iobroker.assistant 1.2.0 → 1.3.0

package/build/lib/piAgent.js

@@ -0,0 +1,89 @@
+"use strict";
+/**
+ * The adapter `pi-agent` as the cloud tier ("harness"): instead of running its own tool loop, the assistant
+ * hands the question to a pi-agent instance via `sendTo(<instance>, 'ask', …)`. pi-agent stores every model
+ * and tool call as a checkpoint before it runs (an interrupted request continues after a restart, a write is
+ * never repeated), enforces its own write permissions and confirmations, and can use the system-wide AI
+ * settings of admin — none of which the built-in `LlmAgent` does.
+ *
+ * Each request source (satellite, `chat`, `telegram:<user>`) gets a conversation of its own in pi-agent
+ * (`conversation`), so two rooms do not see each other's follow-ups. The source-independent part of the
+ * prompt (style, follow-up marker, memory, device list) goes along as the `instructions` of that
+ * conversation; what changes every turn (time, weather, presence) is prepended to the question.
+ * pi-agent 0.3.0 and older ignore both fields and answer everything in their root conversation.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.PiAgentHarness = exports.PI_AGENT_TIMEOUT_SEC = exports.PI_AGENT_MAX_INSTRUCTIONS = void 0;
+exports.composeInstructions = composeInstructions;
+exports.conversationOf = conversationOf;
+/** Longest instructions pi-agent accepts for a conversation; anything beyond is cut there. */
+exports.PI_AGENT_MAX_INSTRUCTIONS = 20000;
+/** Default for `piAgentTimeoutSec`: a tool loop with several reads can take a while. */
+exports.PI_AGENT_TIMEOUT_SEC = 60;
+/**
+ * Join instruction blocks, leaving out optional ones from the end until the result fits into pi-agent's
+ * limit — pi-agent would cut the text in the middle, and a half device list is worse than none.
+ *
+ * @param required blocks that always go along (style, follow-up marker)
+ * @param optional blocks in order of importance (memory before the device list)
+ */
+function composeInstructions(required, optional) {
+    const blocks = [...required.filter(Boolean)];
+    const extra = optional.filter(Boolean);
+    let dropped = 0;
+    while (extra.length) {
+        const text = [...blocks, ...extra].join('\n\n');
+        if (text.length <= exports.PI_AGENT_MAX_INSTRUCTIONS) {
+            return { text, dropped };
+        }
+        extra.pop();
+        dropped++;
+    }
+    return { text: blocks.join('\n\n').slice(0, exports.PI_AGENT_MAX_INSTRUCTIONS), dropped };
+}
+/** Conversation name for a request source; pi-agent uses "/" as separator in its keys. */
+function conversationOf(source) {
+    return (source || 'default').replace(/\//g, '_').slice(0, 100);
+}
+class PiAgentHarness {
+    host;
+    instance;
+    timeoutMs;
+    seq = 0;
+    constructor(host, 
+    /** e.g. `pi-agent.0` */
+    instance, timeoutMs = exports.PI_AGENT_TIMEOUT_SEC * 1000) {
+        this.host = host;
+        this.instance = instance;
+        this.timeoutMs = timeoutMs;
+    }
+    /**
+     * Ask pi-agent. Fails fast when the instance is not running (otherwise the caller would wait for the whole
+     * timeout), and turns pi-agent's `{error}` into an exception so the caller can fall back to the local model.
+     */
+    async ask(req) {
+        const alive = await this.host.getForeignStateAsync(`system.adapter.${this.instance}.alive`);
+        if (!alive?.val) {
+            throw new Error(`${this.instance} is not running`);
+        }
+        // Unique per request: pi-agent answers a repeated requestId only once.
+        const requestId = `assistant-${Date.now().toString(36)}-${(++this.seq).toString(36)}`;
+        const result = (await this.host.sendToAsync(this.instance, 'ask', {
+            text: req.text,
+            conversation: conversationOf(req.conversation),
+            instructions: req.instructions,
+            requestId,
+        }, { timeout: this.timeoutMs }));
+        if (result?.error) {
+            const error = typeof result.error === 'string' ? result.error : JSON.stringify(result.error);
+            throw new Error(`${this.instance}: ${error}`);
+        }
+        const answer = typeof result?.answer === 'string' ? result.answer.trim() : '';
+        if (!answer) {
+            throw new Error(`${this.instance} gave no answer`);
+        }
+        return answer;
+    }
+}
+exports.PiAgentHarness = PiAgentHarness;
+//# sourceMappingURL=piAgent.js.map

package/build/lib/piAgent.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"piAgent.js","sourceRoot":"","sources":["../../src/lib/piAgent.ts"],"names":[],"mappings":";AAAA;;;;;;;;;;;;GAYG;;;AA8BH,kDAaC;AAGD,wCAEC;AA9CD,8FAA8F;AACjF,QAAA,yBAAyB,GAAG,KAAK,CAAC;AAE/C,wFAAwF;AAC3E,QAAA,oBAAoB,GAAG,EAAE,CAAC;AAiBvC;;;;;;GAMG;AACH,SAAgB,mBAAmB,CAAC,QAAkB,EAAE,QAAkB;IACtE,MAAM,MAAM,GAAG,CAAC,GAAG,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC;IAC7C,MAAM,KAAK,GAAG,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;IACvC,IAAI,OAAO,GAAG,CAAC,CAAC;IAChB,OAAO,KAAK,CAAC,MAAM,EAAE,CAAC;QAClB,MAAM,IAAI,GAAG,CAAC,GAAG,MAAM,EAAE,GAAG,KAAK,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAChD,IAAI,IAAI,CAAC,MAAM,IAAI,iCAAyB,EAAE,CAAC;YAC3C,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC;QAC7B,CAAC;QACD,KAAK,CAAC,GAAG,EAAE,CAAC;QACZ,OAAO,EAAE,CAAC;IACd,CAAC;IACD,OAAO,EAAE,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,iCAAyB,CAAC,EAAE,OAAO,EAAE,CAAC;AACtF,CAAC;AAED,0FAA0F;AAC1F,SAAgB,cAAc,CAAC,MAAc;IACzC,OAAO,CAAC,MAAM,IAAI,SAAS,CAAC,CAAC,OAAO,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;AACnE,CAAC;AAED,MAAa,cAAc;IAIF;IAED;IACC;IANb,GAAG,GAAG,CAAC,CAAC;IAEhB,YACqB,IAAiB;IAClC,wBAAwB;IACR,QAAgB,EACf,YAAoB,4BAAoB,GAAG,IAAI;QAH/C,SAAI,GAAJ,IAAI,CAAa;QAElB,aAAQ,GAAR,QAAQ,CAAQ;QACf,cAAS,GAAT,SAAS,CAAsC;IACjE,CAAC;IAEJ;;;OAGG;IACI,KAAK,CAAC,GAAG,CAAC,GAAmB;QAChC,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,oBAAoB,CAAC,kBAAkB,IAAI,CAAC,QAAQ,QAAQ,CAAC,CAAC;QAC5F,IAAI,CAAC,KAAK,EAAE,GAAG,EAAE,CAAC;YACd,MAAM,IAAI,KAAK,CAAC,GAAG,IAAI,CAAC,QAAQ,iBAAiB,CAAC,CAAC;QACvD,CAAC;QACD,uEAAuE;QACvE,MAAM,SAAS,GAAG,aAAa,IAAI,CAAC,GAAG,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,EAAE,CAAC,EAAE,CAAC;QACtF,MAAM,MAAM,GAAG,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,WAAW,CACvC,IAAI,CAAC,QAAQ,EACb,KAAK,EACL;YACI,IAAI,EAAE,GAAG,CAAC,IAAI;YACd,YAAY,EAAE,cAAc,CAAC,GAAG,CAAC,YAAY,CAAC;YAC9C,YAAY,EAAE,GAAG,CAAC,YAAY;YAC9B,SAAS;SACZ,EACD,EAAE,OAAO,EAAE,IAAI,CAAC,SAAS,EAAE,CAC9B,CAA6D,CAAC;QAC/D,IAAI,MAAM,EAAE,KAAK,EAAE,CAAC;YAChB,MAAM,KAAK,GAAG,OAAO,MAAM,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;YAC7F,MAAM,IAAI,KAAK,CAAC,GAAG,IAAI,CAAC,QAAQ,KAAK,KAAK,EAAE,CAAC,CAAC;QAClD,CAAC;QACD,MAAM,MAAM,GAAG,OAAO,MAAM,EAAE,MAAM,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAC9E,IAAI,CAAC,MAAM,EAAE,CAAC;YACV,MAAM,IAAI,KAAK,CAAC,GAAG,IAAI,CAAC,QAAQ,iBAAiB,CAAC,CAAC;QACvD,CAAC;QACD,OAAO,MAAM,CAAC;IAClB,CAAC;CACJ;AA1CD,wCA0CC","sourcesContent":["/**\n * The adapter `pi-agent` as the cloud tier (\"harness\"): instead of running its own tool loop, the assistant\n * hands the question to a pi-agent instance via `sendTo(<instance>, 'ask', …)`. pi-agent stores every model\n * and tool call as a checkpoint before it runs (an interrupted request continues after a restart, a write is\n * never repeated), enforces its own write permissions and confirmations, and can use the system-wide AI\n * settings of admin — none of which the built-in `LlmAgent` does.\n *\n * Each request source (satellite, `chat`, `telegram:<user>`) gets a conversation of its own in pi-agent\n * (`conversation`), so two rooms do not see each other's follow-ups. The source-independent part of the\n * prompt (style, follow-up marker, memory, device list) goes along as the `instructions` of that\n * conversation; what changes every turn (time, weather, presence) is prepended to the question.\n * pi-agent 0.3.0 and older ignore both fields and answer everything in their root conversation.\n */\n\n/** Longest instructions pi-agent accepts for a conversation; anything beyond is cut there. */\nexport const PI_AGENT_MAX_INSTRUCTIONS = 20000;\n\n/** Default for `piAgentTimeoutSec`: a tool loop with several reads can take a while. */\nexport const PI_AGENT_TIMEOUT_SEC = 60;\n\n/** The part of the adapter this module needs (kept small so tests can fake it). */\nexport interface PiAgentHost {\n    sendToAsync(instance: string, command: string, message: unknown, options?: { timeout?: number }): Promise<unknown>;\n    getForeignStateAsync(id: string): Promise<ioBroker.State | null | undefined>;\n}\n\nexport interface PiAgentRequest {\n    /** The text pi-agent answers: context lines + the question. */\n    text: string;\n    /** Conversation of the caller, e.g. the satellite name. */\n    conversation: string;\n    /** Instructions of that conversation. */\n    instructions: string;\n}\n\n/**\n * Join instruction blocks, leaving out optional ones from the end until the result fits into pi-agent's\n * limit — pi-agent would cut the text in the middle, and a half device list is worse than none.\n *\n * @param required blocks that always go along (style, follow-up marker)\n * @param optional blocks in order of importance (memory before the device list)\n */\nexport function composeInstructions(required: string[], optional: string[]): { text: string; dropped: number } {\n    const blocks = [...required.filter(Boolean)];\n    const extra = optional.filter(Boolean);\n    let dropped = 0;\n    while (extra.length) {\n        const text = [...blocks, ...extra].join('\\n\\n');\n        if (text.length <= PI_AGENT_MAX_INSTRUCTIONS) {\n            return { text, dropped };\n        }\n        extra.pop();\n        dropped++;\n    }\n    return { text: blocks.join('\\n\\n').slice(0, PI_AGENT_MAX_INSTRUCTIONS), dropped };\n}\n\n/** Conversation name for a request source; pi-agent uses \"/\" as separator in its keys. */\nexport function conversationOf(source: string): string {\n    return (source || 'default').replace(/\\//g, '_').slice(0, 100);\n}\n\nexport class PiAgentHarness {\n    private seq = 0;\n\n    public constructor(\n        private readonly host: PiAgentHost,\n        /** e.g. `pi-agent.0` */\n        public readonly instance: string,\n        private readonly timeoutMs: number = PI_AGENT_TIMEOUT_SEC * 1000,\n    ) {}\n\n    /**\n     * Ask pi-agent. Fails fast when the instance is not running (otherwise the caller would wait for the whole\n     * timeout), and turns pi-agent's `{error}` into an exception so the caller can fall back to the local model.\n     */\n    public async ask(req: PiAgentRequest): Promise<string> {\n        const alive = await this.host.getForeignStateAsync(`system.adapter.${this.instance}.alive`);\n        if (!alive?.val) {\n            throw new Error(`${this.instance} is not running`);\n        }\n        // Unique per request: pi-agent answers a repeated requestId only once.\n        const requestId = `assistant-${Date.now().toString(36)}-${(++this.seq).toString(36)}`;\n        const result = (await this.host.sendToAsync(\n            this.instance,\n            'ask',\n            {\n                text: req.text,\n                conversation: conversationOf(req.conversation),\n                instructions: req.instructions,\n                requestId,\n            },\n            { timeout: this.timeoutMs },\n        )) as { answer?: unknown; error?: unknown } | null | undefined;\n        if (result?.error) {\n            const error = typeof result.error === 'string' ? result.error : JSON.stringify(result.error);\n            throw new Error(`${this.instance}: ${error}`);\n        }\n        const answer = typeof result?.answer === 'string' ? result.answer.trim() : '';\n        if (!answer) {\n            throw new Error(`${this.instance} gave no answer`);\n        }\n        return answer;\n    }\n}\n"]}

package/build/lib/systemAi.js

@@ -0,0 +1,124 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.SYSTEM_AI_ID = void 0;
+exports.systemAiSelection = systemAiSelection;
+exports.systemAiFingerprint = systemAiFingerprint;
+exports.usesSystemAi = usesSystemAi;
+exports.readSystemAi = readSystemAi;
+exports.resolveSystemLlm = resolveSystemLlm;
+exports.resolveLlm = resolveLlm;
+const credentials_1 = require("./credentials");
+const llm_1 = require("./llm");
+/** ID of the object with the system-wide AI settings */
+exports.SYSTEM_AI_ID = 'system.ai';
+/** Fields of `system.ai` the assistant uses — a change of the others (like admin's chat button) needs no restart */
+const USED_FIELDS = ['provider', 'model', 'credentialId', 'baseUrl'];
+/** Where to set them up, for the log */
+const WHERE = 'admin → System settings → tab "AI"';
+function text(value) {
+    return typeof value === 'string' || typeof value === 'number' ? String(value).trim() : '';
+}
+/**
+ * Provider, model, credential and address of `system.ai`; `null` if it names no provider the assistant knows.
+ *
+ * @param native the `native` of `system.ai`
+ */
+function systemAiSelection(native) {
+    const provider = text(native?.provider);
+    if (!provider || !Object.prototype.hasOwnProperty.call(llm_1.PROVIDER_PRESETS, provider)) {
+        return null;
+    }
+    return {
+        provider: provider,
+        model: text(native?.model),
+        credentialId: text(native?.credentialId),
+        baseUrl: provider === 'custom' ? text(native?.baseUrl) : '',
+    };
+}
+/**
+ * What of `system.ai` the assistant depends on; a change of it restarts the instance.
+ *
+ * @param native the `native` of `system.ai`
+ */
+function systemAiFingerprint(native) {
+    return JSON.stringify(USED_FIELDS.map(field => text(native?.[field])));
+}
+/**
+ * Whether the instance takes the cloud LLM from `system.ai`: provider `system`, or a provider that needs a key
+ * without one (neither in the adapter nor a credential) — the custom endpoint often works without a key.
+ *
+ * @param cfg the configuration of the instance
+ */
+function usesSystemAi(cfg) {
+    const provider = cfg.provider || 'openai';
+    if (provider === 'system') {
+        return true;
+    }
+    if (provider === 'none' || provider === 'pi-agent' || provider === 'custom') {
+        return false;
+    }
+    return cfg.credentialType === 'manager' ? !text(cfg.credentialIdApiKey) : !text(cfg.apiKey);
+}
+/**
+ * Read `system.ai`; a missing object (or one the adapter may not read) gives `undefined`.
+ *
+ * @param adapter the adapter
+ */
+async function readSystemAi(adapter) {
+    try {
+        const obj = await adapter.getForeignObjectAsync(exports.SYSTEM_AI_ID);
+        return obj?.native;
+    }
+    catch {
+        return undefined;
+    }
+}
+/**
+ * The cloud LLM of `system.ai`, or why there is none.
+ *
+ * @param adapter the adapter (reads the credential)
+ * @param native the `native` of `system.ai`
+ * @param explicit the provider `system` was chosen (otherwise the instance only has no key of its own)
+ */
+async function resolveSystemLlm(adapter, native, explicit) {
+    const sel = systemAiSelection(native);
+    if (!sel) {
+        return {
+            error: explicit
+                ? `The system-wide AI settings have no provider — set one up in ${WHERE} or choose a provider here.`
+                : `No API key configured — enter one in the adapter settings or set up the system-wide AI settings in ${WHERE}.`,
+        };
+    }
+    if (sel.provider === 'custom' && !sel.baseUrl) {
+        return { error: `The system-wide AI settings use an OpenAI-compatible server without address (${WHERE}).` };
+    }
+    const apiKey = (0, llm_1.effectiveApiKey)(sel.provider, await (0, credentials_1.readCentralKey)(adapter, sel.credentialId));
+    if (!apiKey) {
+        return { error: `The system-wide AI settings have no API key for ${sel.provider} (${WHERE}).` };
+    }
+    const preset = (0, llm_1.resolveProvider)(sel.provider, sel.baseUrl);
+    return {
+        provider: sel.provider,
+        apiKey,
+        model: sel.model || preset.defaultModel,
+        baseUrl: preset.baseUrl,
+        system: true,
+    };
+}
+/**
+ * The cloud LLM of the instance: its own provider and key, or those of `system.ai` (see {@link usesSystemAi}).
+ * Not for the providers `none` and `pi-agent`, which have no LLM of their own.
+ *
+ * @param adapter the adapter
+ * @param cfg the configuration of the instance
+ */
+async function resolveLlm(adapter, cfg) {
+    if (usesSystemAi(cfg)) {
+        return resolveSystemLlm(adapter, await readSystemAi(adapter), cfg.provider === 'system');
+    }
+    const provider = (cfg.provider || 'openai');
+    const apiKey = (0, llm_1.effectiveApiKey)(provider, await (0, credentials_1.resolveApiKey)(adapter, cfg));
+    const preset = (0, llm_1.resolveProvider)(provider, cfg.baseUrl);
+    return { provider, apiKey, model: cfg.model || preset.defaultModel, baseUrl: preset.baseUrl, system: false };
+}
+//# sourceMappingURL=systemAi.js.map

package/build/lib/systemAi.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"systemAi.js","sourceRoot":"","sources":["../../src/lib/systemAi.ts"],"names":[],"mappings":";;;AAmDA,8CAWC;AAOD,kDAEC;AAQD,oCAWC;AAOD,oCAOC;AASD,4CA4BC;AASD,gCAWC;AAtJD,+CAA8D;AAC9D,+BAA0F;AAE1F,wDAAwD;AAC3C,QAAA,YAAY,GAAG,WAAW,CAAC;AAExC,oHAAoH;AACpH,MAAM,WAAW,GAAG,CAAC,UAAU,EAAE,OAAO,EAAE,cAAc,EAAE,SAAS,CAAU,CAAC;AAE9E,wCAAwC;AACxC,MAAM,KAAK,GAAG,oCAAoC,CAAC;AAqBnD,SAAS,IAAI,CAAC,KAAc;IACxB,OAAO,OAAO,KAAK,KAAK,QAAQ,IAAI,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;AAC9F,CAAC;AAED;;;;GAIG;AACH,SAAgB,iBAAiB,CAAC,MAAkD;IAChF,MAAM,QAAQ,GAAG,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IACxC,IAAI,CAAC,QAAQ,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,cAAc,CAAC,IAAI,CAAC,sBAAgB,EAAE,QAAQ,CAAC,EAAE,CAAC;QACjF,OAAO,IAAI,CAAC;IAChB,CAAC;IACD,OAAO;QACH,QAAQ,EAAE,QAAoB;QAC9B,KAAK,EAAE,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC;QAC1B,YAAY,EAAE,IAAI,CAAC,MAAM,EAAE,YAAY,CAAC;QACxC,OAAO,EAAE,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,EAAE;KAC9D,CAAC;AACN,CAAC;AAED;;;;GAIG;AACH,SAAgB,mBAAmB,CAAC,MAAkD;IAClF,OAAO,IAAI,CAAC,SAAS,CAAC,WAAW,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;AAC3E,CAAC;AAED;;;;;GAKG;AACH,SAAgB,YAAY,CACxB,GAAyF;IAEzF,MAAM,QAAQ,GAAG,GAAG,CAAC,QAAQ,IAAI,QAAQ,CAAC;IAC1C,IAAI,QAAQ,KAAK,QAAQ,EAAE,CAAC;QACxB,OAAO,IAAI,CAAC;IAChB,CAAC;IACD,IAAI,QAAQ,KAAK,MAAM,IAAI,QAAQ,KAAK,UAAU,IAAI,QAAQ,KAAK,QAAQ,EAAE,CAAC;QAC1E,OAAO,KAAK,CAAC;IACjB,CAAC;IACD,OAAO,GAAG,CAAC,cAAc,KAAK,SAAS,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,kBAAkB,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;AAChG,CAAC;AAED;;;;GAIG;AACI,KAAK,UAAU,YAAY,CAAC,OAAyB;IACxD,IAAI,CAAC;QACD,MAAM,GAAG,GAAG,MAAM,OAAO,CAAC,qBAAqB,CAAC,oBAAY,CAAC,CAAC;QAC9D,OAAO,GAAG,EAAE,MAAM,CAAC;IACvB,CAAC;IAAC,MAAM,CAAC;QACL,OAAO,SAAS,CAAC;IACrB,CAAC;AACL,CAAC;AAED;;;;;;GAMG;AACI,KAAK,UAAU,gBAAgB,CAClC,OAAyB,EACzB,MAAkD,EAClD,QAAiB;IAEjB,MAAM,GAAG,GAAG,iBAAiB,CAAC,MAAM,CAAC,CAAC;IACtC,IAAI,CAAC,GAAG,EAAE,CAAC;QACP,OAAO;YACH,KAAK,EAAE,QAAQ;gBACX,CAAC,CAAC,gEAAgE,KAAK,6BAA6B;gBACpG,CAAC,CAAC,sGAAsG,KAAK,GAAG;SACvH,CAAC;IACN,CAAC;IACD,IAAI,GAAG,CAAC,QAAQ,KAAK,QAAQ,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,CAAC;QAC5C,OAAO,EAAE,KAAK,EAAE,gFAAgF,KAAK,IAAI,EAAE,CAAC;IAChH,CAAC;IACD,MAAM,MAAM,GAAG,IAAA,qBAAe,EAAC,GAAG,CAAC,QAAQ,EAAE,MAAM,IAAA,4BAAc,EAAC,OAAO,EAAE,GAAG,CAAC,YAAY,CAAC,CAAC,CAAC;IAC9F,IAAI,CAAC,MAAM,EAAE,CAAC;QACV,OAAO,EAAE,KAAK,EAAE,mDAAmD,GAAG,CAAC,QAAQ,KAAK,KAAK,IAAI,EAAE,CAAC;IACpG,CAAC;IACD,MAAM,MAAM,GAAG,IAAA,qBAAe,EAAC,GAAG,CAAC,QAAQ,EAAE,GAAG,CAAC,OAAO,CAAC,CAAC;IAC1D,OAAO;QACH,QAAQ,EAAE,GAAG,CAAC,QAAQ;QACtB,MAAM;QACN,KAAK,EAAE,GAAG,CAAC,KAAK,IAAI,MAAM,CAAC,YAAY;QACvC,OAAO,EAAE,MAAM,CAAC,OAAO;QACvB,MAAM,EAAE,IAAI;KACf,CAAC;AACN,CAAC;AAED;;;;;;GAMG;AACI,KAAK,UAAU,UAAU,CAC5B,OAAyB,EACzB,GAAkB;IAElB,IAAI,YAAY,CAAC,GAAG,CAAC,EAAE,CAAC;QACpB,OAAO,gBAAgB,CAAC,OAAO,EAAE,MAAM,YAAY,CAAC,OAAO,CAAC,EAAE,GAAG,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC;IAC7F,CAAC;IACD,MAAM,QAAQ,GAAG,CAAC,GAAG,CAAC,QAAQ,IAAI,QAAQ,CAAa,CAAC;IACxD,MAAM,MAAM,GAAG,IAAA,qBAAe,EAAC,QAAQ,EAAE,MAAM,IAAA,2BAAa,EAAC,OAAO,EAAE,GAAG,CAAC,CAAC,CAAC;IAC5E,MAAM,MAAM,GAAG,IAAA,qBAAe,EAAC,QAAQ,EAAE,GAAG,CAAC,OAAO,CAAC,CAAC;IACtD,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,KAAK,EAAE,GAAG,CAAC,KAAK,IAAI,MAAM,CAAC,YAAY,EAAE,OAAO,EAAE,MAAM,CAAC,OAAO,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC;AACjH,CAAC","sourcesContent":["/**\n * The system-wide AI settings of admin (`system.ai`, System settings → tab \"AI\") as the source of the\n * cloud LLM: provider, model, key — always an entry of the central credential store — and the address of\n * a custom endpoint.\n *\n * Used with the provider `system`, and by an instance that has no key of its own: an assistant nobody\n * configured takes what was set up once for the whole system (the same rule as `@iobroker/ai-core`'s\n * `usesSystemAiSettings`). The maximum answer length stays the assistant's own (`maxTokens`), because a\n * spoken answer needs far fewer tokens than the admin chat.\n */\nimport type { AdapterConfig } from '../types';\nimport { readCentralKey, resolveApiKey } from './credentials';\nimport { PROVIDER_PRESETS, effectiveApiKey, resolveProvider, type Provider } from './llm';\n\n/** ID of the object with the system-wide AI settings */\nexport const SYSTEM_AI_ID = 'system.ai';\n\n/** Fields of `system.ai` the assistant uses — a change of the others (like admin's chat button) needs no restart */\nconst USED_FIELDS = ['provider', 'model', 'credentialId', 'baseUrl'] as const;\n\n/** Where to set them up, for the log */\nconst WHERE = 'admin → System settings → tab \"AI\"';\n\n/** Provider, model, credential and address out of `system.ai` */\nexport interface SystemAiSelection {\n    provider: Provider;\n    model: string;\n    credentialId: string;\n    /** Only for `custom` */\n    baseUrl: string;\n}\n\n/** The cloud LLM the instance works with, whether configured here or in `system.ai` */\nexport interface ResolvedLlm {\n    provider: Provider;\n    apiKey: string;\n    model: string;\n    baseUrl: string;\n    /** Taken from `system.ai` */\n    system: boolean;\n}\n\nfunction text(value: unknown): string {\n    return typeof value === 'string' || typeof value === 'number' ? String(value).trim() : '';\n}\n\n/**\n * Provider, model, credential and address of `system.ai`; `null` if it names no provider the assistant knows.\n *\n * @param native the `native` of `system.ai`\n */\nexport function systemAiSelection(native: Record<string, unknown> | null | undefined): SystemAiSelection | null {\n    const provider = text(native?.provider);\n    if (!provider || !Object.prototype.hasOwnProperty.call(PROVIDER_PRESETS, provider)) {\n        return null;\n    }\n    return {\n        provider: provider as Provider,\n        model: text(native?.model),\n        credentialId: text(native?.credentialId),\n        baseUrl: provider === 'custom' ? text(native?.baseUrl) : '',\n    };\n}\n\n/**\n * What of `system.ai` the assistant depends on; a change of it restarts the instance.\n *\n * @param native the `native` of `system.ai`\n */\nexport function systemAiFingerprint(native: Record<string, unknown> | null | undefined): string {\n    return JSON.stringify(USED_FIELDS.map(field => text(native?.[field])));\n}\n\n/**\n * Whether the instance takes the cloud LLM from `system.ai`: provider `system`, or a provider that needs a key\n * without one (neither in the adapter nor a credential) — the custom endpoint often works without a key.\n *\n * @param cfg the configuration of the instance\n */\nexport function usesSystemAi(\n    cfg: Pick<AdapterConfig, 'provider' | 'credentialType' | 'apiKey' | 'credentialIdApiKey'>,\n): boolean {\n    const provider = cfg.provider || 'openai';\n    if (provider === 'system') {\n        return true;\n    }\n    if (provider === 'none' || provider === 'pi-agent' || provider === 'custom') {\n        return false;\n    }\n    return cfg.credentialType === 'manager' ? !text(cfg.credentialIdApiKey) : !text(cfg.apiKey);\n}\n\n/**\n * Read `system.ai`; a missing object (or one the adapter may not read) gives `undefined`.\n *\n * @param adapter the adapter\n */\nexport async function readSystemAi(adapter: ioBroker.Adapter): Promise<Record<string, unknown> | undefined> {\n    try {\n        const obj = await adapter.getForeignObjectAsync(SYSTEM_AI_ID);\n        return obj?.native;\n    } catch {\n        return undefined;\n    }\n}\n\n/**\n * The cloud LLM of `system.ai`, or why there is none.\n *\n * @param adapter the adapter (reads the credential)\n * @param native the `native` of `system.ai`\n * @param explicit the provider `system` was chosen (otherwise the instance only has no key of its own)\n */\nexport async function resolveSystemLlm(\n    adapter: ioBroker.Adapter,\n    native: Record<string, unknown> | null | undefined,\n    explicit: boolean,\n): Promise<ResolvedLlm | { error: string }> {\n    const sel = systemAiSelection(native);\n    if (!sel) {\n        return {\n            error: explicit\n                ? `The system-wide AI settings have no provider — set one up in ${WHERE} or choose a provider here.`\n                : `No API key configured — enter one in the adapter settings or set up the system-wide AI settings in ${WHERE}.`,\n        };\n    }\n    if (sel.provider === 'custom' && !sel.baseUrl) {\n        return { error: `The system-wide AI settings use an OpenAI-compatible server without address (${WHERE}).` };\n    }\n    const apiKey = effectiveApiKey(sel.provider, await readCentralKey(adapter, sel.credentialId));\n    if (!apiKey) {\n        return { error: `The system-wide AI settings have no API key for ${sel.provider} (${WHERE}).` };\n    }\n    const preset = resolveProvider(sel.provider, sel.baseUrl);\n    return {\n        provider: sel.provider,\n        apiKey,\n        model: sel.model || preset.defaultModel,\n        baseUrl: preset.baseUrl,\n        system: true,\n    };\n}\n\n/**\n * The cloud LLM of the instance: its own provider and key, or those of `system.ai` (see {@link usesSystemAi}).\n * Not for the providers `none` and `pi-agent`, which have no LLM of their own.\n *\n * @param adapter the adapter\n * @param cfg the configuration of the instance\n */\nexport async function resolveLlm(\n    adapter: ioBroker.Adapter,\n    cfg: AdapterConfig,\n): Promise<ResolvedLlm | { error: string }> {\n    if (usesSystemAi(cfg)) {\n        return resolveSystemLlm(adapter, await readSystemAi(adapter), cfg.provider === 'system');\n    }\n    const provider = (cfg.provider || 'openai') as Provider;\n    const apiKey = effectiveApiKey(provider, await resolveApiKey(adapter, cfg));\n    const preset = resolveProvider(provider, cfg.baseUrl);\n    return { provider, apiKey, model: cfg.model || preset.defaultModel, baseUrl: preset.baseUrl, system: false };\n}\n"]}

package/build/lib/voice/engines.js

@@ -37,6 +37,7 @@ exports.createSttEngine = createSttEngine;
 exports.createTtsEngine = createTtsEngine;
 exports.listVoices = listVoices;
 exports.listSttModels = listSttModels;
+exports.listTtsModels = listTtsModels;
 /**
  * Factory that builds the configured STT/TTS engine. Cloud engines (OpenAI/Azure/AWS) use resolved
  * credentials; local engines (Vosk/Piper) download their models on demand into the data dir.
@@ -48,6 +49,7 @@ const azure_1 = require("./azure");
 const aws_1 = require("./aws");
 const vosk_1 = require("./vosk");
 const piper_1 = require("./piper");
+const openrouter_1 = require("./openrouter");
 const path = __importStar(require("node:path"));
 const ttsCache_1 = require("./ttsCache");
 const fallback_1 = require("./fallback");
@@ -67,6 +69,11 @@ function requireOpenAi(creds) {
         throw new Error('OpenAI speech needs an API key');
     }
 }
+function requireOpenRouter(creds) {
+    if (!creds.openrouterKey) {
+        throw new Error('OpenRouter speech needs an API key');
+    }
+}
 function createSttEngine(provider, ctx) {
     const primary = { name: provider, engine: buildSttEngine(provider, ctx) };
     const fallback = ctx.sttFallback && ctx.sttFallback !== provider ? ctx.sttFallback : '';
@@ -91,6 +98,9 @@ function buildSttEngine(provider, ctx) {
             return new aws_1.AwsStt(ctx.creds.aws);
         case 'vosk':
             return new vosk_1.VoskStt(ctx.dataDir, ctx.voskModel, ctx.log);
+        case 'openrouter':
+            requireOpenRouter(ctx.creds);
+            return new openrouter_1.OpenRouterStt(ctx.creds.openrouterKey, ctx.openrouterSttModel);
         case 'piper':
             throw new Error('Piper is a text-to-speech engine, not speech-to-text');
         default:
@@ -126,7 +136,9 @@ function cachedTts(provider, ctx) {
             ? ctx.voices.aws
             : provider === 'piper'
                 ? ctx.voices.piper
-                : ctx.voices.openai;
+                : provider === 'openrouter'
+                    ? `${ctx.openrouterTtsModel || openrouter_1.DEFAULT_OPENROUTER_TTS_MODEL}-${ctx.voices.openrouter}`
+                    : ctx.voices.openai;
     const dir = path.join(ctx.dataDir, 'ttscache', `${provider}-${(voice || 'default').replace(/[^\w.-]/g, '_')}`);
     return new ttsCache_1.CachedTts(buildTtsEngine(provider, ctx), { dir, log: ctx.log });
 }
@@ -142,6 +154,9 @@ function buildTtsEngine(provider, ctx) {
             return new piper_1.PiperTts(ctx.dataDir, ctx.voices.piper, ctx.log);
         case 'vosk':
             throw new Error('Vosk is a speech-to-text engine, not text-to-speech');
+        case 'openrouter':
+            requireOpenRouter(ctx.creds);
+            return new openrouter_1.OpenRouterTts(ctx.creds.openrouterKey, ctx.openrouterTtsModel, ctx.voices.openrouter, ctx.decodeAudio);
         default:
             requireOpenAi(ctx.creds);
             return new tts_1.OpenAiTts(ctx.creds.openaiKey, (ctx.voices.openai || 'alloy'), ctx.ttsModel || undefined);
@@ -160,12 +175,25 @@ async function listVoices(provider, ctx, language) {
             return (0, aws_1.listPollyVoices)(ctx.creds.aws, (0, lang_1.isoToLocale)(language));
         case 'piper':
             return (0, piper_1.listPiperVoices)();
+        case 'openrouter': {
+            // public list, no key needed: the voices of the chosen model
+            const model = ctx.openrouterTtsModel || openrouter_1.DEFAULT_OPENROUTER_TTS_MODEL;
+            const models = await (0, openrouter_1.listOpenRouterSpeechModels)('speech');
+            return (0, openrouter_1.voicesForLanguage)(models.find(m => m.id === model)?.voices || [], language);
+        }
         default:
             return OPENAI_VOICES;
     }
 }
-/** Suggested STT model names for the settings dropdown. Only Vosk has a local model catalogue. */
-function listSttModels(provider, language) {
+/** Suggested STT model names for the settings dropdown: the local Vosk catalogue, or the models of OpenRouter. */
+async function listSttModels(provider, language) {
+    if (provider === 'openrouter') {
+        return (await (0, openrouter_1.listOpenRouterSpeechModels)('transcription')).map(m => m.id);
+    }
     return provider === 'vosk' ? (0, vosk_1.listVoskModels)(language) : [];
 }
+/** TTS model names for the settings dropdown — only OpenRouter has a choice worth listing. */
+async function listTtsModels(provider) {
+    return provider === 'openrouter' ? (await (0, openrouter_1.listOpenRouterSpeechModels)('speech')).map(m => m.id) : [];
+}
 //# sourceMappingURL=engines.js.map

package/build/lib/voice/engines.js.map

@@ -1 +1 @@
-{"version":3,"file":"engines.js","sourceRoot":"","sources":["../../../src/lib/voice/engines.ts"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAwEA,0CAYC;AAoBD,0CAeC;AA6CD,gCAaC;AAGD,sCAEC;AAtLD;;;;GAIG;AACH,+BAAkD;AAClD,+BAAoE;AACpE,mCAA8D;AAC9D,+BAAuE;AACvE,iCAAiD;AACjD,mCAAoD;AACpD,gDAAkC;AAElC,yCAAuC;AACvC,yCAAwE;AACxE,iCAAqC;AAuCrC,SAAS,YAAY,CAAC,KAAuB;IACzC,IAAI,CAAC,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,CAAC,WAAW,EAAE,CAAC;QACxC,MAAM,IAAI,KAAK,CAAC,kDAAkD,CAAC,CAAC;IACxE,CAAC;AACL,CAAC;AAED,SAAS,UAAU,CAAC,KAAuB;IACvC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,WAAW,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,eAAe,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,MAAM,EAAE,CAAC;QAC5E,MAAM,IAAI,KAAK,CAAC,iEAAiE,CAAC,CAAC;IACvF,CAAC;AACL,CAAC;AAED,SAAS,aAAa,CAAC,KAAuB;IAC1C,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,CAAC;QACnB,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,CAAC;IACtD,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAClG,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IAC1G,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC;QACnE,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACrC,KAAK,MAAM;YACP,OAAO,IAAI,cAAO,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,SAAS,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAC5D,KAAK,OAAO;YACR,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QAC5E;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,GAAG,CAAC,QAAQ,IAAI,SAAS,CAAC,CAAC;IAC7E,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,sGAAsG;IACtG,iGAAiG;IACjG,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAC7F,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IACrG,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,sGAAsG;QACtG,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED;;;GAGG;AACH,SAAS,SAAS,CAAC,QAAwB,EAAE,GAAkB;IAC3D,MAAM,KAAK,GACP,QAAQ,KAAK,OAAO;QAChB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;QAClB,CAAC,CAAC,QAAQ,KAAK,KAAK;YAClB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG;YAChB,CAAC,CAAC,QAAQ,KAAK,OAAO;gBACpB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;gBAClB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC;IAChC,MAAM,GAAG,GAAG,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,EAAE,GAAG,QAAQ,IAAI,CAAC,KAAK,IAAI,SAAS,CAAC,CAAC,OAAO,CAAC,UAAU,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IAC/G,OAAO,IAAI,oBAAS,CAAC,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,GAAG,EAAE,CAAC,CAAC;AAC/E,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QACrF,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QACrD,KAAK,OAAO;YACR,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAChE,KAAK,MAAM;YACP,MAAM,IAAI,KAAK,CAAC,qDAAqD,CAAC,CAAC;QAC3E;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAChB,GAAG,CAAC,KAAK,CAAC,SAAS,EACnB,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,IAAI,OAAO,CAAgB,EAC7C,GAAG,CAAC,QAAQ,IAAI,SAAS,CAC5B,CAAC;IACV,CAAC;AACL,CAAC;AAED,uDAAuD;AACvD,MAAM,aAAa,GAAG,CAAC,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,CAAC,CAAC;AAE5E,sFAAsF;AAC/E,KAAK,UAAU,UAAU,CAAC,QAAwB,EAAE,GAAkB,EAAE,QAAgB;IAC3F,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAA,uBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QAC7F,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAA,qBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QACjE,KAAK,OAAO;YACR,OAAO,IAAA,uBAAe,GAAE,CAAC;QAC7B;YACI,OAAO,aAAa,CAAC;IAC7B,CAAC;AACL,CAAC;AAED,kGAAkG;AAClG,SAAgB,aAAa,CAAC,QAAwB,EAAE,QAAgB;IACpE,OAAO,QAAQ,KAAK,MAAM,CAAC,CAAC,CAAC,IAAA,qBAAc,EAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AAC/D,CAAC","sourcesContent":["/**\n * Factory that builds the configured STT/TTS engine. Cloud engines (OpenAI/Azure/AWS) use resolved\n * credentials; local engines (Vosk/Piper) download their models on demand into the data dir.\n * Adding a provider = one new engine class + one `case` here.\n */\nimport { OpenAiStt, type SttEngine } from './stt';\nimport { OpenAiTts, type OpenAiVoice, type TtsEngine } from './tts';\nimport { AzureStt, AzureTts, listAzureVoices } from './azure';\nimport { AwsStt, AwsTts, listPollyVoices, type AwsCreds } from './aws';\nimport { VoskStt, listVoskModels } from './vosk';\nimport { PiperTts, listPiperVoices } from './piper';\nimport * as path from 'node:path';\n\nimport { CachedTts } from './ttsCache';\nimport { FallbackStt, FallbackTts, type NamedEngine } from './fallback';\nimport { isoToLocale } from './lang';\nimport type { VoiceLogger } from './download';\n\nexport type SpeechProvider = 'openai' | 'azure' | 'aws' | 'vosk' | 'piper';\n\n/** All speech credentials, resolved from config (or the central credential store). */\nexport interface VoiceCredentials {\n    openaiKey: string;\n    azureKey: string;\n    azureRegion: string;\n    aws: AwsCreds;\n}\n\n/** Provider-specific voice ids (each provider names voices differently). */\nexport interface VoiceNames {\n    openai: string; // alloy/echo/…\n    azure: string; // e.g. de-DE-KatjaNeural\n    aws: string; // e.g. Vicki\n    piper: string; // e.g. de_DE-thorsten-medium\n}\n\n/** Everything the factory needs to build any engine. */\nexport interface EngineContext {\n    creds: VoiceCredentials;\n    voices: VoiceNames;\n    /** Optional second engine, used when the primary one fails ('' = none). */\n    sttFallback?: SpeechProvider | '';\n    ttsFallback?: SpeechProvider | '';\n    /** Writable instance data dir (for the local engines' downloads). */\n    dataDir: string;\n    log: VoiceLogger;\n    /** Local model overrides ('' → default per language). */\n    voskModel: string;\n    /** OpenAI STT model ('' → 'whisper-1'; e.g. 'gpt-4o-transcribe'). */\n    sttModel: string;\n    /** OpenAI TTS model ('' → 'tts-1'; e.g. 'gpt-4o-mini-tts'). */\n    ttsModel: string;\n}\n\nfunction requireAzure(creds: VoiceCredentials): void {\n    if (!creds.azureKey || !creds.azureRegion) {\n        throw new Error('Azure speech needs a subscription key and region');\n    }\n}\n\nfunction requireAws(creds: VoiceCredentials): void {\n    if (!creds.aws.accessKeyId || !creds.aws.secretAccessKey || !creds.aws.region) {\n        throw new Error('AWS speech needs an access key id, secret access key and region');\n    }\n}\n\nfunction requireOpenAi(creds: VoiceCredentials): void {\n    if (!creds.openaiKey) {\n        throw new Error('OpenAI speech needs an API key');\n    }\n}\n\nexport function createSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    const primary: NamedEngine<SttEngine> = { name: provider, engine: buildSttEngine(provider, ctx) };\n    const fallback = ctx.sttFallback && ctx.sttFallback !== provider ? ctx.sttFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackStt([primary, { name: fallback, engine: buildSttEngine(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        ctx.log.warn(`Speech-to-text fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\nfunction buildSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureStt(ctx.creds.azureKey, ctx.creds.azureRegion);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsStt(ctx.creds.aws);\n        case 'vosk':\n            return new VoskStt(ctx.dataDir, ctx.voskModel, ctx.log);\n        case 'piper':\n            throw new Error('Piper is a text-to-speech engine, not speech-to-text');\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiStt(ctx.creds.openaiKey, ctx.sttModel || undefined);\n    }\n}\n\nexport function createTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    // Each engine gets the shared behaviour (disk cache, length limit, SSML) and its own cache directory,\n    // so a fallback engine's voice can never be served later while the primary one is healthy again.\n    const primary: NamedEngine<TtsEngine> = { name: provider, engine: cachedTts(provider, ctx) };\n    const fallback = ctx.ttsFallback && ctx.ttsFallback !== provider ? ctx.ttsFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackTts([primary, { name: fallback, engine: cachedTts(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        // A fallback that cannot even be built (missing key, wrong provider) must not take the main one down.\n        ctx.log.warn(`Text-to-speech fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\n/**\n * One provider binding plus the shared cache/length/SSML layer. One cache directory per provider+voice —\n * a voice change must not serve audio in the old voice.\n */\nfunction cachedTts(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    const voice =\n        provider === 'azure'\n            ? ctx.voices.azure\n            : provider === 'aws'\n              ? ctx.voices.aws\n              : provider === 'piper'\n                ? ctx.voices.piper\n                : ctx.voices.openai;\n    const dir = path.join(ctx.dataDir, 'ttscache', `${provider}-${(voice || 'default').replace(/[^\\w.-]/g, '_')}`);\n    return new CachedTts(buildTtsEngine(provider, ctx), { dir, log: ctx.log });\n}\n\nfunction buildTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureTts(ctx.creds.azureKey, ctx.creds.azureRegion, ctx.voices.azure);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsTts(ctx.creds.aws, ctx.voices.aws);\n        case 'piper':\n            return new PiperTts(ctx.dataDir, ctx.voices.piper, ctx.log);\n        case 'vosk':\n            throw new Error('Vosk is a speech-to-text engine, not text-to-speech');\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiTts(\n                ctx.creds.openaiKey,\n                (ctx.voices.openai || 'alloy') as OpenAiVoice,\n                ctx.ttsModel || undefined,\n            );\n    }\n}\n\n/** OpenAI TTS voices are a fixed set (no list API). */\nconst OPENAI_VOICES = ['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'];\n\n/** List the available TTS voices for a provider (for the settings voice dropdown). */\nexport async function listVoices(provider: SpeechProvider, ctx: EngineContext, language: string): Promise<string[]> {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return listAzureVoices(ctx.creds.azureKey, ctx.creds.azureRegion, isoToLocale(language));\n        case 'aws':\n            requireAws(ctx.creds);\n            return listPollyVoices(ctx.creds.aws, isoToLocale(language));\n        case 'piper':\n            return listPiperVoices();\n        default:\n            return OPENAI_VOICES;\n    }\n}\n\n/** Suggested STT model names for the settings dropdown. Only Vosk has a local model catalogue. */\nexport function listSttModels(provider: SpeechProvider, language: string): string[] {\n    return provider === 'vosk' ? listVoskModels(language) : [];\n}\n"]}
+{"version":3,"file":"engines.js","sourceRoot":"","sources":["../../../src/lib/voice/engines.ts"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA6FA,0CAYC;AAuBD,0CAeC;AAuDD,gCAmBC;AAGD,sCAKC;AAGD,sCAEC;AAtOD;;;;GAIG;AACH,+BAAkD;AAClD,+BAAoE;AACpE,mCAA8D;AAC9D,+BAAuE;AACvE,iCAAiD;AACjD,mCAAoD;AACpD,6CAOsB;AACtB,gDAAkC;AAElC,yCAAuC;AACvC,yCAAwE;AACxE,iCAAqC;AA8CrC,SAAS,YAAY,CAAC,KAAuB;IACzC,IAAI,CAAC,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,CAAC,WAAW,EAAE,CAAC;QACxC,MAAM,IAAI,KAAK,CAAC,kDAAkD,CAAC,CAAC;IACxE,CAAC;AACL,CAAC;AAED,SAAS,UAAU,CAAC,KAAuB;IACvC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,WAAW,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,eAAe,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,MAAM,EAAE,CAAC;QAC5E,MAAM,IAAI,KAAK,CAAC,iEAAiE,CAAC,CAAC;IACvF,CAAC;AACL,CAAC;AAED,SAAS,aAAa,CAAC,KAAuB;IAC1C,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,CAAC;QACnB,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,CAAC;IACtD,CAAC;AACL,CAAC;AAED,SAAS,iBAAiB,CAAC,KAAuB;IAC9C,IAAI,CAAC,KAAK,CAAC,aAAa,EAAE,CAAC;QACvB,MAAM,IAAI,KAAK,CAAC,oCAAoC,CAAC,CAAC;IAC1D,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAClG,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IAC1G,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC;QACnE,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACrC,KAAK,MAAM;YACP,OAAO,IAAI,cAAO,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,SAAS,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAC5D,KAAK,YAAY;YACb,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAC7B,OAAO,IAAI,0BAAa,CAAC,GAAG,CAAC,KAAK,CAAC,aAAa,EAAE,GAAG,CAAC,kBAAkB,CAAC,CAAC;QAC9E,KAAK,OAAO;YACR,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QAC5E;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,GAAG,CAAC,QAAQ,IAAI,SAAS,CAAC,CAAC;IAC7E,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,sGAAsG;IACtG,iGAAiG;IACjG,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAC7F,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IACrG,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,sGAAsG;QACtG,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED;;;GAGG;AACH,SAAS,SAAS,CAAC,QAAwB,EAAE,GAAkB;IAC3D,MAAM,KAAK,GACP,QAAQ,KAAK,OAAO;QAChB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;QAClB,CAAC,CAAC,QAAQ,KAAK,KAAK;YAClB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG;YAChB,CAAC,CAAC,QAAQ,KAAK,OAAO;gBACpB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;gBAClB,CAAC,CAAC,QAAQ,KAAK,YAAY;oBACzB,CAAC,CAAC,GAAG,GAAG,CAAC,kBAAkB,IAAI,yCAA4B,IAAI,GAAG,CAAC,MAAM,CAAC,UAAU,EAAE;oBACtF,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC;IAClC,MAAM,GAAG,GAAG,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,EAAE,GAAG,QAAQ,IAAI,CAAC,KAAK,IAAI,SAAS,CAAC,CAAC,OAAO,CAAC,UAAU,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IAC/G,OAAO,IAAI,oBAAS,CAAC,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,GAAG,EAAE,CAAC,CAAC;AAC/E,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QACrF,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QACrD,KAAK,OAAO;YACR,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAChE,KAAK,MAAM;YACP,MAAM,IAAI,KAAK,CAAC,qDAAqD,CAAC,CAAC;QAC3E,KAAK,YAAY;YACb,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAC7B,OAAO,IAAI,0BAAa,CACpB,GAAG,CAAC,KAAK,CAAC,aAAa,EACvB,GAAG,CAAC,kBAAkB,EACtB,GAAG,CAAC,MAAM,CAAC,UAAU,EACrB,GAAG,CAAC,WAAW,CAClB,CAAC;QACN;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAChB,GAAG,CAAC,KAAK,CAAC,SAAS,EACnB,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,IAAI,OAAO,CAAgB,EAC7C,GAAG,CAAC,QAAQ,IAAI,SAAS,CAC5B,CAAC;IACV,CAAC;AACL,CAAC;AAED,uDAAuD;AACvD,MAAM,aAAa,GAAG,CAAC,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,CAAC,CAAC;AAE5E,sFAAsF;AAC/E,KAAK,UAAU,UAAU,CAAC,QAAwB,EAAE,GAAkB,EAAE,QAAgB;IAC3F,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAA,uBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QAC7F,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAA,qBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QACjE,KAAK,OAAO;YACR,OAAO,IAAA,uBAAe,GAAE,CAAC;QAC7B,KAAK,YAAY,CAAC,CAAC,CAAC;YAChB,6DAA6D;YAC7D,MAAM,KAAK,GAAG,GAAG,CAAC,kBAAkB,IAAI,yCAA4B,CAAC;YACrE,MAAM,MAAM,GAAG,MAAM,IAAA,uCAA0B,EAAC,QAAQ,CAAC,CAAC;YAC1D,OAAO,IAAA,8BAAiB,EAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,KAAK,CAAC,EAAE,MAAM,IAAI,EAAE,EAAE,QAAQ,CAAC,CAAC;QACvF,CAAC;QACD;YACI,OAAO,aAAa,CAAC;IAC7B,CAAC;AACL,CAAC;AAED,kHAAkH;AAC3G,KAAK,UAAU,aAAa,CAAC,QAAwB,EAAE,QAAgB;IAC1E,IAAI,QAAQ,KAAK,YAAY,EAAE,CAAC;QAC5B,OAAO,CAAC,MAAM,IAAA,uCAA0B,EAAC,eAAe,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAC9E,CAAC;IACD,OAAO,QAAQ,KAAK,MAAM,CAAC,CAAC,CAAC,IAAA,qBAAc,EAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AAC/D,CAAC;AAED,8FAA8F;AACvF,KAAK,UAAU,aAAa,CAAC,QAAwB;IACxD,OAAO,QAAQ,KAAK,YAAY,CAAC,CAAC,CAAC,CAAC,MAAM,IAAA,uCAA0B,EAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AACxG,CAAC","sourcesContent":["/**\n * Factory that builds the configured STT/TTS engine. Cloud engines (OpenAI/Azure/AWS) use resolved\n * credentials; local engines (Vosk/Piper) download their models on demand into the data dir.\n * Adding a provider = one new engine class + one `case` here.\n */\nimport { OpenAiStt, type SttEngine } from './stt';\nimport { OpenAiTts, type OpenAiVoice, type TtsEngine } from './tts';\nimport { AzureStt, AzureTts, listAzureVoices } from './azure';\nimport { AwsStt, AwsTts, listPollyVoices, type AwsCreds } from './aws';\nimport { VoskStt, listVoskModels } from './vosk';\nimport { PiperTts, listPiperVoices } from './piper';\nimport {\n    OpenRouterStt,\n    OpenRouterTts,\n    listOpenRouterSpeechModels,\n    voicesForLanguage,\n    DEFAULT_OPENROUTER_TTS_MODEL,\n    type AudioDecoder,\n} from './openrouter';\nimport * as path from 'node:path';\n\nimport { CachedTts } from './ttsCache';\nimport { FallbackStt, FallbackTts, type NamedEngine } from './fallback';\nimport { isoToLocale } from './lang';\nimport type { VoiceLogger } from './download';\n\nexport type SpeechProvider = 'openai' | 'openrouter' | 'azure' | 'aws' | 'vosk' | 'piper';\n\n/** All speech credentials, resolved from config (or the central credential store). */\nexport interface VoiceCredentials {\n    openaiKey: string;\n    openrouterKey: string;\n    azureKey: string;\n    azureRegion: string;\n    aws: AwsCreds;\n}\n\n/** Provider-specific voice ids (each provider names voices differently). */\nexport interface VoiceNames {\n    openai: string; // alloy/echo/…\n    azure: string; // e.g. de-DE-KatjaNeural\n    aws: string; // e.g. Vicki\n    piper: string; // e.g. de_DE-thorsten-medium\n    openrouter: string; // depends on the model, e.g. Kore, george\n}\n\n/** Everything the factory needs to build any engine. */\nexport interface EngineContext {\n    creds: VoiceCredentials;\n    voices: VoiceNames;\n    /** Optional second engine, used when the primary one fails ('' = none). */\n    sttFallback?: SpeechProvider | '';\n    ttsFallback?: SpeechProvider | '';\n    /** Writable instance data dir (for the local engines' downloads). */\n    dataDir: string;\n    log: VoiceLogger;\n    /** Local model overrides ('' → default per language). */\n    voskModel: string;\n    /** OpenAI STT model ('' → 'whisper-1'; e.g. 'gpt-4o-transcribe'). */\n    sttModel: string;\n    /** OpenAI TTS model ('' → 'tts-1'; e.g. 'gpt-4o-mini-tts'). */\n    ttsModel: string;\n    /** OpenRouter models ('' → the defaults of openrouter.ts) */\n    openrouterSttModel: string;\n    openrouterTtsModel: string;\n    /** Decoder for compressed speech (OpenRouter models that only return mp3); absent → those models fail */\n    decodeAudio?: AudioDecoder;\n}\n\nfunction requireAzure(creds: VoiceCredentials): void {\n    if (!creds.azureKey || !creds.azureRegion) {\n        throw new Error('Azure speech needs a subscription key and region');\n    }\n}\n\nfunction requireAws(creds: VoiceCredentials): void {\n    if (!creds.aws.accessKeyId || !creds.aws.secretAccessKey || !creds.aws.region) {\n        throw new Error('AWS speech needs an access key id, secret access key and region');\n    }\n}\n\nfunction requireOpenAi(creds: VoiceCredentials): void {\n    if (!creds.openaiKey) {\n        throw new Error('OpenAI speech needs an API key');\n    }\n}\n\nfunction requireOpenRouter(creds: VoiceCredentials): void {\n    if (!creds.openrouterKey) {\n        throw new Error('OpenRouter speech needs an API key');\n    }\n}\n\nexport function createSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    const primary: NamedEngine<SttEngine> = { name: provider, engine: buildSttEngine(provider, ctx) };\n    const fallback = ctx.sttFallback && ctx.sttFallback !== provider ? ctx.sttFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackStt([primary, { name: fallback, engine: buildSttEngine(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        ctx.log.warn(`Speech-to-text fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\nfunction buildSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureStt(ctx.creds.azureKey, ctx.creds.azureRegion);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsStt(ctx.creds.aws);\n        case 'vosk':\n            return new VoskStt(ctx.dataDir, ctx.voskModel, ctx.log);\n        case 'openrouter':\n            requireOpenRouter(ctx.creds);\n            return new OpenRouterStt(ctx.creds.openrouterKey, ctx.openrouterSttModel);\n        case 'piper':\n            throw new Error('Piper is a text-to-speech engine, not speech-to-text');\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiStt(ctx.creds.openaiKey, ctx.sttModel || undefined);\n    }\n}\n\nexport function createTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    // Each engine gets the shared behaviour (disk cache, length limit, SSML) and its own cache directory,\n    // so a fallback engine's voice can never be served later while the primary one is healthy again.\n    const primary: NamedEngine<TtsEngine> = { name: provider, engine: cachedTts(provider, ctx) };\n    const fallback = ctx.ttsFallback && ctx.ttsFallback !== provider ? ctx.ttsFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackTts([primary, { name: fallback, engine: cachedTts(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        // A fallback that cannot even be built (missing key, wrong provider) must not take the main one down.\n        ctx.log.warn(`Text-to-speech fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\n/**\n * One provider binding plus the shared cache/length/SSML layer. One cache directory per provider+voice —\n * a voice change must not serve audio in the old voice.\n */\nfunction cachedTts(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    const voice =\n        provider === 'azure'\n            ? ctx.voices.azure\n            : provider === 'aws'\n              ? ctx.voices.aws\n              : provider === 'piper'\n                ? ctx.voices.piper\n                : provider === 'openrouter'\n                  ? `${ctx.openrouterTtsModel || DEFAULT_OPENROUTER_TTS_MODEL}-${ctx.voices.openrouter}`\n                  : ctx.voices.openai;\n    const dir = path.join(ctx.dataDir, 'ttscache', `${provider}-${(voice || 'default').replace(/[^\\w.-]/g, '_')}`);\n    return new CachedTts(buildTtsEngine(provider, ctx), { dir, log: ctx.log });\n}\n\nfunction buildTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureTts(ctx.creds.azureKey, ctx.creds.azureRegion, ctx.voices.azure);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsTts(ctx.creds.aws, ctx.voices.aws);\n        case 'piper':\n            return new PiperTts(ctx.dataDir, ctx.voices.piper, ctx.log);\n        case 'vosk':\n            throw new Error('Vosk is a speech-to-text engine, not text-to-speech');\n        case 'openrouter':\n            requireOpenRouter(ctx.creds);\n            return new OpenRouterTts(\n                ctx.creds.openrouterKey,\n                ctx.openrouterTtsModel,\n                ctx.voices.openrouter,\n                ctx.decodeAudio,\n            );\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiTts(\n                ctx.creds.openaiKey,\n                (ctx.voices.openai || 'alloy') as OpenAiVoice,\n                ctx.ttsModel || undefined,\n            );\n    }\n}\n\n/** OpenAI TTS voices are a fixed set (no list API). */\nconst OPENAI_VOICES = ['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'];\n\n/** List the available TTS voices for a provider (for the settings voice dropdown). */\nexport async function listVoices(provider: SpeechProvider, ctx: EngineContext, language: string): Promise<string[]> {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return listAzureVoices(ctx.creds.azureKey, ctx.creds.azureRegion, isoToLocale(language));\n        case 'aws':\n            requireAws(ctx.creds);\n            return listPollyVoices(ctx.creds.aws, isoToLocale(language));\n        case 'piper':\n            return listPiperVoices();\n        case 'openrouter': {\n            // public list, no key needed: the voices of the chosen model\n            const model = ctx.openrouterTtsModel || DEFAULT_OPENROUTER_TTS_MODEL;\n            const models = await listOpenRouterSpeechModels('speech');\n            return voicesForLanguage(models.find(m => m.id === model)?.voices || [], language);\n        }\n        default:\n            return OPENAI_VOICES;\n    }\n}\n\n/** Suggested STT model names for the settings dropdown: the local Vosk catalogue, or the models of OpenRouter. */\nexport async function listSttModels(provider: SpeechProvider, language: string): Promise<string[]> {\n    if (provider === 'openrouter') {\n        return (await listOpenRouterSpeechModels('transcription')).map(m => m.id);\n    }\n    return provider === 'vosk' ? listVoskModels(language) : [];\n}\n\n/** TTS model names for the settings dropdown — only OpenRouter has a choice worth listing. */\nexport async function listTtsModels(provider: SpeechProvider): Promise<string[]> {\n    return provider === 'openrouter' ? (await listOpenRouterSpeechModels('speech')).map(m => m.id) : [];\n}\n"]}

package/build/lib/voice/openrouter.js

@@ -0,0 +1,220 @@
+"use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.OpenRouterTts = exports.OpenRouterStt = exports.DEFAULT_OPENROUTER_TTS_MODEL = exports.DEFAULT_OPENROUTER_STT_MODEL = void 0;
+exports.listOpenRouterSpeechModels = listOpenRouterSpeechModels;
+exports.voicesForLanguage = voicesForLanguage;
+exports.pcmFromResponse = pcmFromResponse;
+/**
+ * Speech over OpenRouter: one key for the speech-to-text and text-to-speech models of many vendors (OpenAI,
+ * ElevenLabs, Google, Deepgram, Mistral …) behind the OpenAI audio API at https://openrouter.ai/api/v1.
+ * Model ids carry the vendor (`openai/whisper-large-v3-turbo`), and the voices depend on the model — the public
+ * model list names them (`supported_voices`), so the settings offer exactly the voices of the chosen model.
+ *
+ * Differences to OpenAI itself: transcription ignores `prompt` (no vocabulary hints), and speech comes back as
+ * `pcm` (most models, the default) or `mp3` (MiniMax only). The PCM rate is not part of the API — OpenRouter
+ * documents 24 kHz, the rate of OpenAI, Gemini and Azure — so a rate in the content type wins over that default.
+ */
+const openai_1 = __importStar(require("openai"));
+const llm_1 = require("../llm");
+const stt_1 = require("./stt");
+/** Speech-to-text model when none is configured: Whisper large v3 turbo, multilingual and the cheapest Whisper */
+exports.DEFAULT_OPENROUTER_STT_MODEL = 'openai/whisper-large-v3-turbo';
+/**
+ * Text-to-speech model when none is configured: multilingual, PCM (no decoder needed) and fast — in a test about
+ * 0.8 s for a sentence where Gemini TTS took 2.5 s, which a spoken answer waits for.
+ */
+exports.DEFAULT_OPENROUTER_TTS_MODEL = 'elevenlabs/eleven-flash-v2.5';
+/** Rate of the PCM OpenRouter returns when the content type does not say otherwise */
+const PCM_RATE = 24000;
+/** Vendors whose speech models only return mp3 (they answer `pcm` with a 400) */
+const MP3_ONLY = ['minimax/'];
+/** How long the public model list is kept */
+const MODEL_LIST_TTL_MS = 60 * 60 * 1000;
+function client(apiKey) {
+    return new openai_1.default({ apiKey, baseURL: llm_1.PROVIDER_PRESETS.openrouter.baseUrl, defaultHeaders: llm_1.OPENROUTER_HEADERS });
+}
+const modelLists = new Map();
+/**
+ * The speech models of OpenRouter. The list is public (no key), cached for an hour; a failed request is not kept.
+ *
+ * @param kind `speech` = text-to-speech, `transcription` = speech-to-text
+ * @param fetchImpl fetch (tests)
+ */
+function listOpenRouterSpeechModels(kind, fetchImpl = fetch) {
+    const cached = modelLists.get(kind);
+    if (cached && Date.now() - cached.at < MODEL_LIST_TTL_MS) {
+        return cached.models;
+    }
+    const models = (async () => {
+        const res = await fetchImpl(`${llm_1.PROVIDER_PRESETS.openrouter.baseUrl}/models?output_modalities=${kind}`, {
+            headers: llm_1.OPENROUTER_HEADERS,
+            signal: AbortSignal.timeout(15000),
+        });
+        if (!res.ok) {
+            throw new Error(`OpenRouter model list: HTTP ${res.status}`);
+        }
+        const body = (await res.json());
+        return (body.data || [])
+            .filter(m => typeof m.id === 'string' && m.id)
+            .map(m => ({
+            id: m.id,
+            name: typeof m.name === 'string' ? m.name : m.id,
+            voices: Array.isArray(m.supported_voices)
+                ? m.supported_voices.filter((v) => typeof v === 'string')
+                : [],
+        }));
+    })();
+    modelLists.set(kind, { at: Date.now(), models });
+    models.catch(() => modelLists.delete(kind));
+    return models;
+}
+/**
+ * The voices of a text-to-speech model. Voices named after a locale (`de-DE-Klaus:MAI-Voice-2`) are narrowed to
+ * the language, if that leaves any; other names (`Kore`, `george`) say nothing about the language.
+ *
+ * @param voices the voices of the model
+ * @param language ISO-639-1 language ('' = all)
+ */
+function voicesForLanguage(voices, language) {
+    const lang = (language || '').split('-')[0].toLowerCase();
+    if (!lang) {
+        return voices;
+    }
+    const matching = voices.filter(v => new RegExp(`^${lang}[-_][A-Z]{2}\\b`).test(v));
+    return matching.length ? matching : voices;
+}
+/**
+ * Raw PCM out of a speech response: WAV is unwrapped, the rate comes from the content type (`audio/pcm;rate=…`)
+ * or is the documented 24 kHz.
+ *
+ * @param audio the response body
+ * @param contentType the content type of the response
+ */
+function pcmFromResponse(audio, contentType) {
+    if (audio.length > 44 && audio.toString('latin1', 0, 4) === 'RIFF' && audio.toString('latin1', 8, 12) === 'WAVE') {
+        const sampleRate = audio.readUInt32LE(24);
+        // the data chunk is not always right after the 16-byte fmt chunk
+        let offset = 12;
+        while (offset + 8 <= audio.length) {
+            const id = audio.toString('latin1', offset, offset + 4);
+            const size = audio.readUInt32LE(offset + 4);
+            if (id === 'data') {
+                return { pcm: audio.subarray(offset + 8, Math.min(audio.length, offset + 8 + size)), sampleRate };
+            }
+            offset += 8 + size + (size % 2);
+        }
+        return { pcm: audio.subarray(44), sampleRate };
+    }
+    const rate = /(?:^|[;\s])(?:rate|sample_?rate)=(\d+)/i.exec(contentType || '');
+    return { pcm: audio, sampleRate: rate ? Number(rate[1]) : PCM_RATE };
+}
+/** Speech-to-text over OpenRouter (OpenAI-compatible multipart upload) */
+class OpenRouterStt {
+    client;
+    model;
+    constructor(apiKey, model = '') {
+        this.client = client(apiKey);
+        this.model = model || exports.DEFAULT_OPENROUTER_STT_MODEL;
+    }
+    async transcribe(pcm, sampleRate, lang) {
+        const file = await (0, openai_1.toFile)((0, stt_1.pcmToWav)(pcm, sampleRate), 'audio.wav', { type: 'audio/wav' });
+        // no vocabulary hints: OpenRouter accepts `prompt`, but ignores it
+        const res = await this.client.audio.transcriptions.create({
+            file,
+            model: this.model,
+            language: (lang || '').split('-')[0] || undefined,
+        });
+        return (res.text || '').trim();
+    }
+}
+exports.OpenRouterStt = OpenRouterStt;
+/** Text-to-speech over OpenRouter */
+class OpenRouterTts {
+    client;
+    model;
+    voice;
+    decode;
+    fetchImpl;
+    /**
+     * @param apiKey OpenRouter key
+     * @param model model id ('' = {@link DEFAULT_OPENROUTER_TTS_MODEL})
+     * @param voice voice of the model ('' = the first voice of the model for the language)
+     * @param decode decoder for the models that only return mp3
+     * @param fetchImpl fetch for the model list (tests)
+     */
+    constructor(apiKey, model = '', voice = '', decode, fetchImpl) {
+        this.client = client(apiKey);
+        this.model = model || exports.DEFAULT_OPENROUTER_TTS_MODEL;
+        this.voice = voice;
+        this.decode = decode;
+        this.fetchImpl = fetchImpl;
+    }
+    /** The configured voice, or the first one the model offers for the language — most models need a voice */
+    async voiceFor(lang) {
+        if (this.voice) {
+            return this.voice;
+        }
+        try {
+            const models = await listOpenRouterSpeechModels('speech', this.fetchImpl);
+            const voices = models.find(m => m.id === this.model)?.voices || [];
+            return voicesForLanguage(voices, lang)[0] || '';
+        }
+        catch {
+            return '';
+        }
+    }
+    async synthesize(text, lang) {
+        const mp3 = MP3_ONLY.some(prefix => this.model.startsWith(prefix));
+        if (mp3 && !this.decode) {
+            throw new Error(`${this.model} only returns mp3, which needs ffmpeg`);
+        }
+        const voice = await this.voiceFor(lang);
+        const res = await this.client.audio.speech.create({
+            model: this.model,
+            input: text,
+            // `voice` is required by the OpenAI types; an empty one is left out (a model with a default voice)
+            ...(voice ? { voice } : {}),
+            response_format: mp3 ? 'mp3' : 'pcm',
+        });
+        const audio = Buffer.from(await res.arrayBuffer());
+        if (mp3) {
+            return this.decode(audio);
+        }
+        return pcmFromResponse(audio, res.headers.get('content-type') || '');
+    }
+}
+exports.OpenRouterTts = OpenRouterTts;
+//# sourceMappingURL=openrouter.js.map