@loomcycle/client 1.139.0 → 1.141.0

package/README.md

@@ -804,6 +804,47 @@ try {
 }
 ```
 
+### Speech: read-aloud and transcription
+
+Read a reply aloud, or turn a recording into text, on the deployment's own speech engines. No run is involved; each engine request is charged to the caller and the caller's hard token budget is checked first. A chat user may call them (the scope is `runs:create`, and access control asks `use gateway:speech` / `use gateway:transcription`); a token limited to certain agents may not.
+
+| Method | Returns | Notes |
+|---|---|---|
+| `speech(input, opts?)` | `Promise<SpeechResult>` | Markdown or plain text → `{bytes, contentType, language?, voice?, seconds?, chars?, pieces?}`. Options: `language`, `voice`, `format`, `model`, `signal`. Mirrors `POST /v1/audio/speech`. |
+| `transcribe(file, opts?)` | `Promise<TranscribeResult>` | A `Blob` or `Uint8Array` → `{text, language?, duration?, segments?}`. Options: `filename`, `contentType`, `language`, `prompt`, `format` (`json` \| `verbose_json`), `signal`. Mirrors `POST /v1/audio/transcriptions`. |
+| `speechInfo()` | `Promise<SpeechInfo>` | Per direction: whether this caller may use it, languages and voices, formats, accepted recording types, caps. Mirrors `GET /v1/_speech`. |
+
+**Whether to show the controls.** `whoami()`'s `effective.surfaces.speech` and `.transcription` are true only when the engine is configured and the caller passes the route's gates, so a client never probes with a request that costs a ledger row.
+
+```ts
+const me = await client.whoami();
+if (me.effective?.surfaces.speech) {
+  const info = await client.speechInfo();
+  const format = info.speech.formats?.includes("ogg") ? "ogg" : "wav";
+  const audio = await client.speech(reply, { format, signal: abort.signal });
+  play(new Blob([audio.bytes], { type: audio.contentType }));
+}
+if (me.effective?.surfaces.transcription) {
+  const { text } = await client.transcribe(wavBlob, { language: "uk" });
+}
+```
+
+- **Pick a format from `speechInfo().speech.formats`.** It is `wav`, plus the one encoded format the deployment makes, if any; it never lists both `mp3` and `ogg`. Any other format is a 400 `speech_encode_not_configured`. `wav` always works.
+- **A recording.** A PCM WAV of any length is cut by the deployment. A browser's own `audio/webm` or `audio/mp4` is taken when `speechInfo().transcription.accepts` lists it, up to `whole_file_max_bytes` when the deployment has no converter. `transcribe` sends `FormData` and sets no `Content-Type`, so the runtime writes the multipart boundary.
+- **Under load a request waits** for the engine rather than being refused. Show progress, and pass `signal` for the user's cancel.
+- **The `seconds` / `language` / `voice` metadata** comes from response headers a browser can read only on a same-origin request; cross-origin they are `undefined`.
+
+| Status | `e.code` | What to do |
+|---|---|---|
+| 400 | `speech_encode_not_configured` / `unsupported_format` | Ask for a format from `speechInfo()`. |
+| 400 | `unsupported_voice` / `unsupported_language` / `empty_input` / `model_not_allowed` / `unreadable_audio` / `unsupported_response_format` | Fix the request; the message names the fault. |
+| 403 | `acl_denied` / `insufficient_scope` / `restricted_token` / `operator_key_restricted` | The caller may not use it. Do not retry. |
+| 413 | `input_too_long` / `audio_too_long` / `audio_too_large` / `request_too_large` | Send it in parts. |
+| 415 | `unsupported_media_type` | Send a PCM WAV. |
+| 429 | `token_limit_exceeded` | The caller is at a hard token budget. |
+| 502 | `speech_failed` / `transcription_failed` | The engine failed; try again later. |
+| 503 | `speech_not_configured` / `transcription_not_configured` / `transcription_unavailable` | The deployment has no such engine, or it is not reachable. |
+
 Two of these have a class of their own: a `prompt_too_large` throws `PromptTooLargeError` and a `token_limit_exceeded` throws `TokenLimitExceededError` (with the budget on `e.scope` / `e.used` / `e.limit`). For the rest, branch on `e.code`, not on the class: the class follows the HTTP status alone.
 
 ### Credentials (RFC AR)

package/dist/cjs/client.js

@@ -1190,6 +1190,108 @@ class LoomcycleClient {
     async listDecisionModels(opts) {
         return (0, fetch_helpers_js_1.jsonFetch)(this.ctx, "/v1/_decide/models", opts);
     }
+    /** Read `input` aloud: Markdown or plain text in, audio out, whole. The
+     *  deployment normalises the Markdown (no "asterisk", no URLs), cuts the
+     *  text, speaks it with the language's voice and joins the pieces. No run
+     *  is involved; each engine request is charged to the caller. Mirrors
+     *  `POST /v1/audio/speech`.
+     *
+     *  Under load a request waits for the engine rather than being refused, so
+     *  show progress and pass `signal` for a cancel. A refusal throws with the
+     *  code on `e.code`: 400 `speech_encode_not_configured` (a format the
+     *  deployment does not make — pick from {@link speechInfo}'s `formats`) /
+     *  `unsupported_voice` / `unsupported_language` / `empty_input` /
+     *  `model_not_allowed`, 403 (scope, access control, a token limited to
+     *  certain agents), 413 `input_too_long`, 429 `token_limit_exceeded`, 502
+     *  `speech_failed`, 503 `speech_not_configured`. */
+    async speech(input, opts) {
+        const body = { input };
+        if (opts?.language !== undefined)
+            body.language = opts.language;
+        if (opts?.voice !== undefined)
+            body.voice = opts.voice;
+        if (opts?.format !== undefined)
+            body.response_format = opts.format;
+        if (opts?.model !== undefined)
+            body.model = opts.model;
+        const headers = { "Content-Type": "application/json" };
+        if (this.ctx.authToken)
+            headers.Authorization = `Bearer ${this.ctx.authToken}`;
+        const resp = await this.ctx.fetchImpl(`${this.ctx.baseUrl}/v1/audio/speech`, {
+            method: "POST",
+            headers,
+            body: JSON.stringify(body),
+            signal: opts?.signal,
+        });
+        if (!resp.ok) {
+            await (0, fetch_helpers_js_1.raiseFromResponse)(resp);
+        }
+        const num = (name) => {
+            const v = resp.headers.get(name);
+            if (v === null || v === "")
+                return undefined;
+            const n = Number(v);
+            return Number.isFinite(n) ? n : undefined;
+        };
+        return {
+            bytes: new Uint8Array(await resp.arrayBuffer()),
+            contentType: resp.headers.get("Content-Type") ?? "",
+            language: resp.headers.get("X-Loomcycle-Speech-Language") ?? undefined,
+            voice: resp.headers.get("X-Loomcycle-Speech-Voice") ?? undefined,
+            seconds: num("X-Loomcycle-Speech-Seconds"),
+            chars: num("X-Loomcycle-Speech-Chars"),
+            pieces: num("X-Loomcycle-Speech-Pieces"),
+        };
+    }
+    /** Transcribe a recording. A PCM WAV of any length is cut by the deployment;
+     *  a browser's own `audio/webm` / `audio/mp4` is taken when the deployment
+     *  converts or its engine reads the type whole (see {@link speechInfo}'s
+     *  `accepts`). No run is involved; each engine request is charged to the
+     *  caller. Mirrors `POST /v1/audio/transcriptions`.
+     *
+     *  The body is `FormData` and no `Content-Type` is set, so the runtime
+     *  writes the multipart boundary. Under load a request waits for the engine
+     *  rather than being refused. A refusal throws with the code on `e.code`:
+     *  400 `unreadable_audio` / `unsupported_language` / `model_not_allowed`,
+     *  403, 413 `request_too_large` / `audio_too_large` / `audio_too_long`, 415
+     *  `unsupported_media_type`, 429 `token_limit_exceeded`, 502
+     *  `transcription_failed`, 503 `transcription_not_configured`. */
+    async transcribe(file, opts) {
+        const type = opts?.contentType ?? (file instanceof Blob ? file.type : "") ?? "";
+        const blob = file instanceof Blob && (opts?.contentType === undefined || opts.contentType === file.type)
+            ? file
+            : new Blob([file], type ? { type } : undefined);
+        const form = new FormData();
+        if (opts?.model !== undefined)
+            form.append("model", opts.model);
+        if (opts?.language !== undefined)
+            form.append("language", opts.language);
+        if (opts?.prompt !== undefined)
+            form.append("prompt", opts.prompt);
+        if (opts?.format !== undefined)
+            form.append("response_format", opts.format);
+        form.append("file", blob, opts?.filename ?? recordingName(type));
+        const headers = { Accept: "application/json" };
+        if (this.ctx.authToken)
+            headers.Authorization = `Bearer ${this.ctx.authToken}`;
+        const resp = await this.ctx.fetchImpl(`${this.ctx.baseUrl}/v1/audio/transcriptions`, {
+            method: "POST",
+            headers,
+            body: form,
+            signal: opts?.signal,
+        });
+        if (!resp.ok) {
+            await (0, fetch_helpers_js_1.raiseFromResponse)(resp);
+        }
+        return (await resp.json());
+    }
+    /** What a client needs to draw its voice controls: per direction, whether
+     *  this caller may use it, the languages and voices, the formats, the types
+     *  a recording may be, and the caps. A direction that is off, or that this
+     *  caller may not use, is `{ enabled: false }`. Mirrors `GET /v1/_speech`. */
+    async speechInfo(opts) {
+        return (0, fetch_helpers_js_1.jsonFetch)(this.ctx, "/v1/_speech", opts);
+    }
     /** Read the full event log for a session. Each entry has seq,
      *  run_id, ts_ns, type, event (the providers.Event payload). */
     async getTranscript(sessionId, opts) {
@@ -1399,6 +1501,8 @@ class LoomcycleClient {
             body.max_active = opts.maxActive;
         if (opts?.agents !== undefined)
             body.agents = opts.agents;
+        if (opts?.surfaces !== undefined)
+            body.surfaces = opts.surfaces;
         if (opts?.overrides !== undefined)
             body.overrides = opts.overrides;
         if (opts?.rate !== undefined) {
@@ -3751,6 +3855,18 @@ class LoomcycleClient {
 exports.LoomcycleClient = LoomcycleClient;
 /** The file name of an RFC 8187 `filename*=UTF-8''…` Content-Disposition, or
  *  a plain quoted `filename="…"`; undefined when there is neither. */
+/** recordingName names an upload that came without a file name, by its type,
+ *  so the runtime's type guess from the extension agrees with the bytes. */
+function recordingName(type) {
+    const base = type.split(";")[0].trim().toLowerCase();
+    const ext = {
+        "audio/wav": ".wav", "audio/x-wav": ".wav", "audio/wave": ".wav",
+        "audio/webm": ".webm", "video/webm": ".webm",
+        "audio/mp4": ".m4a", "audio/x-m4a": ".m4a", "video/mp4": ".mp4",
+        "audio/mpeg": ".mp3", "audio/ogg": ".ogg", "audio/flac": ".flac",
+    };
+    return "recording" + (ext[base] ?? "");
+}
 function dispositionFilename(header) {
     if (!header)
         return undefined;

package/dist/client.d.ts

@@ -26,7 +26,7 @@
 import { InteractiveSession } from "./interactive.js";
 import { LoomcycleOAuth } from "./oauth.js";
 import { ClientToolHost, type ConnectClientToolsOptions } from "./client-tools.js";
-import type { ChangeFeedFrame, StreamChangesOptions, DocumentFeedFrame, DocumentFeedOptions, CredentialListResponse, CredentialMeta, CryptoKeyEncoding, CryptoKeyExport, CryptoKeyExternal, CryptoKeyType, CryptoRequest, CryptoResultByOp, DecideRequest, DecideResponse, DecisionModelsResponse, CredentialScope, DirectoryInspection, DirectoryTenant, DirectoryUser, ErasureReport, ErasureResult, MergeSubjectsResult, ReviewQueueFilter, ReviewQueueResult, RuleStatementsResult, RevertRulingResult, ListRulingsResult, ApplyReviewPolicyResult, BindSubjectPrincipalResult, Agent, AgentEvent, AgentList, CancelAgentResult, ClientOptions, ContinueOptions, RunOverrideOptions, EffectiveConfigResponse, RunFileInput, RunInboxDownload, AgentOverridableResponse, RetuneRunResponse, RunConfigResponse, RunPromptResponse, CreateSnapshotOptions, EnsureCodeAgentOptions, EnsureCodeAgentResult, EnsureHookDefOptions, EnsureHookDefResult, HookDefInput, HookDefOverlay, HookDefRowResponse, CreateDefOptions, ForkDefOptions, HookDefForkOverlay, ScheduleDefInput, ScheduleDefOverlay, ScheduleDefRowResponse, WebhookDefInput, WebhookDefOverlay, WebhookDefRowResponse, EnsureMcpServerOptions, EnsureMcpServerResult, HealthResponse, InterruptListResponse, InterruptStatus, ListUserAgentsOptions, WalkRunsPage, AckChannelOptions, AwaitChannelsOptions, BroadcastChannelsOptions, ChannelAckResult, ChannelAwaitResult, ChannelBroadcastResult, ChannelDescriptor, ChannelPeekResult, ListTeamChannelsOptions, PeekTeamChannelOptions, ReleaseTeamChannelOptions, TeamChannelPeekResult, TeamChannelReleaseResult, TeamChannelsResult, ChannelPublishResult, ChannelPurgeResult, ChannelReleaseResult, ChannelSubscribeResult, CreateChannelOptions, ListChannelsResponse, PeekChannelOptions, PublishChannelOptions, ReleaseChannelOptions, SetMemoryEntryOptions, SetMemoryEntryResponse, SubscribeChannelOptions, UpdateChannelOptions, LibraryAgentDefinition, LibraryListResponse, LibraryMcpServerDefinition, LibrarySkillDefinition, ListUsersResponse, UserRecord, CreateUserBody, UpdateUserBody, MintedUserToken, ListUserTokensResponse, UserIdentitiesResponse, AuthGateListResponse, CreateInviteBody, CreatedInvite, AuthInviteListResponse, CreatedPasswordReset, GateIdentitiesResponse, GateTenantStatus, RunnableAgentsResponse, LLMChatOptions, LLMChatResponse, LLMChatStreamItem, LLMEmbeddingsOptions, LLMEmbeddingsResponse, MCPServerDefVerifyResult, MemoryEmbedStatsResponse, MemoryEntriesResponse, MemoryEntryResponse, MemoryBackfillResponse, MemoryPurgeResponse, MemoryReembedResponse, MemoryScopeIDsResponse, MemoryScopesResponse, MemorySearchInput, MemorySearchResponse, PauseResult, RetentionOverride, PersistentVolumesResponse, EphemeralVolumesResponse, ResolveInterruptOptions, ResumeResult, ResolverMatrix, CompactRunResult, RunTreeDeletion, ReplaySessionResult, CancelTurnResult, ReviewRunResult, RunBatchOptions, RunBatchResult, RunOptions, ConfiguredRun, ConfiguredRunPatch, RunStateStreamItem, RuntimeStateResponse, SnapshotCreateResponse, SnapshotDescriptor, SnapshotEnvelope, SnapshotRestoreResponse, StreamUserRunStatesOptions, SubstrateToolInput, SubstrateToolResponse, CreatedTeam, CreateTeamOptions, DefNamesResponse, ForkTeamOptions, ModelAliasesResponse, OperatorTokenDefNamesResponse, ProviderModelsResponse, RoutingResponse, ScheduleListResponse, SchedulePauseResult, ScheduleRunNowResult, ScheduleStateView, ListTeamsResponse, PromotedTeam, RetiredTeam, TeamBreakpoints, TeamDefDetail, TeamRunDetached, TeamRunTarget, TeamVerification, TeamDraft, TeamDraftVerification, TeamVersionList, TeamDiagram, TeamRunResult, PathToolInput, PathToolResponse, DocumentToolInput, DocumentToolResponse, DocumentCopyInput, DocumentMoveInput, DocumentTransferResult, HistoryToolInput, HistoryToolResponse, EvaluationSubmitInput, EvaluationGetInput, EvaluationListForRunInput, EvaluationListForDefInput, EvaluationAggregateInput, EvaluationInput, EvaluationSubmitResult, EvaluationRecord, EvaluationListForRunResult, EvaluationListForDefResult, EvaluationAggregate, TranscriptResponse, WhoamiResponse, UsageDimension, UsageBucket, UsageReportResponse, TokenLimit, TokenLimitsResponse, ConfigResponse, SetTokenLimitRequest, AclBinding, AclBindingListResponse, AclCatalog, AclDefInput, AclDefListResponse, AclDefNamesResponse, AclDefRow, AclRule, AclValidateResult, AuthzDecision, AuthzEvaluationRequest, AuthzEvaluationsRequest, AuthzEvaluationsResponse, AuthzExplainResponse, AuthzSimulateRequest, AuthzSimulateResponse, CreateAclBindingBody, CreateGroupBody, GroupDetail, GroupEffective, GroupListResponse, GroupMemberChange, GroupMemberKind, GroupRow, AppClientConfig, AppClientEntry, ClientDefInput, ClientDefListResponse, ClientDefNamesResponse, ClientDefRecord, ClientDefRetireResult, ClientDefVerifyResult, DeviceLoginOptions, DeviceLoginResult } from "./types.js";
+import type { ChangeFeedFrame, StreamChangesOptions, DocumentFeedFrame, DocumentFeedOptions, CredentialListResponse, CredentialMeta, CryptoKeyEncoding, CryptoKeyExport, CryptoKeyExternal, CryptoKeyType, CryptoRequest, CryptoResultByOp, DecideRequest, DecideResponse, DecisionModelsResponse, SpeechOptions, SpeechResult, TranscribeOptions, TranscribeResult, SpeechInfo, CredentialScope, DirectoryInspection, DirectoryTenant, DirectoryUser, ErasureReport, ErasureResult, MergeSubjectsResult, ReviewQueueFilter, ReviewQueueResult, RuleStatementsResult, RevertRulingResult, ListRulingsResult, ApplyReviewPolicyResult, BindSubjectPrincipalResult, Agent, AgentEvent, AgentList, CancelAgentResult, ClientOptions, ContinueOptions, RunOverrideOptions, EffectiveConfigResponse, RunFileInput, RunInboxDownload, AgentOverridableResponse, RetuneRunResponse, RunConfigResponse, RunPromptResponse, CreateSnapshotOptions, EnsureCodeAgentOptions, EnsureCodeAgentResult, EnsureHookDefOptions, EnsureHookDefResult, HookDefInput, HookDefOverlay, HookDefRowResponse, CreateDefOptions, ForkDefOptions, HookDefForkOverlay, ScheduleDefInput, ScheduleDefOverlay, ScheduleDefRowResponse, WebhookDefInput, WebhookDefOverlay, WebhookDefRowResponse, EnsureMcpServerOptions, EnsureMcpServerResult, HealthResponse, InterruptListResponse, InterruptStatus, ListUserAgentsOptions, WalkRunsPage, AckChannelOptions, AwaitChannelsOptions, BroadcastChannelsOptions, ChannelAckResult, ChannelAwaitResult, ChannelBroadcastResult, ChannelDescriptor, ChannelPeekResult, ListTeamChannelsOptions, PeekTeamChannelOptions, ReleaseTeamChannelOptions, TeamChannelPeekResult, TeamChannelReleaseResult, TeamChannelsResult, ChannelPublishResult, ChannelPurgeResult, ChannelReleaseResult, ChannelSubscribeResult, CreateChannelOptions, ListChannelsResponse, PeekChannelOptions, PublishChannelOptions, ReleaseChannelOptions, SetMemoryEntryOptions, SetMemoryEntryResponse, SubscribeChannelOptions, UpdateChannelOptions, LibraryAgentDefinition, LibraryListResponse, LibraryMcpServerDefinition, LibrarySkillDefinition, ListUsersResponse, UserRecord, CreateUserBody, UpdateUserBody, MintedUserToken, TokenSurface, ListUserTokensResponse, UserIdentitiesResponse, AuthGateListResponse, CreateInviteBody, CreatedInvite, AuthInviteListResponse, CreatedPasswordReset, GateIdentitiesResponse, GateTenantStatus, RunnableAgentsResponse, LLMChatOptions, LLMChatResponse, LLMChatStreamItem, LLMEmbeddingsOptions, LLMEmbeddingsResponse, MCPServerDefVerifyResult, MemoryEmbedStatsResponse, MemoryEntriesResponse, MemoryEntryResponse, MemoryBackfillResponse, MemoryPurgeResponse, MemoryReembedResponse, MemoryScopeIDsResponse, MemoryScopesResponse, MemorySearchInput, MemorySearchResponse, PauseResult, RetentionOverride, PersistentVolumesResponse, EphemeralVolumesResponse, ResolveInterruptOptions, ResumeResult, ResolverMatrix, CompactRunResult, RunTreeDeletion, ReplaySessionResult, CancelTurnResult, ReviewRunResult, RunBatchOptions, RunBatchResult, RunOptions, ConfiguredRun, ConfiguredRunPatch, RunStateStreamItem, RuntimeStateResponse, SnapshotCreateResponse, SnapshotDescriptor, SnapshotEnvelope, SnapshotRestoreResponse, StreamUserRunStatesOptions, SubstrateToolInput, SubstrateToolResponse, CreatedTeam, CreateTeamOptions, DefNamesResponse, ForkTeamOptions, ModelAliasesResponse, OperatorTokenDefNamesResponse, ProviderModelsResponse, RoutingResponse, ScheduleListResponse, SchedulePauseResult, ScheduleRunNowResult, ScheduleStateView, ListTeamsResponse, PromotedTeam, RetiredTeam, TeamBreakpoints, TeamDefDetail, TeamRunDetached, TeamRunTarget, TeamVerification, TeamDraft, TeamDraftVerification, TeamVersionList, TeamDiagram, TeamRunResult, PathToolInput, PathToolResponse, DocumentToolInput, DocumentToolResponse, DocumentCopyInput, DocumentMoveInput, DocumentTransferResult, HistoryToolInput, HistoryToolResponse, EvaluationSubmitInput, EvaluationGetInput, EvaluationListForRunInput, EvaluationListForDefInput, EvaluationAggregateInput, EvaluationInput, EvaluationSubmitResult, EvaluationRecord, EvaluationListForRunResult, EvaluationListForDefResult, EvaluationAggregate, TranscriptResponse, WhoamiResponse, UsageDimension, UsageBucket, UsageReportResponse, TokenLimit, TokenLimitsResponse, ConfigResponse, SetTokenLimitRequest, AclBinding, AclBindingListResponse, AclCatalog, AclDefInput, AclDefListResponse, AclDefNamesResponse, AclDefRow, AclRule, AclValidateResult, AuthzDecision, AuthzEvaluationRequest, AuthzEvaluationsRequest, AuthzEvaluationsResponse, AuthzExplainResponse, AuthzSimulateRequest, AuthzSimulateResponse, CreateAclBindingBody, CreateGroupBody, GroupDetail, GroupEffective, GroupListResponse, GroupMemberChange, GroupMemberKind, GroupRow, AppClientConfig, AppClientEntry, ClientDefInput, ClientDefListResponse, ClientDefNamesResponse, ClientDefRecord, ClientDefRetireResult, ClientDefVerifyResult, DeviceLoginOptions, DeviceLoginResult } from "./types.js";
 export declare class LoomcycleClient {
     private ctx;
     private readonly tokenProvider?;
@@ -785,6 +785,42 @@ export declare class LoomcycleClient {
     listDecisionModels(opts?: {
         signal?: AbortSignal;
     }): Promise<DecisionModelsResponse>;
+    /** Read `input` aloud: Markdown or plain text in, audio out, whole. The
+     *  deployment normalises the Markdown (no "asterisk", no URLs), cuts the
+     *  text, speaks it with the language's voice and joins the pieces. No run
+     *  is involved; each engine request is charged to the caller. Mirrors
+     *  `POST /v1/audio/speech`.
+     *
+     *  Under load a request waits for the engine rather than being refused, so
+     *  show progress and pass `signal` for a cancel. A refusal throws with the
+     *  code on `e.code`: 400 `speech_encode_not_configured` (a format the
+     *  deployment does not make — pick from {@link speechInfo}'s `formats`) /
+     *  `unsupported_voice` / `unsupported_language` / `empty_input` /
+     *  `model_not_allowed`, 403 (scope, access control, a token limited to
+     *  certain agents), 413 `input_too_long`, 429 `token_limit_exceeded`, 502
+     *  `speech_failed`, 503 `speech_not_configured`. */
+    speech(input: string, opts?: SpeechOptions): Promise<SpeechResult>;
+    /** Transcribe a recording. A PCM WAV of any length is cut by the deployment;
+     *  a browser's own `audio/webm` / `audio/mp4` is taken when the deployment
+     *  converts or its engine reads the type whole (see {@link speechInfo}'s
+     *  `accepts`). No run is involved; each engine request is charged to the
+     *  caller. Mirrors `POST /v1/audio/transcriptions`.
+     *
+     *  The body is `FormData` and no `Content-Type` is set, so the runtime
+     *  writes the multipart boundary. Under load a request waits for the engine
+     *  rather than being refused. A refusal throws with the code on `e.code`:
+     *  400 `unreadable_audio` / `unsupported_language` / `model_not_allowed`,
+     *  403, 413 `request_too_large` / `audio_too_large` / `audio_too_long`, 415
+     *  `unsupported_media_type`, 429 `token_limit_exceeded`, 502
+     *  `transcription_failed`, 503 `transcription_not_configured`. */
+    transcribe(file: Blob | Uint8Array, opts?: TranscribeOptions): Promise<TranscribeResult>;
+    /** What a client needs to draw its voice controls: per direction, whether
+     *  this caller may use it, the languages and voices, the formats, the types
+     *  a recording may be, and the caps. A direction that is off, or that this
+     *  caller may not use, is `{ enabled: false }`. Mirrors `GET /v1/_speech`. */
+    speechInfo(opts?: {
+        signal?: AbortSignal;
+    }): Promise<SpeechInfo>;
     /** Read the full event log for a session. Each entry has seq,
      *  run_id, ts_ns, type, event (the providers.Event payload). */
     getTranscript(sessionId: string, opts?: {
@@ -982,6 +1018,10 @@ export declare class LoomcycleClient {
         name?: string;
         maxActive?: number;
         agents?: string[];
+        /** With `agents`: the off-run surfaces the token may still call, e.g.
+         *  `["speech", "transcription"]` for a chat bridge that exchanges voice
+         *  messages. Refused without `agents`; held to the caller's own. */
+        surfaces?: TokenSurface[];
         overrides?: "none" | string[];
         rate?: {
             perMinute: number;

package/dist/client.js

@@ -1187,6 +1187,108 @@ export class LoomcycleClient {
     async listDecisionModels(opts) {
         return jsonFetch(this.ctx, "/v1/_decide/models", opts);
     }
+    /** Read `input` aloud: Markdown or plain text in, audio out, whole. The
+     *  deployment normalises the Markdown (no "asterisk", no URLs), cuts the
+     *  text, speaks it with the language's voice and joins the pieces. No run
+     *  is involved; each engine request is charged to the caller. Mirrors
+     *  `POST /v1/audio/speech`.
+     *
+     *  Under load a request waits for the engine rather than being refused, so
+     *  show progress and pass `signal` for a cancel. A refusal throws with the
+     *  code on `e.code`: 400 `speech_encode_not_configured` (a format the
+     *  deployment does not make — pick from {@link speechInfo}'s `formats`) /
+     *  `unsupported_voice` / `unsupported_language` / `empty_input` /
+     *  `model_not_allowed`, 403 (scope, access control, a token limited to
+     *  certain agents), 413 `input_too_long`, 429 `token_limit_exceeded`, 502
+     *  `speech_failed`, 503 `speech_not_configured`. */
+    async speech(input, opts) {
+        const body = { input };
+        if (opts?.language !== undefined)
+            body.language = opts.language;
+        if (opts?.voice !== undefined)
+            body.voice = opts.voice;
+        if (opts?.format !== undefined)
+            body.response_format = opts.format;
+        if (opts?.model !== undefined)
+            body.model = opts.model;
+        const headers = { "Content-Type": "application/json" };
+        if (this.ctx.authToken)
+            headers.Authorization = `Bearer ${this.ctx.authToken}`;
+        const resp = await this.ctx.fetchImpl(`${this.ctx.baseUrl}/v1/audio/speech`, {
+            method: "POST",
+            headers,
+            body: JSON.stringify(body),
+            signal: opts?.signal,
+        });
+        if (!resp.ok) {
+            await raiseFromResponse(resp);
+        }
+        const num = (name) => {
+            const v = resp.headers.get(name);
+            if (v === null || v === "")
+                return undefined;
+            const n = Number(v);
+            return Number.isFinite(n) ? n : undefined;
+        };
+        return {
+            bytes: new Uint8Array(await resp.arrayBuffer()),
+            contentType: resp.headers.get("Content-Type") ?? "",
+            language: resp.headers.get("X-Loomcycle-Speech-Language") ?? undefined,
+            voice: resp.headers.get("X-Loomcycle-Speech-Voice") ?? undefined,
+            seconds: num("X-Loomcycle-Speech-Seconds"),
+            chars: num("X-Loomcycle-Speech-Chars"),
+            pieces: num("X-Loomcycle-Speech-Pieces"),
+        };
+    }
+    /** Transcribe a recording. A PCM WAV of any length is cut by the deployment;
+     *  a browser's own `audio/webm` / `audio/mp4` is taken when the deployment
+     *  converts or its engine reads the type whole (see {@link speechInfo}'s
+     *  `accepts`). No run is involved; each engine request is charged to the
+     *  caller. Mirrors `POST /v1/audio/transcriptions`.
+     *
+     *  The body is `FormData` and no `Content-Type` is set, so the runtime
+     *  writes the multipart boundary. Under load a request waits for the engine
+     *  rather than being refused. A refusal throws with the code on `e.code`:
+     *  400 `unreadable_audio` / `unsupported_language` / `model_not_allowed`,
+     *  403, 413 `request_too_large` / `audio_too_large` / `audio_too_long`, 415
+     *  `unsupported_media_type`, 429 `token_limit_exceeded`, 502
+     *  `transcription_failed`, 503 `transcription_not_configured`. */
+    async transcribe(file, opts) {
+        const type = opts?.contentType ?? (file instanceof Blob ? file.type : "") ?? "";
+        const blob = file instanceof Blob && (opts?.contentType === undefined || opts.contentType === file.type)
+            ? file
+            : new Blob([file], type ? { type } : undefined);
+        const form = new FormData();
+        if (opts?.model !== undefined)
+            form.append("model", opts.model);
+        if (opts?.language !== undefined)
+            form.append("language", opts.language);
+        if (opts?.prompt !== undefined)
+            form.append("prompt", opts.prompt);
+        if (opts?.format !== undefined)
+            form.append("response_format", opts.format);
+        form.append("file", blob, opts?.filename ?? recordingName(type));
+        const headers = { Accept: "application/json" };
+        if (this.ctx.authToken)
+            headers.Authorization = `Bearer ${this.ctx.authToken}`;
+        const resp = await this.ctx.fetchImpl(`${this.ctx.baseUrl}/v1/audio/transcriptions`, {
+            method: "POST",
+            headers,
+            body: form,
+            signal: opts?.signal,
+        });
+        if (!resp.ok) {
+            await raiseFromResponse(resp);
+        }
+        return (await resp.json());
+    }
+    /** What a client needs to draw its voice controls: per direction, whether
+     *  this caller may use it, the languages and voices, the formats, the types
+     *  a recording may be, and the caps. A direction that is off, or that this
+     *  caller may not use, is `{ enabled: false }`. Mirrors `GET /v1/_speech`. */
+    async speechInfo(opts) {
+        return jsonFetch(this.ctx, "/v1/_speech", opts);
+    }
     /** Read the full event log for a session. Each entry has seq,
      *  run_id, ts_ns, type, event (the providers.Event payload). */
     async getTranscript(sessionId, opts) {
@@ -1396,6 +1498,8 @@ export class LoomcycleClient {
             body.max_active = opts.maxActive;
         if (opts?.agents !== undefined)
             body.agents = opts.agents;
+        if (opts?.surfaces !== undefined)
+            body.surfaces = opts.surfaces;
         if (opts?.overrides !== undefined)
             body.overrides = opts.overrides;
         if (opts?.rate !== undefined) {
@@ -3747,6 +3851,18 @@ export class LoomcycleClient {
 }
 /** The file name of an RFC 8187 `filename*=UTF-8''…` Content-Disposition, or
  *  a plain quoted `filename="…"`; undefined when there is neither. */
+/** recordingName names an upload that came without a file name, by its type,
+ *  so the runtime's type guess from the extension agrees with the bytes. */
+function recordingName(type) {
+    const base = type.split(";")[0].trim().toLowerCase();
+    const ext = {
+        "audio/wav": ".wav", "audio/x-wav": ".wav", "audio/wave": ".wav",
+        "audio/webm": ".webm", "video/webm": ".webm",
+        "audio/mp4": ".m4a", "audio/x-m4a": ".m4a", "video/mp4": ".mp4",
+        "audio/mpeg": ".mp3", "audio/ogg": ".ogg", "audio/flac": ".flac",
+    };
+    return "recording" + (ext[base] ?? "");
+}
 function dispositionFilename(header) {
     if (!header)
         return undefined;

package/dist/index.d.ts

@@ -218,5 +218,5 @@ export { ClientToolHost, clientToolsURL, CLIENT_TOOL_SUBPROTOCOL } from "./clien
 export type { ClientToolSchema, ClientToolInvocation, ConnectClientToolsOptions, WebSocketLike, WebSocketCtor, } from "./client-tools.js";
 export { LoomcycleOAuth, OAuthTokenProvider, parseAuthorizeResponse, pkceChallenge } from "./oauth.js";
 export type { LoomcycleOAuthOptions, OAuthTokenProviderOptions } from "./oauth.js";
-export type { AgentEvent, ClientOptions, ContinueOptions, EventType, HostWidening, ImageMediaType, ParentContext, PromptContent, PromptSegment, RetryInfo, RunOptions, RunOverrideOptions, SamplingOptions, CompactionOptions, ContextOptions, ToolUse, Usage, CallTiming, Agent, AgentStatus, AwaitedState, AgentUsage, RunResult, RunSpec, RunSpecTeam, ConfiguredRun, ConfiguredRunPatch, ToolChoiceOptions, OutputFormatOptions, CancelAgentResult, ListAgentsResponse, ListUserAgentsOptions, AgentList, WalkRunsPage, RunBatchOptions, RunBatchResult, SpawnRunResult, CompactRunResult, RunTreeDeletion, RunTreeNotReached, DirectoryBudget, DirectoryInspection, DirectoryTenant, DirectoryUser, ErasureAccess, ErasureReport, ErasureResidue, ErasureResult, ErasureTier, MergeSubjectsResult, SubjectRef, BindSubjectPrincipalResult, ReviewQueueFilter, ReviewQueueGroup, ReviewQueueResult, RuleStatementsResult, RevertRulingResult, RulingBreakdown, RulingItem, RulingSelector, RulingRecord, ListRulingsResult, PolicyRuleCount, ApplyReviewPolicyResult, ErasureBoundSubject, ErasureTokenRef, CancelTurnResult, ReviewRunResult, ReplaySessionResult, TranscriptEvent, TranscriptResponse, HealthResponse, ListUsersResponse, UserSummary, UserStatus, UserRecord, CreateUserBody, UpdateUserBody, MintedUserToken, TokenRestrictions, UserTokenMeta, ListUserTokensResponse, UserIdentity, UserIdentitiesResponse, AuthGateKind, AuthGate, AuthGateKey, AuthGateRemoval, AuthGateProbe, AuthGateListResponse, CreateInviteBody, CreatedInvite, AuthInviteState, AuthInvite, AuthInviteListResponse, CreatedPasswordReset, GateIdentity, GateIdentitiesResponse, GateTenantStatus, RunnableAgent, RunnableAgentsResponse, WhoamiResponse, WhoamiEffective, WhoamiAuth, PauseResult, RetentionOverride, ResumeResult, RuntimeStateResponse, RuntimeStateStatus, ResolverMatrix, ResolverModelStatus, ResolverProviderAvailability, CreateSnapshotOptions, SnapshotCreateResponse, SnapshotDescriptor, SnapshotEnvelope, SnapshotListResponse, SnapshotRestoreResponse, MemoryEntriesResponse, MemoryEntry, MemoryEntryResponse, MemoryScopeIDsResponse, MemoryScopeIDSummary, MemoryScopeKind, MemoryScopesResponse, MemorySearchDocumentFilter, MemorySearchInput, MemorySource, MemoryWhen, MemoryTimeFilter, MemorySearchEntry, MemorySearchResponse, MemoryEmbedModelStats, MemoryEmbedStatsResponse, MemoryReembedConfigured, MemoryReembedResponse, MemoryBackfillResponse, MemoryPurgeResponse, InterruptListResponse, InterruptRow, InterruptStatus, ResolveInterruptOptions, ChannelHookCall, ChannelHookResult, HookFailMode, HookPhase, HookToolCall, HookToolResult, PostHookCall, PostHookResult, PreHookCall, PreHookResult, SubstrateToolInput, SubstrateToolResponse, TeamNameSummary, ListTeamsResponse, TeamDiagram, CreatedTeam, PromotedTeam, RetiredTeam, TeamDefDetail, TeamVerification, TeamDraft, TeamDraftVerification, TeamIssue, TeamIssueKind, TeamVersion, TeamVersionList, TeamBreakpoints, TeamRunDetached, TeamRunResult, TeamRunTarget, CreateTeamOptions, ForkTeamOptions, DefNameSummary, DefNamesResponse, OperatorTokenDefNameSummary, OperatorTokenDefNamesResponse, ScheduleListEntry, ScheduleListResponse, ScheduleActiveRun, ScheduleStateView, ScheduleRunNowResult, SchedulePauseResult, ModelAlias, ModelAliasesResponse, RoutingResponse, RoutingUserTier, RoutingTier, RoutingCandidate, RoutingThroughput, RoutingProvider, SearchRoutingProvider, RoutingEmbedder, RoutingDecision, RoutingDecisionModel, ProviderModelsResponse, TeamRunStep, TeamDecisionHandler, TeamDecisionQuestion, TeamDecisionAnswer, PathToolInput, PathToolResponse, DocumentToolInput, DocumentEdgeEvidence, DocumentToolResponse, DocumentCopyInput, DocumentMoveInput, DocumentTransferEntry, DocumentTransferResult, DocumentSearchFilter, DocumentSearchHit, DocumentSearchResult, DocumentRelatedResult, GraphSnapshot, HistoryToolInput, HistoryToolResponse, EvaluationEmitterRole, EvaluationSubmitInput, EvaluationGetInput, EvaluationListForRunInput, EvaluationListForDefInput, EvaluationAggregateInput, EvaluationInput, EvaluationSubmitResult, EvaluationRecord, EvaluationListForRunResult, EvaluationListForDefResult, EvaluationScoreStats, EvaluationAggregate, VolumeMode, PersistentVolumeEntry, PersistentVolumesResponse, EphemeralVolumeEntry, EphemeralVolumesResponse, SystemPromptPayload, UserInputPayload, ChannelDescriptor, ListChannelsResponse, RunStateEvent, RunStateStreamClose, RunStateStreamItem, RunStateStreamOpen, StreamUserRunStatesOptions, ChangeFeedError, ChangeFeedFrame, ChangeFeedStatus, EmbedderHealth, MemoryChange, MemoryChangeScope, MemoryChangeType, StreamChangesOptions, DocumentFeedChange, DocumentFeedChangeType, DocumentFeedFrame, DocumentFeedOptions, DocumentFeedStatus, AckChannelOptions, ChannelAckResult, ChannelMessageItem, ChannelPeekResult, ListTeamChannelsOptions, PeekTeamChannelOptions, ReleaseTeamChannelOptions, TeamChannelReleaseResult, TeamChannelDeclaredIn, TeamChannelDescriptor, TeamChannelPeekResult, TeamChannelsResult, ChannelPublishResult, ChannelPurgeResult, ChannelReleaseResult, ChannelScope, ChannelSubscribeResult, PeekChannelOptions, PublishChannelOptions, SubscribeChannelOptions, AwaitChannelsOptions, BroadcastChannelsOptions, ChannelAwaitEntry, ChannelAwaitMode, ChannelAwaitResult, ChannelBroadcastEntry, ChannelBroadcastResult, CreateChannelOptions, ReleaseChannelOptions, UpdateChannelOptions, SetMemoryEntryOptions, SetMemoryEntryResponse, AgentDefRowResponse, AgentDefVerifyResult, SkillDefVerifyResult, EnsureMcpServerOptions, EnsureMcpServerResult, MCPServerDefRowResponse, MCPServerDefVerifyResult, AgentDefOverlay, AgentRouting, AgentChannelACL, AgentInterruptionACL, AgentDecision, MemoryRerank, AgentSampling, AgentCompaction, AgentTranscription, AgentSpeech, AgentContext, TierCandidate, CoreBlock, HookEvent, InlineWebhook, HookEntry, EventHooks, ToolEntry, ToolHooksByTool, EnsureCodeAgentOptions, EnsureCodeAgentResult, CreateDefOptions, ForkDefOptions, EnsureHookDefOptions, EnsureHookDefResult, HookDefBody, HookDefForkOverlay, HookDefInput, HookDefMatch, HookDefOverlay, HookDefRowResponse, HookDefVerifyResult, ScheduleDefInput, ScheduleDefOverlay, ScheduleDefRowResponse, ScheduledRunHook, ScheduledRunPromptContent, ScheduledRunPromptSegment, WebhookDefAuth, WebhookDefFilter, WebhookDefInput, WebhookDefOverlay, WebhookDefRateLimit, WebhookDefRowResponse, WebhookDefDeliveries, WebhookDefSyncResponse, LibraryAgentDefinition, LibraryEntry, LibraryListResponse, LibraryMcpServerDefinition, LibrarySkillDefinition, LLMChatContent, LLMChatMessage, LLMChatOptions, LLMChatResponse, LLMChatStreamDelta, LLMChatStreamItem, LLMChatToolCall, LLMChatUsage, LLMTool, LLMEmbeddingItem, LLMEmbeddingsOptions, LLMEmbeddingsResponse, LLMEmbeddingsUsage, UsageDimension, UsageBucket, UsageAggregate, UsageReportResponse, CapabilityInertInfo, LimitInfo, FileCapturedInfo, TranscriptionInfo, AdmissionInfo, ErrorInfo, CallFormat, CallRef, HookDecisionInfo, TeamDecisionInfo, OverrideInfo, FallbackInfo, InterruptionEventInfo, SpawnChildInfo, TurnCancelledInfo, ChannelEventInfo, ContextCompactionInfo, ContextRecapInfo, ContextStateInfo, ContextExhaustedInfo, ContextTierVerdict, ContextDistillDeclinedInfo, EffectiveConfigResponse, RunFileInput, RunFileImportSource, RunFileImportHeading, RunInboxDownload, EffectiveConfigSource, EffectiveValue, InertContextSetting, AgentOverridableResponse, OverridableField, OverrideLockReason, RetuneRunResponse, RunConfigRecord, SubagentHoldInfo, RunConfigResponse, RunPromptResponse, RunPromptSnapshot, PromptBlock, TokenLimit, TokenLimitsResponse, ConfigResponse, ConfigProvider, ConfigModel, ConfigSearch, ConfigInstance, SetTokenLimitRequest, CredentialScope, CredentialMeta, CredentialListResponse, CryptoKeyVersion, CryptoKeyExport, CryptoKeyType, CryptoKeyEncoding, CryptoKeyBackend, CryptoKeyExternal, CryptoOp, CryptoRequestBase, CryptoKeyRef, CryptoEncryptRequest, CryptoDecryptRequest, CryptoRewrapRequest, CryptoMacRequest, CryptoVerifyMacRequest, CryptoSignRequest, CryptoVerifyRequest, CryptoPublicKeyRequest, CryptoTotpRequest, CryptoHashRequest, CryptoRandomRequest, CryptoHashPasswordRequest, CryptoVerifyPasswordRequest, CryptoRequest, CryptoEncryptResult, CryptoDecryptResult, CryptoRewrapResult, CryptoMacResult, CryptoVerifyResult, CryptoSignResult, CryptoPublicJWK, CryptoPublicKeyResult, CryptoTotpResult, CryptoHashResult, CryptoRandomResult, CryptoHashPasswordResult, CryptoVerifyPasswordResult, CryptoResultByOp, CryptoResult, DecideRequest, DecideResponse, DecisionQuestion, DecisionChoiceQuestion, DecisionNoulQuestion, DecisionScoreQuestion, DecisionAnswer, DecisionChoiceAnswer, DecisionNoulAnswer, DecisionScoreAnswer, DecisionModel, DecisionModelsResponse, AclMode, AclSource, AclEffect, AclSubjectKind, AclCondition, AclRule, AclList, AclDefRow, AclDefListResponse, AclDefNamesResponse, AclValidateResult, AclCatalogType, AclCatalog, AclDefOp, AclDefInput, GroupRow, GroupMemberKind, GroupMember, GroupListResponse, GroupDetail, GroupEffective, CreateGroupBody, GroupMemberChange, AclBinding, AclBindingListResponse, CreateAclBindingBody, AuthzReason, AuthzDecision, AuthzRequestContext, AuthzEvaluationRequest, AuthzEvaluationsSemantic, AuthzEvaluationsRequest, AuthzEvaluationsResponse, AuthzTrace, AuthzAppliedBinding, AuthzExplainResponse, AuthzDraftBinding, AuthzSimulateRequest, AuthzSimulateResult, AuthzSimulateResponse, OAuthServerMetadata, OAuthTokenResponse, OAuthTokenSet, BeginAuthorizeOptions, AuthorizeRequest, AuthorizeResponseParams, CompleteAuthorizeOptions, RefreshTokenOptions, RevokeTokenOptions, TokenProvider, AppClientRefresh, AppClientConfig, ClientDefInput, ClientDefRecord, ConfigAppClient, AppClientEntry, ClientDefListResponse, ClientDefVerifyResult, ClientDefRetireResult, ClientDefNameSummary, ClientDefNamesResponse, DeviceLoginPrompt, DeviceLoginOptions, DeviceLoginResult, } from "./types.js";
+export type { AgentEvent, ClientOptions, ContinueOptions, EventType, HostWidening, ImageMediaType, ParentContext, PromptContent, PromptSegment, RetryInfo, RunOptions, RunOverrideOptions, SamplingOptions, CompactionOptions, ContextOptions, ToolUse, Usage, CallTiming, Agent, AgentStatus, AwaitedState, AgentUsage, RunResult, RunSpec, RunSpecTeam, ConfiguredRun, ConfiguredRunPatch, ToolChoiceOptions, OutputFormatOptions, CancelAgentResult, ListAgentsResponse, ListUserAgentsOptions, AgentList, WalkRunsPage, RunBatchOptions, RunBatchResult, SpawnRunResult, CompactRunResult, RunTreeDeletion, RunTreeNotReached, DirectoryBudget, DirectoryInspection, DirectoryTenant, DirectoryUser, ErasureAccess, ErasureReport, ErasureResidue, ErasureResult, ErasureTier, MergeSubjectsResult, SubjectRef, BindSubjectPrincipalResult, ReviewQueueFilter, ReviewQueueGroup, ReviewQueueResult, RuleStatementsResult, RevertRulingResult, RulingBreakdown, RulingItem, RulingSelector, RulingRecord, ListRulingsResult, PolicyRuleCount, ApplyReviewPolicyResult, ErasureBoundSubject, ErasureTokenRef, CancelTurnResult, ReviewRunResult, ReplaySessionResult, TranscriptEvent, TranscriptResponse, HealthResponse, ListUsersResponse, UserSummary, UserStatus, UserRecord, CreateUserBody, UpdateUserBody, MintedUserToken, TokenRestrictions, TokenSurface, UserTokenMeta, ListUserTokensResponse, UserIdentity, UserIdentitiesResponse, AuthGateKind, AuthGate, AuthGateKey, AuthGateRemoval, AuthGateProbe, AuthGateListResponse, CreateInviteBody, CreatedInvite, AuthInviteState, AuthInvite, AuthInviteListResponse, CreatedPasswordReset, GateIdentity, GateIdentitiesResponse, GateTenantStatus, RunnableAgent, RunnableAgentsResponse, WhoamiResponse, WhoamiEffective, WhoamiAuth, PauseResult, RetentionOverride, ResumeResult, RuntimeStateResponse, RuntimeStateStatus, ResolverMatrix, ResolverModelStatus, ResolverProviderAvailability, CreateSnapshotOptions, SnapshotCreateResponse, SnapshotDescriptor, SnapshotEnvelope, SnapshotListResponse, SnapshotRestoreResponse, MemoryEntriesResponse, MemoryEntry, MemoryEntryResponse, MemoryScopeIDsResponse, MemoryScopeIDSummary, MemoryScopeKind, MemoryScopesResponse, MemorySearchDocumentFilter, MemorySearchInput, MemorySource, MemoryWhen, MemoryTimeFilter, MemorySearchEntry, MemorySearchResponse, MemoryEmbedModelStats, MemoryEmbedStatsResponse, MemoryReembedConfigured, MemoryReembedResponse, MemoryBackfillResponse, MemoryPurgeResponse, InterruptListResponse, InterruptRow, InterruptStatus, ResolveInterruptOptions, ChannelHookCall, ChannelHookResult, HookFailMode, HookPhase, HookToolCall, HookToolResult, PostHookCall, PostHookResult, PreHookCall, PreHookResult, SubstrateToolInput, SubstrateToolResponse, TeamNameSummary, ListTeamsResponse, TeamDiagram, CreatedTeam, PromotedTeam, RetiredTeam, TeamDefDetail, TeamVerification, TeamDraft, TeamDraftVerification, TeamIssue, TeamIssueKind, TeamVersion, TeamVersionList, TeamBreakpoints, TeamRunDetached, TeamRunResult, TeamRunTarget, CreateTeamOptions, ForkTeamOptions, DefNameSummary, DefNamesResponse, OperatorTokenDefNameSummary, OperatorTokenDefNamesResponse, ScheduleListEntry, ScheduleListResponse, ScheduleActiveRun, ScheduleStateView, ScheduleRunNowResult, SchedulePauseResult, ModelAlias, ModelAliasesResponse, RoutingResponse, RoutingUserTier, RoutingTier, RoutingCandidate, RoutingThroughput, RoutingProvider, SearchRoutingProvider, RoutingEmbedder, RoutingDecision, RoutingDecisionModel, ProviderModelsResponse, TeamRunStep, TeamDecisionHandler, TeamDecisionQuestion, TeamDecisionAnswer, PathToolInput, PathToolResponse, DocumentToolInput, DocumentEdgeEvidence, DocumentToolResponse, DocumentCopyInput, DocumentMoveInput, DocumentTransferEntry, DocumentTransferResult, DocumentSearchFilter, DocumentSearchHit, DocumentSearchResult, DocumentRelatedResult, GraphSnapshot, HistoryToolInput, HistoryToolResponse, EvaluationEmitterRole, EvaluationSubmitInput, EvaluationGetInput, EvaluationListForRunInput, EvaluationListForDefInput, EvaluationAggregateInput, EvaluationInput, EvaluationSubmitResult, EvaluationRecord, EvaluationListForRunResult, EvaluationListForDefResult, EvaluationScoreStats, EvaluationAggregate, VolumeMode, PersistentVolumeEntry, PersistentVolumesResponse, EphemeralVolumeEntry, EphemeralVolumesResponse, SystemPromptPayload, UserInputPayload, ChannelDescriptor, ListChannelsResponse, RunStateEvent, RunStateStreamClose, RunStateStreamItem, RunStateStreamOpen, StreamUserRunStatesOptions, ChangeFeedError, ChangeFeedFrame, ChangeFeedStatus, EmbedderHealth, MemoryChange, MemoryChangeScope, MemoryChangeType, StreamChangesOptions, DocumentFeedChange, DocumentFeedChangeType, DocumentFeedFrame, DocumentFeedOptions, DocumentFeedStatus, AckChannelOptions, ChannelAckResult, ChannelMessageItem, ChannelPeekResult, ListTeamChannelsOptions, PeekTeamChannelOptions, ReleaseTeamChannelOptions, TeamChannelReleaseResult, TeamChannelDeclaredIn, TeamChannelDescriptor, TeamChannelPeekResult, TeamChannelsResult, ChannelPublishResult, ChannelPurgeResult, ChannelReleaseResult, ChannelScope, ChannelSubscribeResult, PeekChannelOptions, PublishChannelOptions, SubscribeChannelOptions, AwaitChannelsOptions, BroadcastChannelsOptions, ChannelAwaitEntry, ChannelAwaitMode, ChannelAwaitResult, ChannelBroadcastEntry, ChannelBroadcastResult, CreateChannelOptions, ReleaseChannelOptions, UpdateChannelOptions, SetMemoryEntryOptions, SetMemoryEntryResponse, AgentDefRowResponse, AgentDefVerifyResult, SkillDefVerifyResult, EnsureMcpServerOptions, EnsureMcpServerResult, MCPServerDefRowResponse, MCPServerDefVerifyResult, AgentDefOverlay, AgentRouting, AgentChannelACL, AgentInterruptionACL, AgentDecision, MemoryRerank, AgentSampling, AgentCompaction, AgentTranscription, AgentSpeech, AgentContext, TierCandidate, CoreBlock, HookEvent, InlineWebhook, HookEntry, EventHooks, ToolEntry, ToolHooksByTool, EnsureCodeAgentOptions, EnsureCodeAgentResult, CreateDefOptions, ForkDefOptions, EnsureHookDefOptions, EnsureHookDefResult, HookDefBody, HookDefForkOverlay, HookDefInput, HookDefMatch, HookDefOverlay, HookDefRowResponse, HookDefVerifyResult, ScheduleDefInput, ScheduleDefOverlay, ScheduleDefRowResponse, ScheduledRunHook, ScheduledRunPromptContent, ScheduledRunPromptSegment, WebhookDefAuth, WebhookDefFilter, WebhookDefInput, WebhookDefOverlay, WebhookDefRateLimit, WebhookDefRowResponse, WebhookDefDeliveries, WebhookDefSyncResponse, LibraryAgentDefinition, LibraryEntry, LibraryListResponse, LibraryMcpServerDefinition, LibrarySkillDefinition, LLMChatContent, LLMChatMessage, LLMChatOptions, LLMChatResponse, LLMChatStreamDelta, LLMChatStreamItem, LLMChatToolCall, LLMChatUsage, LLMTool, LLMEmbeddingItem, LLMEmbeddingsOptions, LLMEmbeddingsResponse, LLMEmbeddingsUsage, UsageDimension, UsageBucket, UsageAggregate, UsageReportResponse, CapabilityInertInfo, LimitInfo, FileCapturedInfo, TranscriptionInfo, AdmissionInfo, ErrorInfo, CallFormat, CallRef, HookDecisionInfo, TeamDecisionInfo, OverrideInfo, FallbackInfo, InterruptionEventInfo, SpawnChildInfo, TurnCancelledInfo, ChannelEventInfo, ContextCompactionInfo, ContextRecapInfo, ContextStateInfo, ContextExhaustedInfo, ContextTierVerdict, ContextDistillDeclinedInfo, EffectiveConfigResponse, RunFileInput, RunFileImportSource, RunFileImportHeading, RunInboxDownload, EffectiveConfigSource, EffectiveValue, InertContextSetting, AgentOverridableResponse, OverridableField, OverrideLockReason, RetuneRunResponse, RunConfigRecord, SubagentHoldInfo, RunConfigResponse, RunPromptResponse, RunPromptSnapshot, PromptBlock, TokenLimit, TokenLimitsResponse, ConfigResponse, ConfigProvider, ConfigModel, ConfigSearch, ConfigInstance, SetTokenLimitRequest, CredentialScope, CredentialMeta, CredentialListResponse, CryptoKeyVersion, CryptoKeyExport, CryptoKeyType, CryptoKeyEncoding, CryptoKeyBackend, CryptoKeyExternal, CryptoOp, CryptoRequestBase, CryptoKeyRef, CryptoEncryptRequest, CryptoDecryptRequest, CryptoRewrapRequest, CryptoMacRequest, CryptoVerifyMacRequest, CryptoSignRequest, CryptoVerifyRequest, CryptoPublicKeyRequest, CryptoTotpRequest, CryptoHashRequest, CryptoRandomRequest, CryptoHashPasswordRequest, CryptoVerifyPasswordRequest, CryptoRequest, CryptoEncryptResult, CryptoDecryptResult, CryptoRewrapResult, CryptoMacResult, CryptoVerifyResult, CryptoSignResult, CryptoPublicJWK, CryptoPublicKeyResult, CryptoTotpResult, CryptoHashResult, CryptoRandomResult, CryptoHashPasswordResult, CryptoVerifyPasswordResult, CryptoResultByOp, CryptoResult, DecideRequest, DecideResponse, DecisionQuestion, DecisionChoiceQuestion, DecisionNoulQuestion, DecisionScoreQuestion, DecisionAnswer, DecisionChoiceAnswer, DecisionNoulAnswer, DecisionScoreAnswer, DecisionModel, DecisionModelsResponse, SpeechFormat, SpeechOptions, SpeechResult, TranscribeOptions, TranscribeResult, TranscriptSegment, SpeechInfo, SpeechDirectionInfo, SpeechLanguageInfo, TranscriptionDirectionInfo, AclMode, AclSource, AclEffect, AclSubjectKind, AclCondition, AclRule, AclList, AclDefRow, AclDefListResponse, AclDefNamesResponse, AclValidateResult, AclCatalogType, AclCatalog, AclDefOp, AclDefInput, GroupRow, GroupMemberKind, GroupMember, GroupListResponse, GroupDetail, GroupEffective, CreateGroupBody, GroupMemberChange, AclBinding, AclBindingListResponse, CreateAclBindingBody, AuthzReason, AuthzDecision, AuthzRequestContext, AuthzEvaluationRequest, AuthzEvaluationsSemantic, AuthzEvaluationsRequest, AuthzEvaluationsResponse, AuthzTrace, AuthzAppliedBinding, AuthzExplainResponse, AuthzDraftBinding, AuthzSimulateRequest, AuthzSimulateResult, AuthzSimulateResponse, OAuthServerMetadata, OAuthTokenResponse, OAuthTokenSet, BeginAuthorizeOptions, AuthorizeRequest, AuthorizeResponseParams, CompleteAuthorizeOptions, RefreshTokenOptions, RevokeTokenOptions, TokenProvider, AppClientRefresh, AppClientConfig, ClientDefInput, ClientDefRecord, ConfigAppClient, AppClientEntry, ClientDefListResponse, ClientDefVerifyResult, ClientDefRetireResult, ClientDefNameSummary, ClientDefNamesResponse, DeviceLoginPrompt, DeviceLoginOptions, DeviceLoginResult, } from "./types.js";
 export { AclDeniedError, AgentIDInUseError, AgentNotFoundError, AlreadyPausingError, AuthError, BackpressureError, DeviceLoginError, NotFoundError, InsufficientScopeError, InvalidArgumentError, ChannelCursorRegressionError, LoomcycleError, NotPausedError, OAuthError, PauseNotConfiguredError, PerUserQuotaExhaustedError, PromptTooLargeError, RateLimitedError, RequestTooLargeError, SessionBusyError, SessionNotFoundError, SnapshotNotFoundError, SnapshotTooLargeError, SnapshotVersionError, SubstrateToolRefusedError, TokenLimitExceededError, UnavailableError, } from "./errors.js";

package/dist/types.d.ts

@@ -2416,6 +2416,9 @@ export interface UserSummary {
     access_mode?: string;
     /** The registered user's status; `""` for a subject seen only in runs. */
     status?: UserStatus | (string & {});
+    /** Set when a subject erasure made the user a tombstone (RFC3339): it can
+     *  never be re-enabled, changed or deleted (409 `user_erased`). */
+    erased_at?: string;
 }
 export interface ListUsersResponse {
     users: UserSummary[];
@@ -2444,6 +2447,8 @@ export interface UserRecord {
     status: UserStatus | (string & {});
     created_at: string;
     created_by: string;
+    /** Set when a subject erasure made the user a tombstone (RFC3339). */
+    erased_at?: string;
 }
 export interface CreateUserBody {
     subject: string;
@@ -2487,6 +2492,10 @@ export interface MintedUserToken {
  *  run endpoints. */
 export interface TokenRestrictions {
     agents?: string[];
+    /** Off-run calls a token with `agents` may still make (a chat bridge's
+     *  voice messages): `"speech"` opens `speech()`, `"transcription"` opens
+     *  `transcribe()`; either opens `speechInfo()`. */
+    surfaces?: TokenSurface[];
     /** `"none"` when every per-run override is locked, or the overrides the
      *  token may still set. Absent when nothing is locked. */
     overrides?: "none" | string[];
@@ -2498,6 +2507,8 @@ export interface TokenRestrictions {
     /** `"none"` when every run the token starts is a no-trace run. */
     retention?: "none";
 }
+/** An off-run surface an agent-limited token may be granted. */
+export type TokenSurface = "speech" | "transcription";
 /** One token row from GET /v1/_users/{subject}/tokens — METADATA ONLY (never
  *  the plaintext or hash). `active` applies the auth-layer validity rule. */
 export interface UserTokenMeta {
@@ -2741,7 +2752,13 @@ export interface WhoamiEffective {
     /** Per Web UI page id (`run`, `runs`, `channels`, `access`, `users`, …),
      *  whether a request to the route the page loads first would pass the
      *  route's scope gate and its access-control question in the current mode.
-     *  Every page is true in open mode. */
+     *  Every page is true in open mode.
+     *
+     *  Two ids are calls rather than pages: `speech` ({@link LoomcycleClient.speech})
+     *  and `transcription` ({@link LoomcycleClient.transcribe}). Each is true
+     *  only when its engine is configured AND the caller would pass its route's
+     *  gates, so a client shows a read-aloud button or a microphone without
+     *  probing. */
     surfaces: Record<string, boolean>;
 }
 export type RuntimeStateStatus = "running" | "pausing" | "paused";
@@ -7030,6 +7047,121 @@ export interface DecisionModelsResponse {
     default: string;
     models: DecisionModel[];
 }
+/** The audio a read-aloud can come back as: `wav` always, `mp3` or `ogg` only
+ *  when the deployment encodes to it (see {@link SpeechDirectionInfo.formats}).
+ *  `opus` is accepted as another name for `ogg`. */
+export type SpeechFormat = "wav" | "mp3" | "ogg" | "opus";
+/** Options of {@link LoomcycleClient.speech}. */
+export interface SpeechOptions {
+    /** A language code from {@link SpeechDirectionInfo.languages}; omitted, the
+     *  text's script decides, else the deployment's default language. */
+    language?: string;
+    /** One of the language's voices; omitted, its default voice. */
+    voice?: string;
+    /** Omitted, the deployment's own format. Ask only for one listed in
+     *  {@link SpeechDirectionInfo.formats}: another is a 400
+     *  `speech_encode_not_configured`. */
+    format?: SpeechFormat;
+    /** A `models:` alias of kind `synthesizer`; omitted, the language's. */
+    model?: string;
+    signal?: AbortSignal;
+}
+/** What {@link LoomcycleClient.speech} returns: the audio, whole. The
+ *  metadata comes from `X-Loomcycle-Speech-*` headers, which a browser can
+ *  read only on a same-origin request (the runtime exposes no CORS headers);
+ *  cross-origin they are undefined. */
+export interface SpeechResult {
+    bytes: Uint8Array;
+    /** `audio/wav`, `audio/mpeg` or `audio/ogg`. */
+    contentType: string;
+    language?: string;
+    voice?: string;
+    /** How long the audio lasts, pauses included. */
+    seconds?: number;
+    /** The characters spoken, after Markdown was normalised. */
+    chars?: number;
+    /** How many engine requests the text was cut into. */
+    pieces?: number;
+}
+/** Options of {@link LoomcycleClient.transcribe}. */
+export interface TranscribeOptions {
+    /** The file name sent with the upload; defaults to `recording` with an
+     *  extension from the content type. */
+    filename?: string;
+    /** The recording's type, e.g. `audio/wav` or `audio/webm`; a Blob's own
+     *  type is used when omitted. */
+    contentType?: string;
+    /** An ISO 639-1 code; omitted, the deployment's default or the engine's
+     *  guess. */
+    language?: string;
+    /** Names and terms the engine should expect. */
+    prompt?: string;
+    /** Must be the deployment's speech-to-text model when given; omit it. */
+    model?: string;
+    /** `json` (the default) or `verbose_json`, which adds `segments`. */
+    format?: "json" | "verbose_json";
+    signal?: AbortSignal;
+}
+/** One stretch of a transcript, in seconds from the start of the recording. */
+export interface TranscriptSegment {
+    start: number;
+    end: number;
+    text: string;
+    /** Present when the engine labels speakers. */
+    speaker?: string;
+}
+/** What {@link LoomcycleClient.transcribe} returns. */
+export interface TranscribeResult {
+    text: string;
+    language?: string;
+    /** The recording's length in seconds. */
+    duration?: number;
+    /** With `format: "verbose_json"`. */
+    segments?: TranscriptSegment[];
+}
+/** One language a read-aloud can speak, and the voices a caller may pick. */
+export interface SpeechLanguageInfo {
+    voices: string[];
+    default_voice: string;
+}
+/** The read-aloud half of {@link SpeechInfo}. Only `enabled` is present when
+ *  it is off, or when the caller may not use it. */
+export interface SpeechDirectionInfo {
+    enabled: boolean;
+    default_language?: string;
+    languages?: Record<string, SpeechLanguageInfo>;
+    /** `wav`, plus the one encoded format the deployment makes, if any. */
+    formats?: string[];
+    /** The size of each piece sent to the engine (information only). */
+    max_chars_per_call?: number;
+    /** What one request may carry; longer text is a 413 `input_too_long`. */
+    max_chars_per_request?: number;
+    max_seconds_per_request?: number;
+}
+/** The transcription half of {@link SpeechInfo}. Only `enabled` is present
+ *  when it is off, or when the caller may not use it. */
+export interface TranscriptionDirectionInfo {
+    enabled: boolean;
+    default_language?: string;
+    /** The recording types the route takes. `audio/wav` always; with a
+     *  converter, `audio/*` and `video/*`; without one, the types the engine
+     *  reads whole. */
+    accepts?: string[];
+    /** The upload limit (LOOMCYCLE_MAX_REQUEST_BYTES). */
+    max_upload_bytes?: number;
+    max_seconds_per_call?: number;
+    /** How long one upload may last; longer is a 413 `audio_too_long`. */
+    max_seconds_per_request?: number;
+    /** Without a converter: the largest non-WAV file the engine takes whole. A
+     *  PCM WAV is cut instead, so it is not bound by this. */
+    whole_file_max_bytes?: number;
+}
+/** GET /v1/_speech: what a client needs to draw its voice controls. Names no
+ *  engine URL, provider or model. */
+export interface SpeechInfo {
+    speech: SpeechDirectionInfo;
+    transcription: TranscriptionDirectionInfo;
+}
 /** One name's version roll-up from a `list*DefNames` method. Shared by every
  *  def family except OperatorTokenDef ({@link OperatorTokenDefNameSummary}). */
 export interface DefNameSummary {
@@ -7152,8 +7284,8 @@ export interface SchedulePauseResult {
 export interface ModelAlias {
     provider: string;
     model: string;
-    /** chat | decision | embedder. Absent for an untagged alias, which is a chat
-     *  model. */
+    /** chat | decision | embedder | transcriber | synthesizer. Absent for an
+     *  untagged alias, which is a chat model. */
     kind?: string;
 }
 /** Response of {@link LoomcycleClient.listModels}: alias name → target. */
@@ -7706,6 +7838,8 @@ export interface AuthzRequestContext {
     user_tier?: string;
     agent?: string;
     depth?: number;
+    /** The client address a `source_cidr` condition reads, e.g. `203.0.113.7`. */
+    source_ip?: string;
 }
 /** One question: may `subject` do `action` to `resource`? An omitted
  *  `subject` (or one naming the caller) asks about the caller; naming another

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@loomcycle/client",
-  "version": "1.139.0",
+  "version": "1.141.0",
   "description": "TypeScript client for the loomcycle sidecar (HTTP+SSE). 71 methods covering run streaming, agent metadata, pause/resume/state, resolver re-probe (resolveProbe — issue #88 operator escape hatch), operator-token admin (operatorTokenDef — RFC L OSS multi-tenant auth) + whoami (RFC L authoritative principal, v0.17.0) + tenant-scoped listUsers / listUserAgents, dynamic MCP registration (mcpServerDef + v0.18.0 typed mcpServerDefVerify + ensureMcpServer idempotent register-if-changed), snapshot lifecycle, memory admin (incl. v0.9.0 Vector Memory embed_stats + reembed AND v0.11.5 setMemoryEntry + deleteMemoryEntry), interruption resolve, hook management, v0.8.22 substrate admin (agentDef + skillDef), v0.9.x n8n Phase 0 (listChannels + streamUserRunStates), v0.9.x Channel CRUD (publishChannel + subscribeChannel + peekChannel + ackChannel), v0.11.5 Channel admin CRUD (createChannel + updateChannel + deleteChannel — runtime substrate; yaml channels refuse mutation with HTTP 409), v0.9.x content_sha256 verify, v0.9.1 transcript first-cycle, v0.9.x dynamic MCP server registration (mcpServerDef + MCPServerDefVerifyResult), v0.10.3 Library v2 enumeration (listLibraryAgents + listLibrarySkills + listLibraryMcpServers), and v0.11.0 LLM Gateway (llmChat + llmStream — direct provider routing without agent overhead; primary target is n8n's LoomCycleChatModel AI Agent sub-node and any LangChain-compatible consumer). v0.10.1 — dual ESM + CommonJS distribution (additive — ESM consumers unchanged; CJS consumers like n8n's community-node loader now work). v0.19.0 — typed inline code-js agent ingestion (AgentDefOverlay.code_body + ensureCodeAgent — register a deterministic code agent through the substrate with no host filesystem bind; RFC J). v0.20.0 — ensureMcpServer surfaces discoveredToolCount straight from create (loomcycle now auto-discovers MCP tools at ingestion); rediscover is now an explicit force-refresh. v0.21.0 — run/continue accept optional non-secret `metadata` (repo name, review policy, …) passed to the agent, symmetric with the WebHook/Schedule trigger paths. v0.22.0 — version-aligned lockstep release; no client-surface change (RFC N tenant isolation of the agent/skill/MCP/Schedule/Webhook definition plane + real op-schemas on the builtin MCP meta-tools are both server-side). v0.23.0 — version-aligned lockstep release; no client-surface change (RFCs O/P/R MCP-server hardening + thin client and the RFC Q DeepSeek tool-content fix are all server-side / MCP-transport-side). v0.24.0 — purgeChannel clears a channel's buffered messages without deleting its definition (allowed on yaml-declared channels too, unlike deleteChannel — F20); AgentDefOverlay gains channels / evaluation_scopes / interruption so a COMPLETE interactive/multi-agent agent round-trips over the substrate (F14). (Consolidates the interim 0.24/0.25 package bumps, which were never tag-published, back into lockstep with the loomcycle v0.24.0 tag.) v0.25.0 — adds the RFC S channel fan-in/fan-out client twins: awaitChannels() (wait for any/all/at_least N messages across channels, or a timeout — non-committing) + broadcastChannels() (publish one payload to N channels in one atomic-pre-flight call), the client-facing counterparts of the in-band Channel.await / Channel.broadcast tool ops (the rest of v0.25.0 — the manual-management Web UI console + Context op=time + max_fires self-retiring schedules — is server-side / in-band, no client-surface change). v0.29.1 — Usage gains optional `max_context_tokens` (the serving model's context-window ceiling, stamped by the loop from Provider.Capabilities() on each usage event) so a consumer can render a 'context used / max' gauge without a hard-coded per-model table; additive + optional, no behavior change. (Lockstep catch-up: the field landed in loomcycle v0.29.0 but the adapter publish skipped on a version mismatch; v0.29.1 realigns the package version with the release tag so it publishes.) v0.33.0 — gRPC + TS client parity for the RFC Y external fan-out and the compaction surface: spawnRunBatch() (POST /v1/runs:batch — spawn up to 32 fresh runs concurrently in one call, combined index-aligned envelope, per-child failures in-envelope) + compactRun(runId) (POST /v1/runs/{run_id}/compact — summarize a parked run's context); plus per-run sampling + compaction overrides now accepted on runStreaming / continueSession (an explicit temperature 0 is preserved as deterministic, not dropped as falsy). (The gRPC half adds the matching SpawnRunBatch / CompactRun RPCs + the sampling/compaction fields on RunRequest/ContinueRequest — server-side.) v0.34.0 — version-aligned lockstep release; no client-surface change (context-transform plugins / RFC Z Phase 1a are server-side config; the exp7 hardening pass is server-side; the R2 cross-provider thinking-model downgrade surfaces a new `model_downgraded` SSE event the generic stream passes through unchanged). v0.35.0 — RFC AH dynamic filesystem volumes: volumeDef() (POST /v1/_volumedef — op-discriminated create/get/list/delete/purge; a Volume is flat, so delete unmaps + purge RemoveAll's, no retire/promote/fork) + listVolumes() / listEphemeralVolumes() (GET /v1/_volumes[/ephemeral] — tenant-scoped; host paths redacted for non-operator callers). Tenant-confined; the runtime derives the path inside an operator-blessed dynamic_root, so callers pass name + mode, never a host path. v1.1.1 — RFC AI interactive agentic sessions: an `interactive: true` flag on runStreaming/continueSession (a run that parks at end_turn for steering) + sendRunInput(runId, text) (POST /v1/runs/{id}/input — steer a live run) + streamRunByID(runId, {fromSeq}) (GET /v1/runs/{id}/stream — re-attach by run_id; the operator's prior turns replay as `steer` events so a cold client reconstructs the whole conversation) + a high-level InteractiveSession driver (client.interactiveSession / attachInteractiveSession — events()/send()/cancel(), the adapter port of the Web UI run terminal). The AgentEvent union gains awaiting_input/steer/context_compaction frames. Version-aligned with the loomcycle v1.1.x line so the v1.1.1 tag publishes it (also carrying the previously-unpublished v0.35.0 volume surface). v1.4.0 — RFC AL Path VFS + RFC AK Document on the wire: path(input) (POST /v1/_path — a Unix-like filesystem over Memory/Volumes/Documents; resolve/ls/stat/mkdir/mv/rm) + document(input) (POST /v1/_document — chunked-graph documents; 13 ops, needs SQL Memory on the sidecar). Scope (agent/user/tenant) + tenant are resolved server-side from the authenticated principal, never the wire; an off-run scope:'user' op keys on the principal subject so it interoperates with that user's agent runs. New PathToolInput / DocumentToolInput types; responses are op-varying (unknown — narrow as needed). v1.7.0 — RFC AT image/vision input: the PromptContent union gains an `image` variant ({ type:'image'; media_type: ImageMediaType; data: base64-no-prefix }) accepted in a user segment by runStreaming / continueSession (segments pass through unchanged — no method change). The model must be vision-capable or the run errors before the call. New ImageMediaType type (image/png|jpeg|gif|webp). v1.12.1 — Path/Document browse-by-subject + the full Document op set (RFC AS/AK): path(input, opts) / document(input, opts) accept optional scopeId / tenant browse overrides sent as ?scope_id= / ?tenant= query params (server reads them from the URL, re-checks authorization; omit both to browse your own subject — byte-identical to the pre-RFC-AS request), and DocumentToolInput.op now covers all 16 backend ops (adds set_path, export_md, import_md) with the matching include_metadata / markdown fields. Additive — existing path() / document() callers are unchanged. v1.16.0 — RFC BC client-executed tools (local tool host): connectClientTools({tools,onInvoke}) opens a persistent WebSocket to /v1/client-tools, registers tools this client runs on the user's machine (browser DOM / files / shell), and answers the invoke frames loomcycle routes when an agent of your principal calls a client:<tool> — returning your reply as an ordinary tool result. Returns a ClientToolHost driver (.close() to stop; auto-reconnect); dependency-free (global WebSocket in browsers / Node 22+, or an injected WebSocketImpl like the ws package on older Node); the bearer rides the Sec-WebSocket-Protocol subprotocol (browsers can't set an Authorization header on a WebSocket). The runtime's first WebSocket surface. v1.45.0 — RFC BL P5 subject erasure: erasureReport(subject, {tenant}) (GET /v1/_erasure — what this deployment holds about one subject, in three tiers: deletable with existing primitives, subject-keyed but not deletable, and facts ABOUT the subject in scopes they do not own) + erasureExecute(subject, {dryRun, confirm, tenant}) (POST /v1/_erasure — removes tiers 1 and 2). DEFAULTS TO A DRY RUN: dryRun defaults true and a live run also requires confirm === subject. The tier-3 residue is traceable only through the subject's chats, which a live run deletes, so a report afterwards shows rows: 0 while those facts remain — the returned object is the only durable record of what was not reached; persist it. New ErasureReport / ErasureResult / ErasureTier / ErasureResidue types. v1.47.0 — RFC BV memory-view SDK: memorySearch() (POST /v1/_memory/search — off-run unified semantic search spanning k/v entries AND document-chunk bodies in one ranked list, each hit tagged kind memory|document with chunk_id on document hits) + memoryEmbedStats(scope) + reembedMemory(scope, scopeId, {dryRun,limit}) (the Vector Memory embed-admin reads the memory-view console needs; dry_run defaults true). Fact reads (list_facts + get_chunk's entity block) ride the existing document() passthrough. Additive — existing callers unchanged. v1.61.0 — RFC CJ per-run context-window override: runStreaming / continueSession accept an optional maxContextTokens (integer tokens) that wins over the agent's own max_context_tokens; omitted inherits it (which defers to the provider/driver default). Distinct from a model's output cap; primarily for local inference (Ollama num_ctx). Serialized as the snake_case max_context_tokens field on POST /v1/runs + the continuation body; additive, existing callers unchanged. v1.72.1 — the TeamDef version lifecycle: listTeamVersions(name) (op=list — every version of one team, newest first), promoteTeam(defId) (op=promote — point the active pointer, which is what a run BY NAME executes; forkTeam defaults to promote:false, so authoring and putting in force stay two steps), retireTeam(defId, retired) (op=retire — reversible and version-scoped, unlike deleteTeam) and verifyTeam(name, contentSha256) (op=verify — the drift check for a workflow kept in source control and pushed to several deployments; an absent team answers deployed:false rather than raising). The ops existed on the substrate and over HTTP; a client that could author a team could not put one in force.",
   "license": "Apache-2.0",
   "type": "module",