic instead of rendering a control that cannot work.\n */\n isSupported: () => boolean;\n /**\n * Start recognizing. `stream` is the microphone stream the composer already\n * opened for the waveform; engines that read audio themselves (Web Speech)\n * ignore it, relay engines encode it.\n */\n start: (options: {\n stream: MediaStream | null;\n lang: string;\n callbacks: DictationEngineCallbacks;\n }) => DictationEngineSession;\n}\n\nexport const DICTATION_ENGINE_ERROR_KINDS = {\n /** The user denied the microphone, or the browser blocked it. */\n permission: 'permission',\n /** No speech engine is available in this browser. */\n unsupported: 'unsupported',\n /** The engine could not reach its recognizer. */\n connection: 'connection',\n /** The engine heard nothing for its silence window. */\n noSpeech: 'no-speech',\n unknown: 'unknown',\n} as const;\n\nexport type DictationEngineErrorKind =\n (typeof DICTATION_ENGINE_ERROR_KINDS)[keyof typeof DICTATION_ENGINE_ERROR_KINDS];\n\nexport interface DictationEngineError {\n kind: DictationEngineErrorKind;\n /** One line the composer can show the user. */\n message: string;\n}\n\n// ─── Web Speech API engine ──────────────────────────────────────────\n\n// The Web Speech API has no lib.dom entry in every TypeScript version the SDK\n// supports, and browsers still vendor-prefix the constructor.\ninterface SpeechRecognitionAlternativeLike {\n transcript: string;\n}\ninterface SpeechRecognitionResultLike {\n isFinal: boolean;\n readonly length: number;\n [index: number]: SpeechRecognitionAlternativeLike;\n}\ninterface SpeechRecognitionEventLike extends Event {\n resultIndex: number;\n results: ArrayLike<SpeechRecognitionResultLike>;\n}\ninterface SpeechRecognitionErrorEventLike extends Event {\n error: string;\n message?: string;\n}\ninterface SpeechRecognitionLike extends EventTarget {\n continuous: boolean;\n interimResults: boolean;\n lang: string;\n start: () => void;\n stop: () => void;\n abort: () => void;\n onresult: ((event: SpeechRecognitionEventLike) => void) | null;\n onerror: ((event: SpeechRecognitionErrorEventLike) => void) | null;\n onend: (() => void) | null;\n}\ntype SpeechRecognitionCtor = new () => SpeechRecognitionLike;\n\nfunction resolveSpeechRecognitionCtor(): SpeechRecognitionCtor | null {\n if (typeof window === 'undefined') return null;\n const scope = window as unknown as {\n SpeechRecognition?: SpeechRecognitionCtor;\n webkitSpeechRecognition?: SpeechRecognitionCtor;\n };\n return scope.SpeechRecognition ?? scope.webkitSpeechRecognition ?? null;\n}\n\n/** Map a Web Speech `error` code to the engine's error vocabulary. */\nexport function classifyWebSpeechError(code: string): DictationEngineError {\n switch (code) {\n case 'not-allowed':\n case 'service-not-allowed':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'Microphone access was denied. Allow the microphone for this site and try again.',\n };\n case 'network':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.connection,\n message: 'Speech recognition could not connect. Check your connection and try again.',\n };\n case 'no-speech':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.noSpeech,\n message: 'No speech was heard.',\n };\n case 'audio-capture':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'No microphone was found.',\n };\n case 'aborted':\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation was interrupted.' };\n default:\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation stopped unexpectedly.' };\n }\n}\n\n/**\n * Fold one Web Speech result event into `{ final, interim }`. Finalized\n * alternatives from `resultIndex` onward are joined with a space; the rest is\n * the interim tail. Pure so it can be tested without a browser.\n */\nexport function foldWebSpeechResults(\n results: ArrayLike<SpeechRecognitionResultLike>,\n resultIndex: number,\n): DictationTranscriptUpdate {\n let final = '';\n let interim = '';\n for (let i = resultIndex; i < results.length; i += 1) {\n const result = results[i];\n const transcript = result?.[0]?.transcript ?? '';\n if (!transcript) continue;\n if (result.isFinal) final = final ? `${final} ${transcript.trim()}` : transcript.trim();\n else interim = interim ? `${interim} ${transcript.trim()}` : transcript.trim();\n }\n return { final, interim };\n}\n\n/**\n * The browser's own recognizer. Zero dependencies and zero backend; Chrome\n * and Edge stream partial results, Safari returns results in chunks, Firefox\n * has no implementation (`isSupported()` is false there).\n */\nexport function createWebSpeechDictationEngine(): DictationEngine {\n return {\n id: 'web-speech',\n isSupported: () => resolveSpeechRecognitionCtor() !== null,\n start: ({ lang, callbacks }) => {\n const Ctor = resolveSpeechRecognitionCtor();\n if (!Ctor) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unsupported,\n message: 'This browser has no speech recognition.',\n });\n callbacks.onEnd();\n return { stop: () => undefined, abort: () => undefined };\n }\n const recognition = new Ctor();\n recognition.continuous = true;\n recognition.interimResults = true;\n recognition.lang = lang;\n let ended = false;\n const end = () => {\n if (ended) return;\n ended = true;\n recognition.onresult = null;\n recognition.onerror = null;\n recognition.onend = null;\n callbacks.onEnd();\n };\n recognition.onresult = (event) => {\n callbacks.onTranscript(foldWebSpeechResults(event.results, event.resultIndex));\n };\n recognition.onerror = (event) => {\n // `aborted` follows our own abort(); the session already ended.\n if (event.error === 'aborted') return;\n callbacks.onError(classifyWebSpeechError(event.error));\n };\n recognition.onend = end;\n try {\n recognition.start();\n } catch (error) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: error instanceof Error ? error.message : 'Dictation could not start.',\n });\n end();\n }\n return {\n stop: () => {\n try {\n recognition.stop();\n } catch {\n end();\n }\n },\n abort: () => {\n try {\n recognition.abort();\n } finally {\n end();\n }\n },\n };\n },\n };\n}\n"],"names":[],"mappings":";;AAoDO,MAAM,+BAA+B;AAAA;AAAA,EAE1C,YAAY;AAAA;AAAA,EAEZ,aAAa;AAAA;AAAA,EAEb,YAAY;AAAA;AAAA,EAEZ,UAAU;AAAA,EACV,SAAS;AACX;AA4CA,SAAS,+BAA6D;AACpE,MAAI,OAAO,WAAW,YAAa,QAAO;AAC1C,QAAM,QAAQ;AAId,SAAO,MAAM,qBAAqB,MAAM,2BAA2B;AACrE;AAGO,SAAS,uBAAuB,MAAoC;AACzE,UAAQ,MAAA;AAAA,IACN,KAAK;AAAA,IACL,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,6BAAA;AAAA,IAChE;AACE,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,kCAAA;AAAA,EAAkC;AAEtG;AAOO,SAAS,qBACd,SACA,aAC2B;;AAC3B,MAAI,QAAQ;AACZ,MAAI,UAAU;AACd,WAAS,IAAI,aAAa,IAAI,QAAQ,QAAQ,KAAK,GAAG;AACpD,UAAM,SAAS,QAAQ,CAAC;AACxB,UAAM,eAAa,sCAAS,OAAT,mBAAa,eAAc;AAC9C,QAAI,CAAC,WAAY;AACjB,QAAI,OAAO,QAAS,SAAQ,QAAQ,GAAG,KAAK,IAAI,WAAW,KAAA,CAAM,KAAK,WAAW,KAAA;AAAA,QAC5E,WAAU,UAAU,GAAG,OAAO,IAAI,WAAW,MAAM,KAAK,WAAW,KAAA;AAAA,EAC1E;AACA,SAAO,EAAE,OAAO,QAAA;AAClB;AAOO,SAAS,iCAAkD;AAChE,SAAO;AAAA,IACL,IAAI;AAAA,IACJ,aAAa,MAAM,6BAAA,MAAmC;AAAA,IACtD,OAAO,CAAC,EAAE,MAAM,gBAAgB;AAC9B,YAAM,OAAO,6BAAA;AACb,UAAI,CAAC,MAAM;AACT,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS;AAAA,QAAA,CACV;AACD,kBAAU,MAAA;AACV,eAAO,EAAE,MAAM,MAAM,QAAW,OAAO,MAAM,OAAA;AAAA,MAC/C;AACA,YAAM,cAAc,IAAI,KAAA;AACxB,kBAAY,aAAa;AACzB,kBAAY,iBAAiB;AAC7B,kBAAY,OAAO;AACnB,UAAI,QAAQ;AACZ,YAAM,MAAM,MAAM;AAChB,YAAI,MAAO;AACX,gBAAQ;AACR,oBAAY,WAAW;AACvB,oBAAY,UAAU;AACtB,oBAAY,QAAQ;AACpB,kBAAU,MAAA;AAAA,MACZ;AACA,kBAAY,WAAW,CAAC,UAAU;AAChC,kBAAU,aAAa,qBAAqB,MAAM,SAAS,MAAM,WAAW,CAAC;AAAA,MAC/E;AACA,kBAAY,UAAU,CAAC,UAAU;AAE/B,YAAI,MAAM,UAAU,UAAW;AAC/B,kBAAU,QAAQ,uBAAuB,MAAM,KAAK,CAAC;AAAA,MACvD;AACA,kBAAY,QAAQ;AACpB,UAAI;AACF,oBAAY,MAAA;AAAA,MACd,SAAS,OAAO;AACd,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAAA,QAAA,CACnD;AACD,YAAA;AAAA,MACF;AACA,aAAO;AAAA,QACL,MAAM,MAAM;AACV,cAAI;AACF,wBAAY,KAAA;AAAA,UACd,QAAQ;AACN,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,QACA,OAAO,MAAM;AACX,cAAI;AACF,wBAAY,MAAA;AAAA,UACd,UAAA;AACE,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,MAAA;AAAA,IAEJ;AAAA,EAAA;AAEJ;;;;;"}
|
|
1
|
+
{"version":3,"file":"dictation-engine.cjs","sources":["../../src/dictation/dictation-engine.ts"],"sourcesContent":["/**\n * Speech-to-text engines behind the composer's dictation mode.\n *\n * An engine turns microphone audio into a stream of transcript updates. The\n * SDK ships the browser's own `SpeechRecognition` engine (Chrome, Edge,\n * Safari), which needs no backend. A host can plug in another engine (for\n * example a relay to a server-side recognizer) through the `engine` option\n * of `useDictation` / `<AthenaChat dictation>`; the composer UI is engine\n * agnostic.\n */\n\nexport interface DictationTranscriptUpdate {\n /** Text the engine has committed. Appended to earlier committed text. */\n final: string;\n /** Text still being recognized; replaced on the next update. */\n interim: string;\n}\n\nexport interface DictationEngineCallbacks {\n onTranscript: (update: DictationTranscriptUpdate) => void;\n /** The engine stopped on its own (silence timeout, network, permission). */\n onEnd: () => void;\n onError: (error: DictationEngineError) => void;\n}\n\nexport interface DictationEngineSession {\n /** Ask the engine to finish. `onEnd` fires once it has flushed. */\n stop: () => void;\n /** Tear the engine down without waiting for a flush. */\n abort: () => void;\n}\n\nexport interface DictationEngine {\n /** Stable identifier surfaced in diagnostics (e.g. `web-speech`). */\n readonly id: string;\n /**\n * Whether this engine can run in the current browser. When false the\n * composer hides the mic instead of rendering a control that cannot work.\n */\n isSupported: () => boolean;\n /**\n * How long `stop()` may wait for the final transcript before the session\n * is forced closed. Engines that transcribe after recording need more than\n * the live default (1.5s).\n */\n readonly flushTimeoutMs?: number;\n /** Status shown while flushing (e.g. `Transcribing…`); defaults to `Finishing…`. */\n readonly stoppingLabel?: string;\n /**\n * Start recognizing. `stream` is the microphone stream the composer already\n * opened for the waveform; engines that read audio themselves (Web Speech)\n * ignore it, relay engines encode it.\n */\n start: (options: {\n stream: MediaStream | null;\n lang: string;\n callbacks: DictationEngineCallbacks;\n }) => DictationEngineSession;\n}\n\nexport const DICTATION_ENGINE_ERROR_KINDS = {\n /** The user denied the microphone, or the browser blocked it. */\n permission: 'permission',\n /** No speech engine is available in this browser. */\n unsupported: 'unsupported',\n /** The engine could not reach its recognizer. */\n connection: 'connection',\n /** The engine heard nothing for its silence window. */\n noSpeech: 'no-speech',\n unknown: 'unknown',\n} as const;\n\nexport type DictationEngineErrorKind =\n (typeof DICTATION_ENGINE_ERROR_KINDS)[keyof typeof DICTATION_ENGINE_ERROR_KINDS];\n\nexport interface DictationEngineError {\n kind: DictationEngineErrorKind;\n /** One line the composer can show the user. */\n message: string;\n}\n\n// ─── Web Speech API engine ──────────────────────────────────────────\n\n// The Web Speech API has no lib.dom entry in every TypeScript version the SDK\n// supports, and browsers still vendor-prefix the constructor.\ninterface SpeechRecognitionAlternativeLike {\n transcript: string;\n}\ninterface SpeechRecognitionResultLike {\n isFinal: boolean;\n readonly length: number;\n [index: number]: SpeechRecognitionAlternativeLike;\n}\ninterface SpeechRecognitionEventLike extends Event {\n resultIndex: number;\n results: ArrayLike<SpeechRecognitionResultLike>;\n}\ninterface SpeechRecognitionErrorEventLike extends Event {\n error: string;\n message?: string;\n}\ninterface SpeechRecognitionLike extends EventTarget {\n continuous: boolean;\n interimResults: boolean;\n lang: string;\n start: () => void;\n stop: () => void;\n abort: () => void;\n onresult: ((event: SpeechRecognitionEventLike) => void) | null;\n onerror: ((event: SpeechRecognitionErrorEventLike) => void) | null;\n onend: (() => void) | null;\n}\ntype SpeechRecognitionCtor = new () => SpeechRecognitionLike;\n\nfunction resolveSpeechRecognitionCtor(): SpeechRecognitionCtor | null {\n if (typeof window === 'undefined') return null;\n const scope = window as unknown as {\n SpeechRecognition?: SpeechRecognitionCtor;\n webkitSpeechRecognition?: SpeechRecognitionCtor;\n };\n return scope.SpeechRecognition ?? scope.webkitSpeechRecognition ?? null;\n}\n\n/** Map a Web Speech `error` code to the engine's error vocabulary. */\nexport function classifyWebSpeechError(code: string): DictationEngineError {\n switch (code) {\n case 'not-allowed':\n case 'service-not-allowed':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'Microphone access was denied. Allow the microphone for this site and try again.',\n };\n case 'network':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.connection,\n message: 'Speech recognition could not connect. Check your connection and try again.',\n };\n case 'no-speech':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.noSpeech,\n message: 'No speech was heard.',\n };\n case 'audio-capture':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'No microphone was found.',\n };\n case 'aborted':\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation was interrupted.' };\n default:\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation stopped unexpectedly.' };\n }\n}\n\n/**\n * Fold one Web Speech result event into `{ final, interim }`. Finalized\n * alternatives from `resultIndex` onward are joined with a space; the rest is\n * the interim tail. Pure so it can be tested without a browser.\n */\nexport function foldWebSpeechResults(\n results: ArrayLike<SpeechRecognitionResultLike>,\n resultIndex: number,\n): DictationTranscriptUpdate {\n let final = '';\n let interim = '';\n for (let i = resultIndex; i < results.length; i += 1) {\n const result = results[i];\n const transcript = result?.[0]?.transcript ?? '';\n if (!transcript) continue;\n if (result.isFinal) final = final ? `${final} ${transcript.trim()}` : transcript.trim();\n else interim = interim ? `${interim} ${transcript.trim()}` : transcript.trim();\n }\n return { final, interim };\n}\n\n/**\n * The browser's own recognizer. Zero dependencies and zero backend; Chrome\n * and Edge stream partial results, Safari returns results in chunks, Firefox\n * has no implementation (`isSupported()` is false there).\n */\nexport function createWebSpeechDictationEngine(): DictationEngine {\n return {\n id: 'web-speech',\n isSupported: () => resolveSpeechRecognitionCtor() !== null,\n start: ({ lang, callbacks }) => {\n const Ctor = resolveSpeechRecognitionCtor();\n if (!Ctor) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unsupported,\n message: 'This browser has no speech recognition.',\n });\n callbacks.onEnd();\n return { stop: () => undefined, abort: () => undefined };\n }\n const recognition = new Ctor();\n recognition.continuous = true;\n recognition.interimResults = true;\n recognition.lang = lang;\n let ended = false;\n const end = () => {\n if (ended) return;\n ended = true;\n recognition.onresult = null;\n recognition.onerror = null;\n recognition.onend = null;\n callbacks.onEnd();\n };\n recognition.onresult = (event) => {\n callbacks.onTranscript(foldWebSpeechResults(event.results, event.resultIndex));\n };\n recognition.onerror = (event) => {\n // `aborted` follows our own abort(); the session already ended.\n if (event.error === 'aborted') return;\n callbacks.onError(classifyWebSpeechError(event.error));\n };\n recognition.onend = end;\n try {\n recognition.start();\n } catch (error) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: error instanceof Error ? error.message : 'Dictation could not start.',\n });\n end();\n }\n return {\n stop: () => {\n try {\n recognition.stop();\n } catch {\n end();\n }\n },\n abort: () => {\n try {\n recognition.abort();\n } finally {\n end();\n }\n },\n };\n },\n };\n}\n"],"names":[],"mappings":";;AA4DO,MAAM,+BAA+B;AAAA;AAAA,EAE1C,YAAY;AAAA;AAAA,EAEZ,aAAa;AAAA;AAAA,EAEb,YAAY;AAAA;AAAA,EAEZ,UAAU;AAAA,EACV,SAAS;AACX;AA4CA,SAAS,+BAA6D;AACpE,MAAI,OAAO,WAAW,YAAa,QAAO;AAC1C,QAAM,QAAQ;AAId,SAAO,MAAM,qBAAqB,MAAM,2BAA2B;AACrE;AAGO,SAAS,uBAAuB,MAAoC;AACzE,UAAQ,MAAA;AAAA,IACN,KAAK;AAAA,IACL,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,6BAAA;AAAA,IAChE;AACE,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,kCAAA;AAAA,EAAkC;AAEtG;AAOO,SAAS,qBACd,SACA,aAC2B;;AAC3B,MAAI,QAAQ;AACZ,MAAI,UAAU;AACd,WAAS,IAAI,aAAa,IAAI,QAAQ,QAAQ,KAAK,GAAG;AACpD,UAAM,SAAS,QAAQ,CAAC;AACxB,UAAM,eAAa,sCAAS,OAAT,mBAAa,eAAc;AAC9C,QAAI,CAAC,WAAY;AACjB,QAAI,OAAO,QAAS,SAAQ,QAAQ,GAAG,KAAK,IAAI,WAAW,KAAA,CAAM,KAAK,WAAW,KAAA;AAAA,QAC5E,WAAU,UAAU,GAAG,OAAO,IAAI,WAAW,MAAM,KAAK,WAAW,KAAA;AAAA,EAC1E;AACA,SAAO,EAAE,OAAO,QAAA;AAClB;AAOO,SAAS,iCAAkD;AAChE,SAAO;AAAA,IACL,IAAI;AAAA,IACJ,aAAa,MAAM,6BAAA,MAAmC;AAAA,IACtD,OAAO,CAAC,EAAE,MAAM,gBAAgB;AAC9B,YAAM,OAAO,6BAAA;AACb,UAAI,CAAC,MAAM;AACT,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS;AAAA,QAAA,CACV;AACD,kBAAU,MAAA;AACV,eAAO,EAAE,MAAM,MAAM,QAAW,OAAO,MAAM,OAAA;AAAA,MAC/C;AACA,YAAM,cAAc,IAAI,KAAA;AACxB,kBAAY,aAAa;AACzB,kBAAY,iBAAiB;AAC7B,kBAAY,OAAO;AACnB,UAAI,QAAQ;AACZ,YAAM,MAAM,MAAM;AAChB,YAAI,MAAO;AACX,gBAAQ;AACR,oBAAY,WAAW;AACvB,oBAAY,UAAU;AACtB,oBAAY,QAAQ;AACpB,kBAAU,MAAA;AAAA,MACZ;AACA,kBAAY,WAAW,CAAC,UAAU;AAChC,kBAAU,aAAa,qBAAqB,MAAM,SAAS,MAAM,WAAW,CAAC;AAAA,MAC/E;AACA,kBAAY,UAAU,CAAC,UAAU;AAE/B,YAAI,MAAM,UAAU,UAAW;AAC/B,kBAAU,QAAQ,uBAAuB,MAAM,KAAK,CAAC;AAAA,MACvD;AACA,kBAAY,QAAQ;AACpB,UAAI;AACF,oBAAY,MAAA;AAAA,MACd,SAAS,OAAO;AACd,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAAA,QAAA,CACnD;AACD,YAAA;AAAA,MACF;AACA,aAAO;AAAA,QACL,MAAM,MAAM;AACV,cAAI;AACF,wBAAY,KAAA;AAAA,UACd,QAAQ;AACN,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,QACA,OAAO,MAAM;AACX,cAAI;AACF,wBAAY,MAAA;AAAA,UACd,UAAA;AACE,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,MAAA;AAAA,IAEJ;AAAA,EAAA;AAEJ;;;;;"}
|
|
@@ -34,6 +34,14 @@ export interface DictationEngine {
|
|
|
34
34
|
* composer hides the mic instead of rendering a control that cannot work.
|
|
35
35
|
*/
|
|
36
36
|
isSupported: () => boolean;
|
|
37
|
+
/**
|
|
38
|
+
* How long `stop()` may wait for the final transcript before the session
|
|
39
|
+
* is forced closed. Engines that transcribe after recording need more than
|
|
40
|
+
* the live default (1.5s).
|
|
41
|
+
*/
|
|
42
|
+
readonly flushTimeoutMs?: number;
|
|
43
|
+
/** Status shown while flushing (e.g. `Transcribing…`); defaults to `Finishing…`. */
|
|
44
|
+
readonly stoppingLabel?: string;
|
|
37
45
|
/**
|
|
38
46
|
* Start recognizing. `stream` is the microphone stream the composer already
|
|
39
47
|
* opened for the waveform; engines that read audio themselves (Web Speech)
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"dictation-engine.js","sources":["../../src/dictation/dictation-engine.ts"],"sourcesContent":["/**\n * Speech-to-text engines behind the composer's dictation mode.\n *\n * An engine turns microphone audio into a stream of transcript updates. The\n * SDK ships the browser's own `SpeechRecognition` engine (Chrome, Edge,\n * Safari), which needs no backend. A host can plug in another engine (for\n * example a relay to a server-side recognizer) through the `engine` option\n * of `useDictation` / `<AthenaChat dictation>`; the composer UI is engine\n * agnostic.\n */\n\nexport interface DictationTranscriptUpdate {\n /** Text the engine has committed. Appended to earlier committed text. */\n final: string;\n /** Text still being recognized; replaced on the next update. */\n interim: string;\n}\n\nexport interface DictationEngineCallbacks {\n onTranscript: (update: DictationTranscriptUpdate) => void;\n /** The engine stopped on its own (silence timeout, network, permission). */\n onEnd: () => void;\n onError: (error: DictationEngineError) => void;\n}\n\nexport interface DictationEngineSession {\n /** Ask the engine to finish. `onEnd` fires once it has flushed. */\n stop: () => void;\n /** Tear the engine down without waiting for a flush. */\n abort: () => void;\n}\n\nexport interface DictationEngine {\n /** Stable identifier surfaced in diagnostics (e.g. `web-speech`). */\n readonly id: string;\n /**\n * Whether this engine can run in the current browser. When false the\n * composer hides the mic instead of rendering a control that cannot work.\n */\n isSupported: () => boolean;\n /**\n * Start recognizing. `stream` is the microphone stream the composer already\n * opened for the waveform; engines that read audio themselves (Web Speech)\n * ignore it, relay engines encode it.\n */\n start: (options: {\n stream: MediaStream | null;\n lang: string;\n callbacks: DictationEngineCallbacks;\n }) => DictationEngineSession;\n}\n\nexport const DICTATION_ENGINE_ERROR_KINDS = {\n /** The user denied the microphone, or the browser blocked it. */\n permission: 'permission',\n /** No speech engine is available in this browser. */\n unsupported: 'unsupported',\n /** The engine could not reach its recognizer. */\n connection: 'connection',\n /** The engine heard nothing for its silence window. */\n noSpeech: 'no-speech',\n unknown: 'unknown',\n} as const;\n\nexport type DictationEngineErrorKind =\n (typeof DICTATION_ENGINE_ERROR_KINDS)[keyof typeof DICTATION_ENGINE_ERROR_KINDS];\n\nexport interface DictationEngineError {\n kind: DictationEngineErrorKind;\n /** One line the composer can show the user. */\n message: string;\n}\n\n// ─── Web Speech API engine ──────────────────────────────────────────\n\n// The Web Speech API has no lib.dom entry in every TypeScript version the SDK\n// supports, and browsers still vendor-prefix the constructor.\ninterface SpeechRecognitionAlternativeLike {\n transcript: string;\n}\ninterface SpeechRecognitionResultLike {\n isFinal: boolean;\n readonly length: number;\n [index: number]: SpeechRecognitionAlternativeLike;\n}\ninterface SpeechRecognitionEventLike extends Event {\n resultIndex: number;\n results: ArrayLike<SpeechRecognitionResultLike>;\n}\ninterface SpeechRecognitionErrorEventLike extends Event {\n error: string;\n message?: string;\n}\ninterface SpeechRecognitionLike extends EventTarget {\n continuous: boolean;\n interimResults: boolean;\n lang: string;\n start: () => void;\n stop: () => void;\n abort: () => void;\n onresult: ((event: SpeechRecognitionEventLike) => void) | null;\n onerror: ((event: SpeechRecognitionErrorEventLike) => void) | null;\n onend: (() => void) | null;\n}\ntype SpeechRecognitionCtor = new () => SpeechRecognitionLike;\n\nfunction resolveSpeechRecognitionCtor(): SpeechRecognitionCtor | null {\n if (typeof window === 'undefined') return null;\n const scope = window as unknown as {\n SpeechRecognition?: SpeechRecognitionCtor;\n webkitSpeechRecognition?: SpeechRecognitionCtor;\n };\n return scope.SpeechRecognition ?? scope.webkitSpeechRecognition ?? null;\n}\n\n/** Map a Web Speech `error` code to the engine's error vocabulary. */\nexport function classifyWebSpeechError(code: string): DictationEngineError {\n switch (code) {\n case 'not-allowed':\n case 'service-not-allowed':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'Microphone access was denied. Allow the microphone for this site and try again.',\n };\n case 'network':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.connection,\n message: 'Speech recognition could not connect. Check your connection and try again.',\n };\n case 'no-speech':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.noSpeech,\n message: 'No speech was heard.',\n };\n case 'audio-capture':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'No microphone was found.',\n };\n case 'aborted':\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation was interrupted.' };\n default:\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation stopped unexpectedly.' };\n }\n}\n\n/**\n * Fold one Web Speech result event into `{ final, interim }`. Finalized\n * alternatives from `resultIndex` onward are joined with a space; the rest is\n * the interim tail. Pure so it can be tested without a browser.\n */\nexport function foldWebSpeechResults(\n results: ArrayLike<SpeechRecognitionResultLike>,\n resultIndex: number,\n): DictationTranscriptUpdate {\n let final = '';\n let interim = '';\n for (let i = resultIndex; i < results.length; i += 1) {\n const result = results[i];\n const transcript = result?.[0]?.transcript ?? '';\n if (!transcript) continue;\n if (result.isFinal) final = final ? `${final} ${transcript.trim()}` : transcript.trim();\n else interim = interim ? `${interim} ${transcript.trim()}` : transcript.trim();\n }\n return { final, interim };\n}\n\n/**\n * The browser's own recognizer. Zero dependencies and zero backend; Chrome\n * and Edge stream partial results, Safari returns results in chunks, Firefox\n * has no implementation (`isSupported()` is false there).\n */\nexport function createWebSpeechDictationEngine(): DictationEngine {\n return {\n id: 'web-speech',\n isSupported: () => resolveSpeechRecognitionCtor() !== null,\n start: ({ lang, callbacks }) => {\n const Ctor = resolveSpeechRecognitionCtor();\n if (!Ctor) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unsupported,\n message: 'This browser has no speech recognition.',\n });\n callbacks.onEnd();\n return { stop: () => undefined, abort: () => undefined };\n }\n const recognition = new Ctor();\n recognition.continuous = true;\n recognition.interimResults = true;\n recognition.lang = lang;\n let ended = false;\n const end = () => {\n if (ended) return;\n ended = true;\n recognition.onresult = null;\n recognition.onerror = null;\n recognition.onend = null;\n callbacks.onEnd();\n };\n recognition.onresult = (event) => {\n callbacks.onTranscript(foldWebSpeechResults(event.results, event.resultIndex));\n };\n recognition.onerror = (event) => {\n // `aborted` follows our own abort(); the session already ended.\n if (event.error === 'aborted') return;\n callbacks.onError(classifyWebSpeechError(event.error));\n };\n recognition.onend = end;\n try {\n recognition.start();\n } catch (error) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: error instanceof Error ? error.message : 'Dictation could not start.',\n });\n end();\n }\n return {\n stop: () => {\n try {\n recognition.stop();\n } catch {\n end();\n }\n },\n abort: () => {\n try {\n recognition.abort();\n } finally {\n end();\n }\n },\n };\n },\n };\n}\n"],"names":[],"mappings":"AAoDO,MAAM,+BAA+B;AAAA;AAAA,EAE1C,YAAY;AAAA;AAAA,EAEZ,aAAa;AAAA;AAAA,EAEb,YAAY;AAAA;AAAA,EAEZ,UAAU;AAAA,EACV,SAAS;AACX;AA4CA,SAAS,+BAA6D;AACpE,MAAI,OAAO,WAAW,YAAa,QAAO;AAC1C,QAAM,QAAQ;AAId,SAAO,MAAM,qBAAqB,MAAM,2BAA2B;AACrE;AAGO,SAAS,uBAAuB,MAAoC;AACzE,UAAQ,MAAA;AAAA,IACN,KAAK;AAAA,IACL,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,6BAAA;AAAA,IAChE;AACE,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,kCAAA;AAAA,EAAkC;AAEtG;AAOO,SAAS,qBACd,SACA,aAC2B;AAtGtB;AAuGL,MAAI,QAAQ;AACZ,MAAI,UAAU;AACd,WAAS,IAAI,aAAa,IAAI,QAAQ,QAAQ,KAAK,GAAG;AACpD,UAAM,SAAS,QAAQ,CAAC;AACxB,UAAM,eAAa,sCAAS,OAAT,mBAAa,eAAc;AAC9C,QAAI,CAAC,WAAY;AACjB,QAAI,OAAO,QAAS,SAAQ,QAAQ,GAAG,KAAK,IAAI,WAAW,KAAA,CAAM,KAAK,WAAW,KAAA;AAAA,QAC5E,WAAU,UAAU,GAAG,OAAO,IAAI,WAAW,MAAM,KAAK,WAAW,KAAA;AAAA,EAC1E;AACA,SAAO,EAAE,OAAO,QAAA;AAClB;AAOO,SAAS,iCAAkD;AAChE,SAAO;AAAA,IACL,IAAI;AAAA,IACJ,aAAa,MAAM,6BAAA,MAAmC;AAAA,IACtD,OAAO,CAAC,EAAE,MAAM,gBAAgB;AAC9B,YAAM,OAAO,6BAAA;AACb,UAAI,CAAC,MAAM;AACT,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS;AAAA,QAAA,CACV;AACD,kBAAU,MAAA;AACV,eAAO,EAAE,MAAM,MAAM,QAAW,OAAO,MAAM,OAAA;AAAA,MAC/C;AACA,YAAM,cAAc,IAAI,KAAA;AACxB,kBAAY,aAAa;AACzB,kBAAY,iBAAiB;AAC7B,kBAAY,OAAO;AACnB,UAAI,QAAQ;AACZ,YAAM,MAAM,MAAM;AAChB,YAAI,MAAO;AACX,gBAAQ;AACR,oBAAY,WAAW;AACvB,oBAAY,UAAU;AACtB,oBAAY,QAAQ;AACpB,kBAAU,MAAA;AAAA,MACZ;AACA,kBAAY,WAAW,CAAC,UAAU;AAChC,kBAAU,aAAa,qBAAqB,MAAM,SAAS,MAAM,WAAW,CAAC;AAAA,MAC/E;AACA,kBAAY,UAAU,CAAC,UAAU;AAE/B,YAAI,MAAM,UAAU,UAAW;AAC/B,kBAAU,QAAQ,uBAAuB,MAAM,KAAK,CAAC;AAAA,MACvD;AACA,kBAAY,QAAQ;AACpB,UAAI;AACF,oBAAY,MAAA;AAAA,MACd,SAAS,OAAO;AACd,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAAA,QAAA,CACnD;AACD,YAAA;AAAA,MACF;AACA,aAAO;AAAA,QACL,MAAM,MAAM;AACV,cAAI;AACF,wBAAY,KAAA;AAAA,UACd,QAAQ;AACN,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,QACA,OAAO,MAAM;AACX,cAAI;AACF,wBAAY,MAAA;AAAA,UACd,UAAA;AACE,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,MAAA;AAAA,IAEJ;AAAA,EAAA;AAEJ;"}
|
|
1
|
+
{"version":3,"file":"dictation-engine.js","sources":["../../src/dictation/dictation-engine.ts"],"sourcesContent":["/**\n * Speech-to-text engines behind the composer's dictation mode.\n *\n * An engine turns microphone audio into a stream of transcript updates. The\n * SDK ships the browser's own `SpeechRecognition` engine (Chrome, Edge,\n * Safari), which needs no backend. A host can plug in another engine (for\n * example a relay to a server-side recognizer) through the `engine` option\n * of `useDictation` / `<AthenaChat dictation>`; the composer UI is engine\n * agnostic.\n */\n\nexport interface DictationTranscriptUpdate {\n /** Text the engine has committed. Appended to earlier committed text. */\n final: string;\n /** Text still being recognized; replaced on the next update. */\n interim: string;\n}\n\nexport interface DictationEngineCallbacks {\n onTranscript: (update: DictationTranscriptUpdate) => void;\n /** The engine stopped on its own (silence timeout, network, permission). */\n onEnd: () => void;\n onError: (error: DictationEngineError) => void;\n}\n\nexport interface DictationEngineSession {\n /** Ask the engine to finish. `onEnd` fires once it has flushed. */\n stop: () => void;\n /** Tear the engine down without waiting for a flush. */\n abort: () => void;\n}\n\nexport interface DictationEngine {\n /** Stable identifier surfaced in diagnostics (e.g. `web-speech`). */\n readonly id: string;\n /**\n * Whether this engine can run in the current browser. When false the\n * composer hides the mic instead of rendering a control that cannot work.\n */\n isSupported: () => boolean;\n /**\n * How long `stop()` may wait for the final transcript before the session\n * is forced closed. Engines that transcribe after recording need more than\n * the live default (1.5s).\n */\n readonly flushTimeoutMs?: number;\n /** Status shown while flushing (e.g. `Transcribing…`); defaults to `Finishing…`. */\n readonly stoppingLabel?: string;\n /**\n * Start recognizing. `stream` is the microphone stream the composer already\n * opened for the waveform; engines that read audio themselves (Web Speech)\n * ignore it, relay engines encode it.\n */\n start: (options: {\n stream: MediaStream | null;\n lang: string;\n callbacks: DictationEngineCallbacks;\n }) => DictationEngineSession;\n}\n\nexport const DICTATION_ENGINE_ERROR_KINDS = {\n /** The user denied the microphone, or the browser blocked it. */\n permission: 'permission',\n /** No speech engine is available in this browser. */\n unsupported: 'unsupported',\n /** The engine could not reach its recognizer. */\n connection: 'connection',\n /** The engine heard nothing for its silence window. */\n noSpeech: 'no-speech',\n unknown: 'unknown',\n} as const;\n\nexport type DictationEngineErrorKind =\n (typeof DICTATION_ENGINE_ERROR_KINDS)[keyof typeof DICTATION_ENGINE_ERROR_KINDS];\n\nexport interface DictationEngineError {\n kind: DictationEngineErrorKind;\n /** One line the composer can show the user. */\n message: string;\n}\n\n// ─── Web Speech API engine ──────────────────────────────────────────\n\n// The Web Speech API has no lib.dom entry in every TypeScript version the SDK\n// supports, and browsers still vendor-prefix the constructor.\ninterface SpeechRecognitionAlternativeLike {\n transcript: string;\n}\ninterface SpeechRecognitionResultLike {\n isFinal: boolean;\n readonly length: number;\n [index: number]: SpeechRecognitionAlternativeLike;\n}\ninterface SpeechRecognitionEventLike extends Event {\n resultIndex: number;\n results: ArrayLike<SpeechRecognitionResultLike>;\n}\ninterface SpeechRecognitionErrorEventLike extends Event {\n error: string;\n message?: string;\n}\ninterface SpeechRecognitionLike extends EventTarget {\n continuous: boolean;\n interimResults: boolean;\n lang: string;\n start: () => void;\n stop: () => void;\n abort: () => void;\n onresult: ((event: SpeechRecognitionEventLike) => void) | null;\n onerror: ((event: SpeechRecognitionErrorEventLike) => void) | null;\n onend: (() => void) | null;\n}\ntype SpeechRecognitionCtor = new () => SpeechRecognitionLike;\n\nfunction resolveSpeechRecognitionCtor(): SpeechRecognitionCtor | null {\n if (typeof window === 'undefined') return null;\n const scope = window as unknown as {\n SpeechRecognition?: SpeechRecognitionCtor;\n webkitSpeechRecognition?: SpeechRecognitionCtor;\n };\n return scope.SpeechRecognition ?? scope.webkitSpeechRecognition ?? null;\n}\n\n/** Map a Web Speech `error` code to the engine's error vocabulary. */\nexport function classifyWebSpeechError(code: string): DictationEngineError {\n switch (code) {\n case 'not-allowed':\n case 'service-not-allowed':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'Microphone access was denied. Allow the microphone for this site and try again.',\n };\n case 'network':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.connection,\n message: 'Speech recognition could not connect. Check your connection and try again.',\n };\n case 'no-speech':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.noSpeech,\n message: 'No speech was heard.',\n };\n case 'audio-capture':\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.permission,\n message: 'No microphone was found.',\n };\n case 'aborted':\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation was interrupted.' };\n default:\n return { kind: DICTATION_ENGINE_ERROR_KINDS.unknown, message: 'Dictation stopped unexpectedly.' };\n }\n}\n\n/**\n * Fold one Web Speech result event into `{ final, interim }`. Finalized\n * alternatives from `resultIndex` onward are joined with a space; the rest is\n * the interim tail. Pure so it can be tested without a browser.\n */\nexport function foldWebSpeechResults(\n results: ArrayLike<SpeechRecognitionResultLike>,\n resultIndex: number,\n): DictationTranscriptUpdate {\n let final = '';\n let interim = '';\n for (let i = resultIndex; i < results.length; i += 1) {\n const result = results[i];\n const transcript = result?.[0]?.transcript ?? '';\n if (!transcript) continue;\n if (result.isFinal) final = final ? `${final} ${transcript.trim()}` : transcript.trim();\n else interim = interim ? `${interim} ${transcript.trim()}` : transcript.trim();\n }\n return { final, interim };\n}\n\n/**\n * The browser's own recognizer. Zero dependencies and zero backend; Chrome\n * and Edge stream partial results, Safari returns results in chunks, Firefox\n * has no implementation (`isSupported()` is false there).\n */\nexport function createWebSpeechDictationEngine(): DictationEngine {\n return {\n id: 'web-speech',\n isSupported: () => resolveSpeechRecognitionCtor() !== null,\n start: ({ lang, callbacks }) => {\n const Ctor = resolveSpeechRecognitionCtor();\n if (!Ctor) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unsupported,\n message: 'This browser has no speech recognition.',\n });\n callbacks.onEnd();\n return { stop: () => undefined, abort: () => undefined };\n }\n const recognition = new Ctor();\n recognition.continuous = true;\n recognition.interimResults = true;\n recognition.lang = lang;\n let ended = false;\n const end = () => {\n if (ended) return;\n ended = true;\n recognition.onresult = null;\n recognition.onerror = null;\n recognition.onend = null;\n callbacks.onEnd();\n };\n recognition.onresult = (event) => {\n callbacks.onTranscript(foldWebSpeechResults(event.results, event.resultIndex));\n };\n recognition.onerror = (event) => {\n // `aborted` follows our own abort(); the session already ended.\n if (event.error === 'aborted') return;\n callbacks.onError(classifyWebSpeechError(event.error));\n };\n recognition.onend = end;\n try {\n recognition.start();\n } catch (error) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: error instanceof Error ? error.message : 'Dictation could not start.',\n });\n end();\n }\n return {\n stop: () => {\n try {\n recognition.stop();\n } catch {\n end();\n }\n },\n abort: () => {\n try {\n recognition.abort();\n } finally {\n end();\n }\n },\n };\n },\n };\n}\n"],"names":[],"mappings":"AA4DO,MAAM,+BAA+B;AAAA;AAAA,EAE1C,YAAY;AAAA;AAAA,EAEZ,aAAa;AAAA;AAAA,EAEb,YAAY;AAAA;AAAA,EAEZ,UAAU;AAAA,EACV,SAAS;AACX;AA4CA,SAAS,+BAA6D;AACpE,MAAI,OAAO,WAAW,YAAa,QAAO;AAC1C,QAAM,QAAQ;AAId,SAAO,MAAM,qBAAqB,MAAM,2BAA2B;AACrE;AAGO,SAAS,uBAAuB,MAAoC;AACzE,UAAQ,MAAA;AAAA,IACN,KAAK;AAAA,IACL,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO;AAAA,QACL,MAAM,6BAA6B;AAAA,QACnC,SAAS;AAAA,MAAA;AAAA,IAEb,KAAK;AACH,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,6BAAA;AAAA,IAChE;AACE,aAAO,EAAE,MAAM,6BAA6B,SAAS,SAAS,kCAAA;AAAA,EAAkC;AAEtG;AAOO,SAAS,qBACd,SACA,aAC2B;AAtGtB;AAuGL,MAAI,QAAQ;AACZ,MAAI,UAAU;AACd,WAAS,IAAI,aAAa,IAAI,QAAQ,QAAQ,KAAK,GAAG;AACpD,UAAM,SAAS,QAAQ,CAAC;AACxB,UAAM,eAAa,sCAAS,OAAT,mBAAa,eAAc;AAC9C,QAAI,CAAC,WAAY;AACjB,QAAI,OAAO,QAAS,SAAQ,QAAQ,GAAG,KAAK,IAAI,WAAW,KAAA,CAAM,KAAK,WAAW,KAAA;AAAA,QAC5E,WAAU,UAAU,GAAG,OAAO,IAAI,WAAW,MAAM,KAAK,WAAW,KAAA;AAAA,EAC1E;AACA,SAAO,EAAE,OAAO,QAAA;AAClB;AAOO,SAAS,iCAAkD;AAChE,SAAO;AAAA,IACL,IAAI;AAAA,IACJ,aAAa,MAAM,6BAAA,MAAmC;AAAA,IACtD,OAAO,CAAC,EAAE,MAAM,gBAAgB;AAC9B,YAAM,OAAO,6BAAA;AACb,UAAI,CAAC,MAAM;AACT,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS;AAAA,QAAA,CACV;AACD,kBAAU,MAAA;AACV,eAAO,EAAE,MAAM,MAAM,QAAW,OAAO,MAAM,OAAA;AAAA,MAC/C;AACA,YAAM,cAAc,IAAI,KAAA;AACxB,kBAAY,aAAa;AACzB,kBAAY,iBAAiB;AAC7B,kBAAY,OAAO;AACnB,UAAI,QAAQ;AACZ,YAAM,MAAM,MAAM;AAChB,YAAI,MAAO;AACX,gBAAQ;AACR,oBAAY,WAAW;AACvB,oBAAY,UAAU;AACtB,oBAAY,QAAQ;AACpB,kBAAU,MAAA;AAAA,MACZ;AACA,kBAAY,WAAW,CAAC,UAAU;AAChC,kBAAU,aAAa,qBAAqB,MAAM,SAAS,MAAM,WAAW,CAAC;AAAA,MAC/E;AACA,kBAAY,UAAU,CAAC,UAAU;AAE/B,YAAI,MAAM,UAAU,UAAW;AAC/B,kBAAU,QAAQ,uBAAuB,MAAM,KAAK,CAAC;AAAA,MACvD;AACA,kBAAY,QAAQ;AACpB,UAAI;AACF,oBAAY,MAAA;AAAA,MACd,SAAS,OAAO;AACd,kBAAU,QAAQ;AAAA,UAChB,MAAM,6BAA6B;AAAA,UACnC,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAAA,QAAA,CACnD;AACD,YAAA;AAAA,MACF;AACA,aAAO;AAAA,QACL,MAAM,MAAM;AACV,cAAI;AACF,wBAAY,KAAA;AAAA,UACd,QAAQ;AACN,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,QACA,OAAO,MAAM;AACX,cAAI;AACF,wBAAY,MAAA;AAAA,UACd,UAAA;AACE,gBAAA;AAAA,UACF;AAAA,QACF;AAAA,MAAA;AAAA,IAEJ;AAAA,EAAA;AAEJ;"}
|
|
@@ -0,0 +1,153 @@
|
|
|
1
|
+
"use strict";
|
|
2
|
+
Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
|
|
3
|
+
const dictationEngine = require("./dictation-engine.cjs");
|
|
4
|
+
const DEFAULT_MIME_TYPES = ["audio/webm;codecs=opus", "audio/webm", "audio/mp4", "audio/ogg;codecs=opus"];
|
|
5
|
+
const DEFAULT_FLUSH_TIMEOUT_MS = 9e4;
|
|
6
|
+
const DEFAULT_MAX_RECORDING_MS = 5 * 6e4;
|
|
7
|
+
function resolveMediaRecorderCtor() {
|
|
8
|
+
if (typeof window === "undefined") return null;
|
|
9
|
+
return window.MediaRecorder ?? null;
|
|
10
|
+
}
|
|
11
|
+
function pickRecorderMimeType(Ctor, candidates) {
|
|
12
|
+
if (!Ctor.isTypeSupported) return void 0;
|
|
13
|
+
return candidates.find((type) => {
|
|
14
|
+
var _a;
|
|
15
|
+
return (_a = Ctor.isTypeSupported) == null ? void 0 : _a.call(Ctor, type);
|
|
16
|
+
});
|
|
17
|
+
}
|
|
18
|
+
function transcriptionError(error) {
|
|
19
|
+
return {
|
|
20
|
+
kind: dictationEngine.DICTATION_ENGINE_ERROR_KINDS.connection,
|
|
21
|
+
message: error instanceof Error && error.message ? error.message : "Transcription failed. Try again."
|
|
22
|
+
};
|
|
23
|
+
}
|
|
24
|
+
function createRecordedDictationEngine(options) {
|
|
25
|
+
const {
|
|
26
|
+
transcribe,
|
|
27
|
+
mimeTypes = DEFAULT_MIME_TYPES,
|
|
28
|
+
flushTimeoutMs = DEFAULT_FLUSH_TIMEOUT_MS,
|
|
29
|
+
maxDurationMs = DEFAULT_MAX_RECORDING_MS
|
|
30
|
+
} = options;
|
|
31
|
+
return {
|
|
32
|
+
id: "recorded-upload",
|
|
33
|
+
flushTimeoutMs,
|
|
34
|
+
stoppingLabel: "Transcribing…",
|
|
35
|
+
isSupported: () => {
|
|
36
|
+
var _a;
|
|
37
|
+
return resolveMediaRecorderCtor() !== null && typeof navigator !== "undefined" && typeof ((_a = navigator.mediaDevices) == null ? void 0 : _a.getUserMedia) === "function";
|
|
38
|
+
},
|
|
39
|
+
start: ({ stream, lang, callbacks }) => {
|
|
40
|
+
const Ctor = resolveMediaRecorderCtor();
|
|
41
|
+
const noop = { stop: () => void 0, abort: () => void 0 };
|
|
42
|
+
if (!Ctor || !stream) {
|
|
43
|
+
callbacks.onError({
|
|
44
|
+
kind: !Ctor ? dictationEngine.DICTATION_ENGINE_ERROR_KINDS.unsupported : dictationEngine.DICTATION_ENGINE_ERROR_KINDS.permission,
|
|
45
|
+
message: !Ctor ? "This browser cannot record audio." : "No microphone stream was available."
|
|
46
|
+
});
|
|
47
|
+
callbacks.onEnd();
|
|
48
|
+
return noop;
|
|
49
|
+
}
|
|
50
|
+
const mimeType = pickRecorderMimeType(Ctor, mimeTypes);
|
|
51
|
+
let recorder;
|
|
52
|
+
try {
|
|
53
|
+
recorder = new Ctor(stream, mimeType ? { mimeType } : void 0);
|
|
54
|
+
} catch (error) {
|
|
55
|
+
callbacks.onError({
|
|
56
|
+
kind: dictationEngine.DICTATION_ENGINE_ERROR_KINDS.unknown,
|
|
57
|
+
message: error instanceof Error ? error.message : "Recording could not start."
|
|
58
|
+
});
|
|
59
|
+
callbacks.onEnd();
|
|
60
|
+
return noop;
|
|
61
|
+
}
|
|
62
|
+
const chunks = [];
|
|
63
|
+
const upload = new AbortController();
|
|
64
|
+
let ended = false;
|
|
65
|
+
let aborted = false;
|
|
66
|
+
let limitTimer;
|
|
67
|
+
const clearLimit = () => {
|
|
68
|
+
if (limitTimer === void 0) return;
|
|
69
|
+
clearTimeout(limitTimer);
|
|
70
|
+
limitTimer = void 0;
|
|
71
|
+
};
|
|
72
|
+
const end = () => {
|
|
73
|
+
if (ended) return;
|
|
74
|
+
ended = true;
|
|
75
|
+
clearLimit();
|
|
76
|
+
recorder.ondataavailable = null;
|
|
77
|
+
recorder.onstop = null;
|
|
78
|
+
recorder.onerror = null;
|
|
79
|
+
callbacks.onEnd();
|
|
80
|
+
};
|
|
81
|
+
const stopRecorder = () => {
|
|
82
|
+
if (recorder.state === "inactive") return false;
|
|
83
|
+
try {
|
|
84
|
+
recorder.stop();
|
|
85
|
+
return true;
|
|
86
|
+
} catch {
|
|
87
|
+
return false;
|
|
88
|
+
}
|
|
89
|
+
};
|
|
90
|
+
recorder.ondataavailable = (event) => {
|
|
91
|
+
if (event.data && event.data.size > 0) chunks.push(event.data);
|
|
92
|
+
};
|
|
93
|
+
recorder.onerror = () => {
|
|
94
|
+
callbacks.onError({
|
|
95
|
+
kind: dictationEngine.DICTATION_ENGINE_ERROR_KINDS.unknown,
|
|
96
|
+
message: "Recording stopped unexpectedly."
|
|
97
|
+
});
|
|
98
|
+
end();
|
|
99
|
+
};
|
|
100
|
+
recorder.onstop = () => {
|
|
101
|
+
if (aborted || ended) {
|
|
102
|
+
end();
|
|
103
|
+
return;
|
|
104
|
+
}
|
|
105
|
+
const clip = new Blob(chunks, { type: recorder.mimeType || mimeType || "audio/webm" });
|
|
106
|
+
if (clip.size === 0) {
|
|
107
|
+
end();
|
|
108
|
+
return;
|
|
109
|
+
}
|
|
110
|
+
void transcribe(clip, { lang, signal: upload.signal }).then((text) => {
|
|
111
|
+
if (ended) return;
|
|
112
|
+
const trimmed = text.trim();
|
|
113
|
+
if (trimmed) callbacks.onTranscript({ final: trimmed, interim: "" });
|
|
114
|
+
}).catch((error) => {
|
|
115
|
+
if (ended) return;
|
|
116
|
+
callbacks.onError(transcriptionError(error));
|
|
117
|
+
}).finally(end);
|
|
118
|
+
};
|
|
119
|
+
try {
|
|
120
|
+
recorder.start();
|
|
121
|
+
} catch (error) {
|
|
122
|
+
callbacks.onError({
|
|
123
|
+
kind: dictationEngine.DICTATION_ENGINE_ERROR_KINDS.unknown,
|
|
124
|
+
message: error instanceof Error ? error.message : "Recording could not start."
|
|
125
|
+
});
|
|
126
|
+
end();
|
|
127
|
+
return noop;
|
|
128
|
+
}
|
|
129
|
+
if (Number.isFinite(maxDurationMs) && maxDurationMs > 0) {
|
|
130
|
+
limitTimer = setTimeout(() => {
|
|
131
|
+
limitTimer = void 0;
|
|
132
|
+
if (!ended) stopRecorder();
|
|
133
|
+
}, maxDurationMs);
|
|
134
|
+
}
|
|
135
|
+
return {
|
|
136
|
+
stop: () => {
|
|
137
|
+
clearLimit();
|
|
138
|
+
if (!stopRecorder()) end();
|
|
139
|
+
},
|
|
140
|
+
abort: () => {
|
|
141
|
+
aborted = true;
|
|
142
|
+
upload.abort();
|
|
143
|
+
stopRecorder();
|
|
144
|
+
end();
|
|
145
|
+
}
|
|
146
|
+
};
|
|
147
|
+
}
|
|
148
|
+
};
|
|
149
|
+
}
|
|
150
|
+
exports.DEFAULT_MAX_RECORDING_MS = DEFAULT_MAX_RECORDING_MS;
|
|
151
|
+
exports.createRecordedDictationEngine = createRecordedDictationEngine;
|
|
152
|
+
exports.pickRecorderMimeType = pickRecorderMimeType;
|
|
153
|
+
//# sourceMappingURL=recorded-dictation-engine.cjs.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"recorded-dictation-engine.cjs","sources":["../../src/dictation/recorded-dictation-engine.ts"],"sourcesContent":["import {\n DICTATION_ENGINE_ERROR_KINDS,\n type DictationEngine,\n type DictationEngineError,\n} from './dictation-engine';\n\nexport interface DictationTranscribeContext {\n /** BCP-47 tag the session was started with (host `lang` or the document's). */\n lang: string;\n /** Fires when the user cancels or the composer gives up waiting. */\n signal: AbortSignal;\n}\n\n/**\n * Turn a finished recording into text. Receives the whole clip once the user\n * stops; resolves with the transcript (empty when nothing was said). Should\n * honour `signal` so a cancelled dictation does not keep uploading.\n */\nexport type DictationTranscriber = (clip: Blob, context: DictationTranscribeContext) => Promise<string>;\n\nexport interface RecordedDictationEngineOptions {\n transcribe: DictationTranscriber;\n /**\n * Preferred `MediaRecorder` container, tried in order until the browser\n * accepts one. Defaults to Opus in WebM (Chrome, Firefox, Edge) then MP4\n * (Safari); the browser's own default is the last resort.\n */\n mimeTypes?: readonly string[];\n /**\n * Upper bound on the transcription round trip; the in-flight request is\n * aborted when it elapses. Defaults to 90s, above agora's 60s Deepgram\n * budget plus upload time.\n */\n flushTimeoutMs?: number;\n /**\n * Longest recording the engine accepts before it stops on its own and\n * transcribes what it has, exactly as if Stop had been pressed. Bounds a\n * mic left open by mistake: every recorded minute is a minute billed by the\n * recognizer, and agora refuses clips over its size cap. Defaults to\n * 5 minutes; pass `Infinity` to disable.\n */\n maxDurationMs?: number;\n}\n\nconst DEFAULT_MIME_TYPES = ['audio/webm;codecs=opus', 'audio/webm', 'audio/mp4', 'audio/ogg;codecs=opus'];\nconst DEFAULT_FLUSH_TIMEOUT_MS = 90_000;\nexport const DEFAULT_MAX_RECORDING_MS = 5 * 60_000;\n\n// lib.dom typings for MediaRecorder vary across the TypeScript versions the\n// SDK supports; describe only what the engine touches.\ninterface MediaRecorderLike extends EventTarget {\n readonly state: 'inactive' | 'recording' | 'paused';\n readonly mimeType: string;\n start: (timeslice?: number) => void;\n stop: () => void;\n ondataavailable: ((event: { data: Blob }) => void) | null;\n onstop: (() => void) | null;\n onerror: ((event: Event) => void) | null;\n}\ninterface MediaRecorderCtor {\n new (stream: MediaStream, options?: { mimeType?: string }): MediaRecorderLike;\n isTypeSupported?: (type: string) => boolean;\n}\n\nfunction resolveMediaRecorderCtor(): MediaRecorderCtor | null {\n if (typeof window === 'undefined') return null;\n return (window as unknown as { MediaRecorder?: MediaRecorderCtor }).MediaRecorder ?? null;\n}\n\n/** First container the recorder accepts, or undefined for the browser default. */\nexport function pickRecorderMimeType(\n Ctor: Pick<MediaRecorderCtor, 'isTypeSupported'>,\n candidates: readonly string[],\n): string | undefined {\n if (!Ctor.isTypeSupported) return undefined;\n return candidates.find((type) => Ctor.isTypeSupported?.(type));\n}\n\nfunction transcriptionError(error: unknown): DictationEngineError {\n return {\n kind: DICTATION_ENGINE_ERROR_KINDS.connection,\n message: error instanceof Error && error.message ? error.message : 'Transcription failed. Try again.',\n };\n}\n\n/**\n * Record-then-transcribe: capture the microphone with `MediaRecorder` while\n * the user talks, then hand the finished clip to `transcribe` on stop. No\n * interim words are shown; the composer shows a \"transcribing\" state until\n * the text lands. Works in every browser with `MediaRecorder`, so the mic is\n * available where Web Speech is not (Firefox) and the quality is whatever the\n * server-side recognizer provides.\n */\nexport function createRecordedDictationEngine(options: RecordedDictationEngineOptions): DictationEngine {\n const {\n transcribe,\n mimeTypes = DEFAULT_MIME_TYPES,\n flushTimeoutMs = DEFAULT_FLUSH_TIMEOUT_MS,\n maxDurationMs = DEFAULT_MAX_RECORDING_MS,\n } = options;\n return {\n id: 'recorded-upload',\n flushTimeoutMs,\n stoppingLabel: 'Transcribing…',\n isSupported: () =>\n resolveMediaRecorderCtor() !== null &&\n typeof navigator !== 'undefined' &&\n typeof navigator.mediaDevices?.getUserMedia === 'function',\n start: ({ stream, lang, callbacks }) => {\n const Ctor = resolveMediaRecorderCtor();\n const noop = { stop: () => undefined, abort: () => undefined };\n if (!Ctor || !stream) {\n callbacks.onError({\n kind: !Ctor ? DICTATION_ENGINE_ERROR_KINDS.unsupported : DICTATION_ENGINE_ERROR_KINDS.permission,\n message: !Ctor ? 'This browser cannot record audio.' : 'No microphone stream was available.',\n });\n callbacks.onEnd();\n return noop;\n }\n\n const mimeType = pickRecorderMimeType(Ctor, mimeTypes);\n let recorder: MediaRecorderLike;\n try {\n recorder = new Ctor(stream, mimeType ? { mimeType } : undefined);\n } catch (error) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: error instanceof Error ? error.message : 'Recording could not start.',\n });\n callbacks.onEnd();\n return noop;\n }\n\n const chunks: Blob[] = [];\n const upload = new AbortController();\n let ended = false;\n let aborted = false;\n let limitTimer: ReturnType<typeof setTimeout> | undefined;\n const clearLimit = () => {\n if (limitTimer === undefined) return;\n clearTimeout(limitTimer);\n limitTimer = undefined;\n };\n const end = () => {\n if (ended) return;\n ended = true;\n clearLimit();\n recorder.ondataavailable = null;\n recorder.onstop = null;\n recorder.onerror = null;\n callbacks.onEnd();\n };\n const stopRecorder = () => {\n if (recorder.state === 'inactive') return false;\n try {\n recorder.stop();\n return true;\n } catch {\n return false;\n }\n };\n\n recorder.ondataavailable = (event) => {\n if (event.data && event.data.size > 0) chunks.push(event.data);\n };\n recorder.onerror = () => {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: 'Recording stopped unexpectedly.',\n });\n end();\n };\n recorder.onstop = () => {\n if (aborted || ended) {\n end();\n return;\n }\n const clip = new Blob(chunks, { type: recorder.mimeType || mimeType || 'audio/webm' });\n if (clip.size === 0) {\n end();\n return;\n }\n void transcribe(clip, { lang, signal: upload.signal })\n .then((text) => {\n if (ended) return;\n const trimmed = text.trim();\n if (trimmed) callbacks.onTranscript({ final: trimmed, interim: '' });\n })\n .catch((error: unknown) => {\n if (ended) return;\n callbacks.onError(transcriptionError(error));\n })\n .finally(end);\n };\n\n try {\n recorder.start();\n } catch (error) {\n callbacks.onError({\n kind: DICTATION_ENGINE_ERROR_KINDS.unknown,\n message: error instanceof Error ? error.message : 'Recording could not start.',\n });\n end();\n return noop;\n }\n\n if (Number.isFinite(maxDurationMs) && maxDurationMs > 0) {\n limitTimer = setTimeout(() => {\n limitTimer = undefined;\n // Same path as the user's Stop: onstop transcribes the clip and the\n // words land in the draft. A recorder that is already inactive is a\n // stop in flight, which must run to completion untouched.\n if (!ended) stopRecorder();\n }, maxDurationMs);\n }\n\n return {\n stop: () => {\n clearLimit();\n if (!stopRecorder()) end();\n },\n abort: () => {\n aborted = true;\n upload.abort();\n stopRecorder();\n end();\n },\n };\n },\n };\n}\n"],"names":["DICTATION_ENGINE_ERROR_KINDS"],"mappings":";;;AA4CA,MAAM,qBAAqB,CAAC,0BAA0B,cAAc,aAAa,uBAAuB;AACxG,MAAM,2BAA2B;AAC1B,MAAM,2BAA2B,IAAI;AAkB5C,SAAS,2BAAqD;AAC5D,MAAI,OAAO,WAAW,YAAa,QAAO;AAC1C,SAAQ,OAA4D,iBAAiB;AACvF;AAGO,SAAS,qBACd,MACA,YACoB;AACpB,MAAI,CAAC,KAAK,gBAAiB,QAAO;AAClC,SAAO,WAAW,KAAK,CAAC;;AAAS,sBAAK,oBAAL,8BAAuB;AAAA,GAAK;AAC/D;AAEA,SAAS,mBAAmB,OAAsC;AAChE,SAAO;AAAA,IACL,MAAMA,gBAAAA,6BAA6B;AAAA,IACnC,SAAS,iBAAiB,SAAS,MAAM,UAAU,MAAM,UAAU;AAAA,EAAA;AAEvE;AAUO,SAAS,8BAA8B,SAA0D;AACtG,QAAM;AAAA,IACJ;AAAA,IACA,YAAY;AAAA,IACZ,iBAAiB;AAAA,IACjB,gBAAgB;AAAA,EAAA,IACd;AACJ,SAAO;AAAA,IACL,IAAI;AAAA,IACJ;AAAA,IACA,eAAe;AAAA,IACf,aAAa,MAAA;;AACX,4CAA+B,QAC/B,OAAO,cAAc,eACrB,SAAO,eAAU,iBAAV,mBAAwB,kBAAiB;AAAA;AAAA,IAClD,OAAO,CAAC,EAAE,QAAQ,MAAM,gBAAgB;AACtC,YAAM,OAAO,yBAAA;AACb,YAAM,OAAO,EAAE,MAAM,MAAM,QAAW,OAAO,MAAM,OAAA;AACnD,UAAI,CAAC,QAAQ,CAAC,QAAQ;AACpB,kBAAU,QAAQ;AAAA,UAChB,MAAM,CAAC,OAAOA,gBAAAA,6BAA6B,cAAcA,gBAAAA,6BAA6B;AAAA,UACtF,SAAS,CAAC,OAAO,sCAAsC;AAAA,QAAA,CACxD;AACD,kBAAU,MAAA;AACV,eAAO;AAAA,MACT;AAEA,YAAM,WAAW,qBAAqB,MAAM,SAAS;AACrD,UAAI;AACJ,UAAI;AACF,mBAAW,IAAI,KAAK,QAAQ,WAAW,EAAE,SAAA,IAAa,MAAS;AAAA,MACjE,SAAS,OAAO;AACd,kBAAU,QAAQ;AAAA,UAChB,MAAMA,gBAAAA,6BAA6B;AAAA,UACnC,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAAA,QAAA,CACnD;AACD,kBAAU,MAAA;AACV,eAAO;AAAA,MACT;AAEA,YAAM,SAAiB,CAAA;AACvB,YAAM,SAAS,IAAI,gBAAA;AACnB,UAAI,QAAQ;AACZ,UAAI,UAAU;AACd,UAAI;AACJ,YAAM,aAAa,MAAM;AACvB,YAAI,eAAe,OAAW;AAC9B,qBAAa,UAAU;AACvB,qBAAa;AAAA,MACf;AACA,YAAM,MAAM,MAAM;AAChB,YAAI,MAAO;AACX,gBAAQ;AACR,mBAAA;AACA,iBAAS,kBAAkB;AAC3B,iBAAS,SAAS;AAClB,iBAAS,UAAU;AACnB,kBAAU,MAAA;AAAA,MACZ;AACA,YAAM,eAAe,MAAM;AACzB,YAAI,SAAS,UAAU,WAAY,QAAO;AAC1C,YAAI;AACF,mBAAS,KAAA;AACT,iBAAO;AAAA,QACT,QAAQ;AACN,iBAAO;AAAA,QACT;AAAA,MACF;AAEA,eAAS,kBAAkB,CAAC,UAAU;AACpC,YAAI,MAAM,QAAQ,MAAM,KAAK,OAAO,EAAG,QAAO,KAAK,MAAM,IAAI;AAAA,MAC/D;AACA,eAAS,UAAU,MAAM;AACvB,kBAAU,QAAQ;AAAA,UAChB,MAAMA,gBAAAA,6BAA6B;AAAA,UACnC,SAAS;AAAA,QAAA,CACV;AACD,YAAA;AAAA,MACF;AACA,eAAS,SAAS,MAAM;AACtB,YAAI,WAAW,OAAO;AACpB,cAAA;AACA;AAAA,QACF;AACA,cAAM,OAAO,IAAI,KAAK,QAAQ,EAAE,MAAM,SAAS,YAAY,YAAY,cAAc;AACrF,YAAI,KAAK,SAAS,GAAG;AACnB,cAAA;AACA;AAAA,QACF;AACA,aAAK,WAAW,MAAM,EAAE,MAAM,QAAQ,OAAO,QAAQ,EAClD,KAAK,CAAC,SAAS;AACd,cAAI,MAAO;AACX,gBAAM,UAAU,KAAK,KAAA;AACrB,cAAI,mBAAmB,aAAa,EAAE,OAAO,SAAS,SAAS,IAAI;AAAA,QACrE,CAAC,EACA,MAAM,CAAC,UAAmB;AACzB,cAAI,MAAO;AACX,oBAAU,QAAQ,mBAAmB,KAAK,CAAC;AAAA,QAC7C,CAAC,EACA,QAAQ,GAAG;AAAA,MAChB;AAEA,UAAI;AACF,iBAAS,MAAA;AAAA,MACX,SAAS,OAAO;AACd,kBAAU,QAAQ;AAAA,UAChB,MAAMA,gBAAAA,6BAA6B;AAAA,UACnC,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAAA,QAAA,CACnD;AACD,YAAA;AACA,eAAO;AAAA,MACT;AAEA,UAAI,OAAO,SAAS,aAAa,KAAK,gBAAgB,GAAG;AACvD,qBAAa,WAAW,MAAM;AAC5B,uBAAa;AAIb,cAAI,CAAC,MAAO,cAAA;AAAA,QACd,GAAG,aAAa;AAAA,MAClB;AAEA,aAAO;AAAA,QACL,MAAM,MAAM;AACV,qBAAA;AACA,cAAI,CAAC,aAAA,EAAgB,KAAA;AAAA,QACvB;AAAA,QACA,OAAO,MAAM;AACX,oBAAU;AACV,iBAAO,MAAA;AACP,uBAAA;AACA,cAAA;AAAA,QACF;AAAA,MAAA;AAAA,IAEJ;AAAA,EAAA;AAEJ;;;;"}
|
|
@@ -0,0 +1,66 @@
|
|
|
1
|
+
import { type DictationEngine } from './dictation-engine';
|
|
2
|
+
export interface DictationTranscribeContext {
|
|
3
|
+
/** BCP-47 tag the session was started with (host `lang` or the document's). */
|
|
4
|
+
lang: string;
|
|
5
|
+
/** Fires when the user cancels or the composer gives up waiting. */
|
|
6
|
+
signal: AbortSignal;
|
|
7
|
+
}
|
|
8
|
+
/**
|
|
9
|
+
* Turn a finished recording into text. Receives the whole clip once the user
|
|
10
|
+
* stops; resolves with the transcript (empty when nothing was said). Should
|
|
11
|
+
* honour `signal` so a cancelled dictation does not keep uploading.
|
|
12
|
+
*/
|
|
13
|
+
export type DictationTranscriber = (clip: Blob, context: DictationTranscribeContext) => Promise<string>;
|
|
14
|
+
export interface RecordedDictationEngineOptions {
|
|
15
|
+
transcribe: DictationTranscriber;
|
|
16
|
+
/**
|
|
17
|
+
* Preferred `MediaRecorder` container, tried in order until the browser
|
|
18
|
+
* accepts one. Defaults to Opus in WebM (Chrome, Firefox, Edge) then MP4
|
|
19
|
+
* (Safari); the browser's own default is the last resort.
|
|
20
|
+
*/
|
|
21
|
+
mimeTypes?: readonly string[];
|
|
22
|
+
/**
|
|
23
|
+
* Upper bound on the transcription round trip; the in-flight request is
|
|
24
|
+
* aborted when it elapses. Defaults to 90s, above agora's 60s Deepgram
|
|
25
|
+
* budget plus upload time.
|
|
26
|
+
*/
|
|
27
|
+
flushTimeoutMs?: number;
|
|
28
|
+
/**
|
|
29
|
+
* Longest recording the engine accepts before it stops on its own and
|
|
30
|
+
* transcribes what it has, exactly as if Stop had been pressed. Bounds a
|
|
31
|
+
* mic left open by mistake: every recorded minute is a minute billed by the
|
|
32
|
+
* recognizer, and agora refuses clips over its size cap. Defaults to
|
|
33
|
+
* 5 minutes; pass `Infinity` to disable.
|
|
34
|
+
*/
|
|
35
|
+
maxDurationMs?: number;
|
|
36
|
+
}
|
|
37
|
+
export declare const DEFAULT_MAX_RECORDING_MS: number;
|
|
38
|
+
interface MediaRecorderLike extends EventTarget {
|
|
39
|
+
readonly state: 'inactive' | 'recording' | 'paused';
|
|
40
|
+
readonly mimeType: string;
|
|
41
|
+
start: (timeslice?: number) => void;
|
|
42
|
+
stop: () => void;
|
|
43
|
+
ondataavailable: ((event: {
|
|
44
|
+
data: Blob;
|
|
45
|
+
}) => void) | null;
|
|
46
|
+
onstop: (() => void) | null;
|
|
47
|
+
onerror: ((event: Event) => void) | null;
|
|
48
|
+
}
|
|
49
|
+
interface MediaRecorderCtor {
|
|
50
|
+
new (stream: MediaStream, options?: {
|
|
51
|
+
mimeType?: string;
|
|
52
|
+
}): MediaRecorderLike;
|
|
53
|
+
isTypeSupported?: (type: string) => boolean;
|
|
54
|
+
}
|
|
55
|
+
/** First container the recorder accepts, or undefined for the browser default. */
|
|
56
|
+
export declare function pickRecorderMimeType(Ctor: Pick<MediaRecorderCtor, 'isTypeSupported'>, candidates: readonly string[]): string | undefined;
|
|
57
|
+
/**
|
|
58
|
+
* Record-then-transcribe: capture the microphone with `MediaRecorder` while
|
|
59
|
+
* the user talks, then hand the finished clip to `transcribe` on stop. No
|
|
60
|
+
* interim words are shown; the composer shows a "transcribing" state until
|
|
61
|
+
* the text lands. Works in every browser with `MediaRecorder`, so the mic is
|
|
62
|
+
* available where Web Speech is not (Firefox) and the quality is whatever the
|
|
63
|
+
* server-side recognizer provides.
|
|
64
|
+
*/
|
|
65
|
+
export declare function createRecordedDictationEngine(options: RecordedDictationEngineOptions): DictationEngine;
|
|
66
|
+
export {};
|