Merge pull request #496 from ai16z/feat/embeddings

fix: improve embeddings

Author: ponderingdemocritus

.github/workflows/pre-release.yml

@@ -0,0 +1,75 @@
+name: Release
+
+on:
+    push:
+        branches:
+            - main
+    workflow_dispatch:
+        inputs:
+            release_type:
+                description: "Type of release (prerelease, prepatch, patch, minor, preminor, major)"
+                required: true
+                default: "prerelease"
+
+jobs:
+    release:
+        runs-on: ubuntu-latest
+        steps:
+            - uses: actions/checkout@v4
+              with:
+                  fetch-depth: 0
+
+            - uses: pnpm/action-setup@v3
+              with:
+                  version: 8
+
+            - name: Configure Git
+              run: |
+                  git config user.name "${{ github.actor }}"
+                  git config user.email "${{ github.actor }}@users.noreply.github.com"
+
+            - name: "Setup npm for npmjs"
+              run: |
+                  npm config set registry https://registry.npmjs.org/
+                  echo "//registry.npmjs.org/:_authToken=${{ secrets.NPM_TOKEN }}" > ~/.npmrc
+
+            - name: Install Protobuf Compiler
+              run: sudo apt-get install -y protobuf-compiler
+
+            - name: Install dependencies
+              run: pnpm install
+
+            - name: Build packages
+              run: pnpm run build
+
+            - name: Tag and Publish Packages
+              id: tag_publish
+              run: |
+                  RELEASE_TYPE=${{ github.event_name == 'push' && 'prerelease' || github.event.inputs.release_type }}
+                  npx lerna version $RELEASE_TYPE --conventional-commits --yes --no-private --force-publish
+                  npx lerna publish from-git --yes --dist-tag next
+
+            - name: Get Version Tag
+              id: get_tag
+              run: echo "TAG=$(git describe --tags --abbrev=0)" >> $GITHUB_OUTPUT
+
+            - name: Generate Release Body
+              id: release_body
+              run: |
+                  if [ -f CHANGELOG.md ]; then
+                    echo "body=$(cat CHANGELOG.md)" >> $GITHUB_OUTPUT
+                  else
+                    echo "body=No changelog provided for this release." >> $GITHUB_OUTPUT
+                  fi
+
+            - name: Create GitHub Release
+              uses: actions/create-release@v1
+              env:
+                  GITHUB_TOKEN: ${{ secrets.GH_TOKEN }}
+                  PNPM_HOME: /home/runner/setup-pnpm/node_modules/.bin
+              with:
+                  tag_name: ${{ steps.get_tag.outputs.TAG }}
+                  release_name: Release
+                  body_path: CHANGELOG.md
+                  draft: false
+                  prerelease: ${{ github.event_name == 'push' }}

packages/adapter-postgres/src/index.ts

@@ -11,6 +11,7 @@ import {
     type IDatabaseCacheAdapter,
     Participant,
     DatabaseAdapter,
+    elizaLogger,
 } from "@ai16z/eliza";
 import fs from "fs";
 import { fileURLToPath } from "url";
@@ -28,15 +29,50 @@ export class PostgresDatabaseAdapter
     constructor(connectionConfig: any) {
         super();
 
-        this.pool = new pg.Pool({
-            ...connectionConfig,
+        const defaultConfig = {
             max: 20,
             idleTimeoutMillis: 30000,
             connectionTimeoutMillis: 2000,
+        };
+
+        this.pool = new pg.Pool({
+            ...defaultConfig,
+            ...connectionConfig, // Allow overriding defaults
         });
 
-        this.pool.on("error", (err) => {
-            console.error("Unexpected error on idle client", err);
+        this.pool.on("error", async (err) => {
+            elizaLogger.error("Unexpected error on idle client", err);
+
+            // Attempt to reconnect with exponential backoff
+            let retryCount = 0;
+            const maxRetries = 5;
+            const baseDelay = 1000; // Start with 1 second delay
+
+            while (retryCount < maxRetries) {
+                try {
+                    const delay = baseDelay * Math.pow(2, retryCount);
+                    elizaLogger.log(`Attempting to reconnect in ${delay}ms...`);
+                    await new Promise((resolve) => setTimeout(resolve, delay));
+
+                    // Create new pool with same config
+                    this.pool = new pg.Pool(this.pool.options);
+                    await this.testConnection();
+
+                    elizaLogger.log("Successfully reconnected to database");
+                    return;
+                } catch (error) {
+                    retryCount++;
+                    elizaLogger.error(
+                        `Reconnection attempt ${retryCount} failed:`,
+                        error
+                    );
+                }
+            }
+
+            elizaLogger.error(
+                `Failed to reconnect after ${maxRetries} attempts`
+            );
+            throw new Error("Database connection lost and unable to reconnect");
         });
     }
 
@@ -51,7 +87,7 @@ export class PostgresDatabaseAdapter
             );
             await client.query(schema);
         } catch (error) {
-            console.error(error);
+            elizaLogger.error(error);
             throw error;
         }
     }
@@ -61,10 +97,13 @@ export class PostgresDatabaseAdapter
         try {
             client = await this.pool.connect();
             const result = await client.query("SELECT NOW()");
-            console.log("Database connection test successful:", result.rows[0]);
+            elizaLogger.log(
+                "Database connection test successful:",
+                result.rows[0]
+            );
             return true;
         } catch (error) {
-            console.error("Database connection test failed:", error);
+            elizaLogger.error("Database connection test failed:", error);
             throw new Error(`Failed to connect to database: ${error.message}`);
         } finally {
             if (client) client.release();
@@ -187,7 +226,7 @@ export class PostgresDatabaseAdapter
             if (rows.length === 0) return null;
 
             const account = rows[0];
-            console.log("account", account);
+            elizaLogger.log("account", account);
             return {
                 ...account,
                 details:
@@ -217,7 +256,7 @@ export class PostgresDatabaseAdapter
             );
             return true;
         } catch (error) {
-            console.log("Error creating account", error);
+            elizaLogger.log("Error creating account", error);
             return false;
         } finally {
             client.release();
@@ -370,7 +409,7 @@ export class PostgresDatabaseAdapter
                 values.push(params.count);
             }
 
-            console.log("sql", sql, values);
+            elizaLogger.log("sql", sql, values);
 
             const { rows } = await client.query(sql, values);
             return rows.map((row) => ({

packages/client-discord/src/actions/summarize_conversation.ts

@@ -251,12 +251,7 @@ const summarizeAction = {
         const model = models[runtime.character.settings.model];
         const chunkSize = model.settings.maxContextLength - 1000;
 
-        const chunks = await splitChunks(
-            formattedMemories,
-            chunkSize,
-            "gpt-4o-mini",
-            0
-        );
+        const chunks = await splitChunks(formattedMemories, chunkSize, 0);
 
         const datestr = new Date().toUTCString().replace(/:/g, "-");
 

packages/core/package.json

@@ -70,6 +70,7 @@
         "gaxios": "6.7.1",
         "glob": "11.0.0",
         "js-sha1": "0.7.0",
+        "langchain": "^0.3.6",
         "ollama-ai-provider": "^0.16.1",
         "openai": "4.69.0",
         "tiktoken": "1.0.17",

packages/core/src/embedding.ts

@@ -37,7 +37,7 @@ async function getRemoteEmbedding(
                 : {}),
         },
         body: JSON.stringify({
-            input,
+            input: trimTokens(input, 8191, "gpt-4o-mini"),
             model: options.model,
             length: options.length || 384,
         }),
@@ -70,25 +70,39 @@ async function getRemoteEmbedding(
  * @param input The input to be embedded.
  * @returns The embedding of the input.
  */
+/**
+ * Generate embeddings for input text using configured model provider
+ * @param runtime The agent runtime containing model configuration
+ * @param input The text to generate embeddings for
+ * @returns Array of embedding numbers
+ */
 export async function embed(runtime: IAgentRuntime, input: string) {
+    // Get model provider configuration
     const modelProvider = models[runtime.character.modelProvider];
-    //need to have env override for this to select what to use for embedding if provider doesnt provide or using openai
+
+    // Determine which embedding model to use:
+    // 1. OpenAI if USE_OPENAI_EMBEDDING is true
+    // 2. Provider's own embedding model if available
+    // 3. Fallback to OpenAI embedding model
     const embeddingModel = settings.USE_OPENAI_EMBEDDING
-        ? "text-embedding-3-small" // Use OpenAI if specified
-        : modelProvider.model?.[ModelClass.EMBEDDING] || // Use provider's embedding model if available
-          models[ModelProviderName.OPENAI].model[ModelClass.EMBEDDING]; // Fallback to OpenAI
+        ? "text-embedding-3-small"
+        : modelProvider.model?.[ModelClass.EMBEDDING] ||
+          models[ModelProviderName.OPENAI].model[ModelClass.EMBEDDING];
 
     if (!embeddingModel) {
         throw new Error("No embedding model configured");
     }
 
-    // // Try local embedding first
-    // Check if we're in Node.js environment
+    // Check if running in Node.js environment
     const isNode =
         typeof process !== "undefined" &&
         process.versions != null &&
         process.versions.node != null;
 
+    // Use local embedding if:
+    // - Running in Node.js
+    // - Not using OpenAI provider
+    // - Not forcing OpenAI embeddings
     if (
         isNode &&
         runtime.character.modelProvider !== ModelProviderName.OPENAI &&
@@ -97,28 +111,30 @@ export async function embed(runtime: IAgentRuntime, input: string) {
         return await getLocalEmbedding(input);
     }
 
-    // Check cache
+    // Try to get cached embedding first
     const cachedEmbedding = await retrieveCachedEmbedding(runtime, input);
     if (cachedEmbedding) {
         return cachedEmbedding;
     }
 
-    // Get remote embedding
+    // Generate new embedding remotely
     return await getRemoteEmbedding(input, {
         model: embeddingModel,
+        // Use OpenAI endpoint if specified, otherwise use provider endpoint
         endpoint: settings.USE_OPENAI_EMBEDDING
-            ? "https://api.openai.com/v1" // Always use OpenAI endpoint when USE_OPENAI_EMBEDDING is true
+            ? "https://api.openai.com/v1"
             : runtime.character.modelEndpointOverride || modelProvider.endpoint,
+        // Use OpenAI API key if specified, otherwise use runtime token
         apiKey: settings.USE_OPENAI_EMBEDDING
-            ? settings.OPENAI_API_KEY // Use OpenAI key from settings when USE_OPENAI_EMBEDDING is true
-            : runtime.token, // Use runtime token for other providers
+            ? settings.OPENAI_API_KEY
+            : runtime.token,
+        // Special handling for Ollama provider
         isOllama:
             runtime.character.modelProvider === ModelProviderName.OLLAMA &&
             !settings.USE_OPENAI_EMBEDDING,
     });
 }
 
-//  TODO: Add back in when it can work in browser and locally
 async function getLocalEmbedding(input: string): Promise<number[]> {
     // Check if we're in Node.js environment
     const isNode =
@@ -153,7 +169,7 @@ async function getLocalEmbedding(input: string): Promise<number[]> {
             cacheDir: cacheDir,
         });
 
-        const trimmedInput = trimTokens(input, 8000, "gpt-4o-mini");
+        const trimmedInput = trimTokens(input, 8191, "gpt-4o-mini");
         const embedding = await embeddingModel.queryEmbed(trimmedInput);
         return embedding;
     } else {