@andreprado/agentkit 0.1.0-alpha.14 → 0.1.0-alpha.16

package/src/runtime/config.ts

@@ -16,8 +16,11 @@ import type {
   TelegramAllowedUpdate,
   AgentKnowledgeEmbeddingProvider,
   ChannelBufferMode,
+  ChannelAudioMode,
+  TranscriptionProviderName,
 } from "../index";
 import { AgentKitError } from "./errors";
+import { isValidTimeZone } from "./prompt-context";
 
 const CONFIG_FILE = "agentkit.config.ts";
 const AGENT_RUNTIMES = new Set<AgentRuntime>(["local", "edge"]);
@@ -27,8 +30,10 @@ const TOOL_VISIBILITIES = new Set<ToolVisibility>(["user", "internal"]);
 const CHANNEL_TYPES = new Set<ChannelType>(["website", "telegram", "whatsapp"]);
 const CHANNEL_PROVIDERS = new Set<ChannelProvider>(["agentkit", "telegram", "zapster", "meta"]);
 const CHANNEL_BUFFER_MODES = new Set<ChannelBufferMode>(["off", "debounce"]);
+const CHANNEL_AUDIO_MODES = new Set<ChannelAudioMode>(["off", "reject", "transcribe"]);
 const TELEGRAM_ALLOWED_UPDATES = new Set<TelegramAllowedUpdate>(["message"]);
 const KNOWLEDGE_EMBEDDING_PROVIDERS = new Set<AgentKnowledgeEmbeddingProvider>(["none", "test", "openai"]);
+const TRANSCRIPTION_PROVIDERS = new Set<TranscriptionProviderName>(["test", "openai", "groq"]);
 const CHANNEL_NAME_PATTERN = /^[a-z][a-z0-9-]*$/;
 const SECRET_NAME_PATTERN = /^[A-Z_][A-Z0-9_]*$/;
 const moduleDir = dirname(fileURLToPath(import.meta.url));
@@ -207,8 +212,17 @@ function validateAgentConfig(value: unknown): asserts value is AgentConfig {
     throw invalidConfig('provider.model must be "fake" when provider.name is "test".');
   }
 
+  if (value.timeZone !== undefined) {
+    expectString(value.timeZone, "timeZone");
+
+    if (!isValidTimeZone(value.timeZone)) {
+      throw invalidConfig('timeZone must be a valid IANA time zone such as "UTC" or "America/New_York".');
+    }
+  }
+
   expectString(value.instructions, "instructions");
   expectStringArray(value.secrets, "secrets");
+  validateSecretNames(value.secrets, "secrets");
 
   if (value.tools !== undefined) {
     if (!Array.isArray(value.tools)) {
@@ -220,7 +234,8 @@ function validateAgentConfig(value: unknown): asserts value is AgentConfig {
     }
   }
 
-  validateChannels(value.channels, value.runtime as AgentRuntime);
+  validateTranscriptionConfig(value.transcription);
+  validateChannels(value.channels, value.runtime as AgentRuntime, Boolean(value.transcription));
   validateKnowledgeConfig(value.knowledge);
 
   if (!isRecord(value.access) || !ACCESS_MODES.has(value.access.mode as AccessMode)) {
@@ -243,6 +258,46 @@ function validateAgentConfig(value: unknown): asserts value is AgentConfig {
   validateFilesConfig(value.storage.files);
 }
 
+function validateTranscriptionConfig(value: unknown): void {
+  if (value === undefined) {
+    return;
+  }
+
+  if (!isRecord(value)) {
+    throw invalidConfig("transcription must be an object when provided.");
+  }
+
+  if (!TRANSCRIPTION_PROVIDERS.has(value.provider as TranscriptionProviderName)) {
+    throw invalidConfig(`transcription.provider must be one of: ${Array.from(TRANSCRIPTION_PROVIDERS).join(", ")}.`);
+  }
+
+  expectString(value.model, "transcription.model");
+
+  if (value.secret !== undefined) {
+    expectString(value.secret, "transcription.secret");
+    validateSecretNames([value.secret], "transcription.secret");
+  }
+
+  if (value.language !== undefined) {
+    expectString(value.language, "transcription.language");
+  }
+
+  if (value.prompt !== undefined) {
+    expectString(value.prompt, "transcription.prompt");
+  }
+
+  if (value.limits !== undefined) {
+    if (!isRecord(value.limits)) {
+      throw invalidConfig("transcription.limits must be an object when provided.");
+    }
+
+    validateOptionalPositiveInteger(value.limits.maxDurationSeconds, "transcription.limits.maxDurationSeconds");
+    validateOptionalPositiveInteger(value.limits.maxBytes, "transcription.limits.maxBytes");
+  }
+
+  validateOptionalPositiveInteger(value.rawAudioTtlSeconds, "transcription.rawAudioTtlSeconds");
+}
+
 function validateKnowledgeConfig(value: unknown): void {
   if (value === undefined) {
     return;
@@ -371,7 +426,7 @@ function validateKnowledgeRetrieval(value: unknown): void {
   }
 }
 
-function validateChannels(value: unknown, runtime: AgentRuntime): void {
+function validateChannels(value: unknown, runtime: AgentRuntime, hasTranscriptionConfig: boolean): void {
   if (value === undefined) {
     return;
   }
@@ -387,7 +442,7 @@ function validateChannels(value: unknown, runtime: AgentRuntime): void {
   const names = new Set<string>();
 
   for (const [index, channel] of value.entries()) {
-    validateChannel(channel, index);
+    validateChannel(channel, index, hasTranscriptionConfig);
 
     if (names.has(channel.name)) {
       throw invalidConfig(`channels[${index}].name must be unique within the Agent Capsule.`);
@@ -397,7 +452,7 @@ function validateChannels(value: unknown, runtime: AgentRuntime): void {
   }
 }
 
-function validateChannel(value: unknown, index: number): asserts value is AgentChannel {
+function validateChannel(value: unknown, index: number, hasTranscriptionConfig: boolean): asserts value is AgentChannel {
   if (!isRecord(value)) {
     throw invalidConfig(`channels[${index}] must be an object.`);
   }
@@ -422,6 +477,7 @@ function validateChannel(value: unknown, index: number): asserts value is AgentC
   validateSecretNames(value.secrets, `channels[${index}].secrets`);
   validateChannelLimits(value.limits, index);
   validateChannelBuffer(value.buffer, index);
+  validateChannelAudio(value.audio, index, hasTranscriptionConfig);
 
   if (value.type === "website") {
     validateWebsiteChannel(value, index);
@@ -436,6 +492,31 @@ function validateChannel(value: unknown, index: number): asserts value is AgentC
   validateWhatsappChannel(value, index);
 }
 
+function validateChannelAudio(value: unknown, index: number, hasTranscriptionConfig: boolean): void {
+  if (value === undefined) {
+    return;
+  }
+
+  if (!isRecord(value)) {
+    throw invalidConfig(`channels[${index}].audio must be an object when provided.`);
+  }
+
+  if (!CHANNEL_AUDIO_MODES.has(value.mode as ChannelAudioMode)) {
+    throw invalidConfig(`channels[${index}].audio.mode must be "off", "reject", or "transcribe".`);
+  }
+
+  if (value.mode === "transcribe" && !hasTranscriptionConfig) {
+    throw invalidConfig(`channels[${index}].audio.mode "transcribe" requires top-level transcription config.`);
+  }
+
+  if (value.mode !== "off") {
+    expectString(value.fallbackMessage, `channels[${index}].audio.fallbackMessage`);
+  }
+
+  validateOptionalPositiveInteger(value.maxDurationSeconds, `channels[${index}].audio.maxDurationSeconds`);
+  validateOptionalPositiveInteger(value.maxBytes, `channels[${index}].audio.maxBytes`);
+}
+
 function validateWebsiteChannel(value: Record<string, unknown>, index: number): void {
   if (value.provider !== "agentkit") {
     throw invalidConfig(`channels[${index}].provider must be "agentkit" when type is "website".`);
@@ -552,6 +633,16 @@ function validateSecretNames(secrets: string[], field: string): void {
   }
 }
 
+function validateOptionalPositiveInteger(value: unknown, field: string): void {
+  if (value === undefined) {
+    return;
+  }
+
+  if (typeof value !== "number" || !Number.isInteger(value) || value <= 0) {
+    throw invalidConfig(`${field} must be a positive integer when provided.`);
+  }
+}
+
 function expectRequiredSecrets(value: unknown, requiredSecrets: string[], field: string): void {
   const declaredSecrets = new Set(Array.isArray(value) ? value : []);
   const missingSecrets = requiredSecrets.filter((secret) => !declaredSecrets.has(secret));
@@ -636,6 +727,7 @@ function validateTool(value: unknown, index: number): asserts value is AgentTool
 
   if (value.secrets !== undefined) {
     expectStringArray(value.secrets, `tools[${index}].secrets`);
+    validateSecretNames(value.secrets, `tools[${index}].secrets`);
   }
 
   if (value.permissions !== undefined) {

package/src/runtime/core/manifest.ts

@@ -4,6 +4,7 @@ import type { AgentConfig, AgentTool } from "../../index";
 import type { LoadedAgentCapsule } from "../config";
 import { resolveKnowledgeConfig } from "../knowledge/config";
 import { KNOWLEDGE_SEARCH_TOOL_NAME } from "../knowledge/tool";
+import { resolveTranscriptionConfig } from "../transcription";
 
 export type ManifestSchema = {
   source: string;
@@ -15,6 +16,7 @@ export type SharedAgentManifest = {
   name: string;
   runtime: AgentConfig["runtime"];
   provider: AgentConfig["provider"];
+  timeZone: string | null;
   prompt: string;
   secrets: string[];
   tools: Array<{
@@ -27,6 +29,19 @@ export type SharedAgentManifest = {
     timeoutMs: number | null;
   }>;
   channels: NonNullable<AgentConfig["channels"]>;
+  transcription: {
+    enabled: boolean;
+    provider: string | null;
+    model: string | null;
+    secret: string | null;
+    language: string | null;
+    prompt: string | null;
+    limits: {
+      maxDurationSeconds: number | null;
+      maxBytes: number | null;
+    };
+    rawAudioTtlSeconds: number | null;
+  };
   knowledge: {
     enabled: boolean;
     sources: Array<{
@@ -79,10 +94,9 @@ export function buildSharedAgentManifest(
   const database = buildDatabaseManifest(capsule, options.databaseSchema ?? null);
   const files = buildFilesManifest(capsule);
   const knowledge = resolveKnowledgeConfig(capsule.config);
+  const transcription = resolveTranscriptionConfig(capsule.config.transcription);
 
-  if (capsule.config.access.mode === "private") {
-    requiredSecrets.add("AGENTKIT_PRIVATE_TOKEN");
-  }
+  requiredSecrets.add("AGENTKIT_PRIVATE_TOKEN");
 
   if (capsule.config.access.mode === "token") {
     requiredSecrets.add("AGENTKIT_ACCESS_TOKEN");
@@ -101,6 +115,10 @@ export function buildSharedAgentManifest(
     requiredSecrets.add(knowledge.embedding.secret);
   }
 
+  if (transcription.secret) {
+    requiredSecrets.add(transcription.secret);
+  }
+
   for (const tool of capsule.config.tools ?? []) {
     for (const secret of tool.secrets ?? []) {
       requiredSecrets.add(secret);
@@ -118,6 +136,7 @@ export function buildSharedAgentManifest(
     name: capsule.config.name,
     runtime: capsule.config.runtime,
     provider: capsule.config.provider,
+    timeZone: capsule.config.timeZone ?? null,
     prompt: toCapsulePath(capsule, capsule.instructionsPath),
     secrets: Array.from(requiredSecrets).sort(),
     tools: (capsule.config.tools ?? []).map((tool) => ({
@@ -130,6 +149,19 @@ export function buildSharedAgentManifest(
       timeoutMs: tool.timeoutMs ?? null,
     })),
     channels: capsule.config.channels ?? [],
+    transcription: {
+      enabled: transcription.enabled,
+      provider: transcription.enabled ? transcription.provider : null,
+      model: transcription.enabled ? transcription.model : null,
+      secret: transcription.secret,
+      language: transcription.language,
+      prompt: transcription.prompt,
+      limits: {
+        maxDurationSeconds: transcription.limits.maxDurationSeconds ?? null,
+        maxBytes: transcription.limits.maxBytes ?? null,
+      },
+      rawAudioTtlSeconds: transcription.rawAudioTtlSeconds,
+    },
     knowledge: {
       enabled: knowledge.enabled,
       sources: knowledge.sources.map((source) => ({

package/src/runtime/deploy-readiness.ts

@@ -85,8 +85,8 @@ export async function loadDeployReadinessContext(
   };
 }
 
-export function stableCloudProjectId(name: string): string {
-  return `proj_${createHash("sha256").update(name).digest("hex").slice(0, 16)}`;
+export function stableCloudProjectId(name: string, accountId?: string | null): string {
+  return `proj_${createHash("sha256").update(`${accountId ? `account:${accountId}` : "anonymous"}:${name}`).digest("hex").slice(0, 16)}`;
 }
 
 function isAgentKitManagedDeploySecret(
@@ -97,7 +97,7 @@ function isAgentKitManagedDeploySecret(
     return true;
   }
 
-  if (context.accessMode === "private" && name === "AGENTKIT_PRIVATE_TOKEN") {
+  if (name === "AGENTKIT_PRIVATE_TOKEN") {
     return true;
   }