Merge branch 'fix/character-voice-settings' of https://github.com/augchan42/eliza into HEAD

lalalune · lalalune · commit 84466c483e95 · 2024-12-01T16:18:21.000-08:00
diff --git a/package.json b/package.json
@@ -2,7 +2,7 @@
     "name": "eliza",
     "scripts": {
         "preinstall": "npx only-allow pnpm",
-        "build": "turbo check-types build",
+        "build": "turbo run build --filter=@ai16z/plugin-bootstrap^... --filter=@ai16z/eliza^...",
         "start": "pnpm --filter \"@ai16z/agent\" start --isRoot",
         "start:client": "pnpm --dir client start --isRoot",
         "start:debug": "cross-env NODE_ENV=development VERBOSE=true DEBUG=eliza:* pnpm --filter \"@ai16z/agent\" start --isRoot",
@@ -63,4 +63,4 @@
     "workspaces": [
         "packages/*"
     ]
-}
+}
diff --git a/packages/adapter-postgres/src/index.ts b/packages/adapter-postgres/src/index.ts
@@ -1,8 +1,8 @@
 import { v4 } from "uuid";
 
-// Import the entire module as default
-import pg from "pg";
-type Pool = pg.Pool;
+import postgres from "pg";
+const { Pool } = postgres;
+type PoolType = typeof postgres.Pool;
 
 import {
     QueryConfig,
@@ -32,10 +32,10 @@ const __filename = fileURLToPath(import.meta.url); // get the resolved path to t
 const __dirname = path.dirname(__filename); // get the name of the directory
 
 export class PostgresDatabaseAdapter
-    extends DatabaseAdapter<Pool>
+    extends DatabaseAdapter<PoolType>
     implements IDatabaseCacheAdapter
 {
-    private pool: Pool;
+    private pool: InstanceType<PoolType>;
     private readonly maxRetries: number = 3;
     private readonly baseDelay: number = 1000; // 1 second
     private readonly maxDelay: number = 10000; // 10 seconds
@@ -51,7 +51,7 @@ export class PostgresDatabaseAdapter
             connectionTimeoutMillis: this.connectionTimeout,
         };
 
-        this.pool = new pg.Pool({
+        this.pool = new Pool({
             ...defaultConfig,
             ...connectionConfig, // Allow overriding defaults
         });
@@ -137,7 +137,7 @@ export class PostgresDatabaseAdapter
             await this.pool.end();
 
             // Create new pool
-            this.pool = new pg.Pool({
+            this.pool = new Pool({
                 ...this.pool.options,
                 connectionTimeoutMillis: this.connectionTimeout,
             });
@@ -174,12 +174,33 @@ export class PostgresDatabaseAdapter
     async init() {
         await this.testConnection();
 
-        const schema = fs.readFileSync(
-            path.resolve(__dirname, "../schema.sql"),
-            "utf8"
-        );
+        const client = await this.pool.connect();
+        try {
+            await client.query("BEGIN");
+
+            // Check if schema already exists (check for a core table)
+            const { rows } = await client.query(`
+                SELECT EXISTS (
+                    SELECT FROM information_schema.tables
+                    WHERE table_name = 'rooms'
+                );
+            `);
+
+            if (!rows[0].exists) {
+                const schema = fs.readFileSync(
+                    path.resolve(__dirname, "../schema.sql"),
+                    "utf8"
+                );
+                await client.query(schema);
+            }
 
-        await this.query(schema);
+            await client.query("COMMIT");
+        } catch (error) {
+            await client.query("ROLLBACK");
+            throw error;
+        } finally {
+            client.release();
+        }
     }
 
     async close() {
diff --git a/packages/core/src/types.ts b/packages/core/src/types.ts
@@ -680,8 +680,17 @@ export type Character = {
         secrets?: { [key: string]: string };
         buttplug?: boolean;
         voice?: {
-            model?: string;
-            url?: string;
+            model?: string; // For VITS
+            url?: string; // Legacy VITS support
+            elevenlabs?: {
+                // New structured ElevenLabs config
+                voiceId: string;
+                model?: string;
+                stability?: string;
+                similarityBoost?: string;
+                style?: string;
+                useSpeakerBoost?: string;
+            };
         };
         model?: string;
         embeddingModel?: string;
diff --git a/packages/plugin-node/src/enviroment.ts b/packages/plugin-node/src/enviroment.ts
@@ -3,27 +3,21 @@ import { z } from "zod";
 
 export const nodeEnvSchema = z.object({
     OPENAI_API_KEY: z.string().min(1, "OpenAI API key is required"),
+
+    // Core settings
     ELEVENLABS_XI_API_KEY: z.string().optional(),
-    ELEVENLABS_MODEL_ID: z.string().min(1, "ElevenLabs model ID is required"),
-    ELEVENLABS_VOICE_ID: z.string().min(1, "ElevenLabs voice ID is required"),
-    ELEVENLABS_VOICE_STABILITY: z
-        .string()
-        .min(1, "ElevenLabs voice stability is required"),
-    ELEVENLABS_VOICE_SIMILARITY_BOOST: z
-        .string()
-        .min(1, "ElevenLabs voice similarity boost is required"),
-    ELEVENLABS_VOICE_STYLE: z
-        .string()
-        .min(1, "ElevenLabs voice style is required"),
-    ELEVENLABS_VOICE_USE_SPEAKER_BOOST: z
-        .string()
-        .min(1, "ElevenLabs voice speaker boost setting is required"),
-    ELEVENLABS_OPTIMIZE_STREAMING_LATENCY: z
-        .string()
-        .min(1, "ElevenLabs streaming latency optimization is required"),
-    ELEVENLABS_OUTPUT_FORMAT: z
-        .string()
-        .min(1, "ElevenLabs output format is required"),
+
+    // All other settings optional with defaults
+    ELEVENLABS_MODEL_ID: z.string().optional(),
+    ELEVENLABS_VOICE_ID: z.string().optional(),
+    ELEVENLABS_VOICE_STABILITY: z.string().optional(),
+    ELEVENLABS_VOICE_SIMILARITY_BOOST: z.string().optional(),
+    ELEVENLABS_VOICE_STYLE: z.string().optional(),
+    ELEVENLABS_VOICE_USE_SPEAKER_BOOST: z.string().optional(),
+    ELEVENLABS_OPTIMIZE_STREAMING_LATENCY: z.string().optional(),
+    ELEVENLABS_OUTPUT_FORMAT: z.string().optional(),
+    VITS_VOICE: z.string().optional(),
+    VITS_MODEL: z.string().optional(),
 });
 
 export type NodeConfig = z.infer<typeof nodeEnvSchema>;
@@ -32,34 +26,51 @@ export async function validateNodeConfig(
     runtime: IAgentRuntime
 ): Promise<NodeConfig> {
     try {
+        const voiceSettings = runtime.character.settings?.voice;
+        const elevenlabs = voiceSettings?.elevenlabs;
+
+        // Only include what's absolutely required
         const config = {
             OPENAI_API_KEY:
                 runtime.getSetting("OPENAI_API_KEY") ||
                 process.env.OPENAI_API_KEY,
-            ELEVENLABS_MODEL_ID:
-                runtime.getSetting("ELEVENLABS_MODEL_ID") ||
-                process.env.ELEVENLABS_MODEL_ID,
-            ELEVENLABS_VOICE_ID:
-                runtime.getSetting("ELEVENLABS_VOICE_ID") ||
-                process.env.ELEVENLABS_VOICE_ID,
-            ELEVENLABS_VOICE_STABILITY:
-                runtime.getSetting("ELEVENLABS_VOICE_STABILITY") ||
-                process.env.ELEVENLABS_VOICE_STABILITY,
-            ELEVENLABS_VOICE_SIMILARITY_BOOST:
-                runtime.getSetting("ELEVENLABS_VOICE_SIMILARITY_BOOST") ||
-                process.env.ELEVENLABS_VOICE_SIMILARITY_BOOST,
-            ELEVENLABS_VOICE_STYLE:
-                runtime.getSetting("ELEVENLABS_VOICE_STYLE") ||
-                process.env.ELEVENLABS_VOICE_STYLE,
-            ELEVENLABS_VOICE_USE_SPEAKER_BOOST:
-                runtime.getSetting("ELEVENLABS_VOICE_USE_SPEAKER_BOOST") ||
-                process.env.ELEVENLABS_VOICE_USE_SPEAKER_BOOST,
-            ELEVENLABS_OPTIMIZE_STREAMING_LATENCY:
-                runtime.getSetting("ELEVENLABS_OPTIMIZE_STREAMING_LATENCY") ||
-                process.env.ELEVENLABS_OPTIMIZE_STREAMING_LATENCY,
-            ELEVENLABS_OUTPUT_FORMAT:
-                runtime.getSetting("ELEVENLABS_OUTPUT_FORMAT") ||
-                process.env.ELEVENLABS_OUTPUT_FORMAT,
+            ELEVENLABS_XI_API_KEY:
+                runtime.getSetting("ELEVENLABS_XI_API_KEY") ||
+                process.env.ELEVENLABS_XI_API_KEY,
+
+            // Use character card settings first, fall back to env vars, then defaults
+            ...(runtime.getSetting("ELEVENLABS_XI_API_KEY") && {
+                ELEVENLABS_MODEL_ID:
+                    elevenlabs?.model ||
+                    process.env.ELEVENLABS_MODEL_ID ||
+                    "eleven_monolingual_v1",
+                ELEVENLABS_VOICE_ID:
+                    elevenlabs?.voiceId || process.env.ELEVENLABS_VOICE_ID,
+                ELEVENLABS_VOICE_STABILITY:
+                    elevenlabs?.stability ||
+                    process.env.ELEVENLABS_VOICE_STABILITY ||
+                    "0.5",
+                ELEVENLABS_VOICE_SIMILARITY_BOOST:
+                    elevenlabs?.similarityBoost ||
+                    process.env.ELEVENLABS_VOICE_SIMILARITY_BOOST ||
+                    "0.75",
+                ELEVENLABS_VOICE_STYLE:
+                    elevenlabs?.style ||
+                    process.env.ELEVENLABS_VOICE_STYLE ||
+                    "0",
+                ELEVENLABS_VOICE_USE_SPEAKER_BOOST:
+                    elevenlabs?.useSpeakerBoost ||
+                    process.env.ELEVENLABS_VOICE_USE_SPEAKER_BOOST ||
+                    "true",
+                ELEVENLABS_OPTIMIZE_STREAMING_LATENCY:
+                    process.env.ELEVENLABS_OPTIMIZE_STREAMING_LATENCY || "0",
+                ELEVENLABS_OUTPUT_FORMAT:
+                    process.env.ELEVENLABS_OUTPUT_FORMAT || "pcm_16000",
+            }),
+
+            // VITS settings
+            VITS_VOICE: voiceSettings?.model || process.env.VITS_VOICE,
+            VITS_MODEL: process.env.VITS_MODEL,
         };
 
         return nodeEnvSchema.parse(config);
diff --git a/packages/plugin-node/src/services/speech.ts b/packages/plugin-node/src/services/speech.ts
diff --git a/packages/plugin-node/src/vendor/vitsVoiceList.ts b/packages/plugin-node/src/vendor/vitsVoiceList.ts
