doc2vec 2.4.0 → 2.6.0

package/README.md

@@ -36,6 +36,10 @@ The primary goal is to prepare documentation content for Retrieval-Augmented Gen
 *   **Vector Storage:** Supports storing chunks, metadata, and embeddings in:
     *   **SQLite:** Using `better-sqlite3` and the `sqlite-vec` extension for efficient vector search.
     *   **Qdrant:** A dedicated vector database, using the `@qdrant/js-client-rest`.
+*   **[Postgres Markdown Store](MARKDOWN_STORE.md):** Optionally stores the generated markdown for each crawled URL in a Postgres table. Useful for maintaining a searchable, raw-text copy of all documentation pages alongside the vector embeddings.
+    *   **Automatic population:** On the first sync, all pages are force-processed (bypassing lastmod/ETag caching) to fully populate the store. Subsequent syncs only update rows when a change is detected.
+    *   **404 cleanup:** Pages that return 404 are automatically removed from the store.
+    *   **Shared table:** A single table (configurable name, default `markdown_pages`) is shared across all sources, with a `product_name` column to distinguish them.
 *   **Multi-Layer Change Detection:** Four layers of change detection minimize unnecessary re-processing:
     1. **Sitemap `lastmod`:** When available, compares the sitemap's `<lastmod>` date against the stored value — skips without any HTTP request. Child URLs inherit `lastmod` from their most specific parent directory.
     2. **ETag via HEAD request:** For URLs without `lastmod`, sends a lightweight HEAD request and compares the ETag header against the stored value. Adaptive backoff prevents rate limiting (starts at 0ms delay, increases on 429 responses, decays on success).
@@ -43,6 +47,8 @@ The primary goal is to prepare documentation content for Retrieval-Augmented Gen
     4. **Embedding:** Only re-embeds chunks when content has actually changed.
     
     ETag and lastmod values are only stored when chunking and embedding succeed, ensuring failed pages are retried on the next run.
+    
+    A `sync_complete` metadata flag tracks whether a full sync has ever completed successfully. If a sync is interrupted (process killed), the next run force-processes all pages regardless of lastmod/ETag values, ensuring no pages are permanently skipped.
 *   **Incremental Updates:** For GitHub and Zendesk sources, tracks the last run date to only fetch new or updated issues/tickets.
 *   **Cleanup:** Removes obsolete chunks from the database corresponding to pages or files that are no longer found during processing.
 *   **Configuration:** Driven by a YAML configuration file (`config.yaml`) specifying sites, repositories, local directories, Zendesk instances, database types, metadata, and other parameters.
@@ -145,6 +151,9 @@ Configuration is managed through two files:
     OPENAI_API_KEY="sk-..."
     OPENAI_MODEL="text-embedding-3-large"  # Optional, defaults to text-embedding-3-large
 
+    # Optional: Embedding dimension size (defaults to 3072)
+    EMBEDDING_DIMENSION="3072"
+
     # Required: Your Azure OpenAI credentials (if using Azure provider)
     AZURE_OPENAI_KEY="your-azure-key"
     AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com"
@@ -172,6 +181,7 @@ Configuration is managed through two files:
         For websites (`type: 'website'`):
         *   `url`: The starting URL for crawling the documentation site.
         *   `sitemap_url`: (Optional) URL to the site's XML sitemap for discovering additional pages not linked in navigation.
+        *   `markdown_store`: (Optional) Set to `true` to store generated markdown in the Postgres markdown store (requires top-level `markdown_store` config). Defaults to `false`.
         
         For GitHub repositories (`type: 'github'`):
         *   `repo`: Repository name in the format `'owner/repo'` (e.g., `'istio/istio'`).
@@ -223,6 +233,21 @@ Configuration is managed through two files:
                     *   `qdrant_port`: (Optional) Port for the Qdrant REST API. Defaults to `443` if `qdrant_url` starts with `https`, otherwise `6333`.
                     *   `collection_name`: (Optional) Name of the Qdrant collection to use. Defaults to `<product_name>_<version>` (lowercased, spaces replaced with underscores).
 
+        Optional embedding configuration:
+        *   `embedding.provider`: Provider for embeddings (`openai` or `azure`).
+        *   `embedding.dimension`: Embedding vector size. Defaults to `3072` when not set.
+
+        Optional Postgres markdown store (top-level):
+        *   `markdown_store.connection_string`: (Optional) Full Postgres connection string (e.g., `'postgres://user:pass@host:5432/db'`). Takes priority over individual fields.
+        *   `markdown_store.host`: (Optional) Postgres host.
+        *   `markdown_store.port`: (Optional) Postgres port.
+        *   `markdown_store.database`: (Optional) Postgres database name.
+        *   `markdown_store.user`: (Optional) Postgres user.
+        *   `markdown_store.password`: (Optional) Postgres password. Supports `${PG_PASSWORD}` env var substitution.
+        *   `markdown_store.table_name`: (Optional) Table name. Defaults to `'markdown_pages'`.
+        
+        When configured, website sources with `markdown_store: true` will store the generated markdown for each URL in this Postgres table. On the first sync, all pages are force-processed (bypassing lastmod/ETag skip logic) to populate the table. On subsequent syncs, only pages with detected changes get their rows updated.
+
     **Example (`config.yaml`):**
     ```yaml
     # Optional: Configure embedding provider
@@ -230,6 +255,7 @@ Configuration is managed through two files:
     # Defaults to OpenAI if not specified
     embedding:
       provider: 'openai'  # or 'azure'
+      dimension: 3072  # Optional, defaults to 3072
       openai:
         api_key: '${OPENAI_API_KEY}'  # Optional, uses env var by default
         model: 'text-embedding-3-large'  # Optional, defaults to text-embedding-3-large
@@ -240,13 +266,25 @@ Configuration is managed through two files:
       #   deployment_name: 'text-embedding-3-large'
       #   api_version: '2024-10-21'  # Optional
 
+    # Optional: Store generated markdown in Postgres
+    # markdown_store:
+    #   connection_string: 'postgres://user:pass@host:5432/db'
+    #   # OR use individual fields:
+    #   # host: 'localhost'
+    #   # port: 5432
+    #   # database: 'doc2vec'
+    #   # user: 'myuser'
+    #   # password: '${PG_PASSWORD}'
+    #   # table_name: 'markdown_pages'  # Optional, defaults to 'markdown_pages'
+
     sources:
-      # Website source example
+      # Website source example (with markdown store enabled)
       - type: 'website'
         product_name: 'argo'
         version: 'stable'
         url: 'https://argo-cd.readthedocs.io/en/stable/'
         sitemap_url: 'https://argo-cd.readthedocs.io/en/stable/sitemap.xml'
+        markdown_store: true  # Store generated markdown in Postgres
         max_size: 1048576
         database_config:
           type: 'sqlite'
@@ -540,6 +578,20 @@ If you don't specify a config path, it will look for config.yaml in the current
 
 ## Recent Changes
 
+### Postgres Markdown Store
+- **New feature:** Optionally store the generated markdown for each crawled website URL in a Postgres table (`url`, `product_name`, `markdown`, `updated_at`)
+- **Top-level configuration:** Configure Postgres connection once at the top level via `connection_string` or individual `host`/`port`/`database`/`user`/`password` fields, with environment variable substitution support
+- **Per-source opt-in:** Enable per website source with `markdown_store: true` (disabled by default)
+- **First-sync force-processing:** When the markdown store is enabled, pages that aren't yet in the Postgres table bypass lastmod and ETag skip logic, ensuring all pages are stored on the first sync
+- **Change-only updates:** On subsequent syncs, only pages with detected content changes (via lastmod/ETag) have their Postgres rows updated
+- **404 cleanup:** Pages that return 404 during HEAD checks are automatically removed from the Postgres store
+
+### Incomplete Sync Recovery
+- **New feature:** Tracks whether a full sync has ever completed successfully for each website source via a `sync_complete:<url_prefix>` metadata key
+- **Interrupted sync handling:** If a sync is killed mid-crawl (process terminated, crash, etc.), the stored ETags/lastmods from the partial run would otherwise cause remaining pages to be skipped permanently. The `sync_complete` flag prevents this — when absent, all pages are force-processed regardless of caching signals
+- **Gated on clean completion:** The flag is only set when the crawl completes without network errors (DNS failures, connection refused, timeouts). If the site is unreachable, the next run will force a full sync again
+- **Scoped per source:** Each website source has its own `sync_complete` key based on its URL prefix. Changing the source URL naturally triggers a new full sync
+
 ### Multi-Layer Change Detection for Websites
 - **Sitemap `lastmod` support:** When a sitemap includes `<lastmod>` dates, pages are skipped entirely if the date hasn't changed — no HEAD request, no Puppeteer load, no chunking. One sitemap fetch replaces hundreds of individual HEAD requests.
 - **`lastmod` inheritance:** Child URLs without their own `<lastmod>` inherit from the most specific parent directory URL in the sitemap (e.g., `/docs/2.10.x/reference/cli/` inherits from `/docs/2.10.x/`).

package/dist/content-processor.js

@@ -218,6 +218,7 @@ class ContentProcessor {
         const queue = [baseUrl];
         const brokenLinks = [];
         const brokenLinkKeys = new Set();
+        const notFoundUrls = new Set();
         const referrers = new Map();
         const addReferrer = (targetUrl, sourceUrl) => {
             const existing = referrers.get(targetUrl);
@@ -236,7 +237,7 @@ class ContentProcessor {
             brokenLinks.push({ source: sourceUrl, target: targetUrl });
         };
         addReferrer(baseUrl, baseUrl);
-        const { knownUrls, etagStore, lastmodStore } = options ?? {};
+        const { knownUrls, etagStore, lastmodStore, markdownStoreUrls, forceFullSync } = options ?? {};
         // Pre-seed queue with previously-known URLs so link discovery isn't lost
         // when pages are skipped via ETag/lastmod matching on re-runs
         if (knownUrls && knownUrls.size > 0) {
@@ -422,14 +423,26 @@ class ContentProcessor {
                 // this URL and it matches the stored value, skip entirely. This avoids
                 // both HEAD requests and full page loads — the sitemap is fetched once
                 // for the entire crawl.
+                // Bypass this skip when:
+                //   - forceFullSync is true (previous sync never completed)
+                //   - markdownStoreUrls is set and the URL is not in it (markdown store gap)
                 const urlLastmod = sitemapLastmod.get(url);
+                const urlInMarkdownStore = !markdownStoreUrls || markdownStoreUrls.has(url);
                 if (urlLastmod && lastmodStore) {
                     const storedLastmod = await lastmodStore.get(url);
                     if (storedLastmod && storedLastmod === urlLastmod) {
-                        logger.info(`Lastmod unchanged (${urlLastmod}), skipping: ${url}`);
-                        visitedUrls.add(url);
-                        lastmodSkippedCount++;
-                        continue;
+                        if (forceFullSync) {
+                            logger.info(`Lastmod unchanged but full sync not yet completed, forcing processing: ${url}`);
+                        }
+                        else if (!urlInMarkdownStore) {
+                            logger.info(`Lastmod unchanged but URL not in markdown store, forcing processing: ${url}`);
+                        }
+                        else {
+                            logger.info(`Lastmod unchanged (${urlLastmod}), skipping: ${url}`);
+                            visitedUrls.add(url);
+                            lastmodSkippedCount++;
+                            continue;
+                        }
                     }
                     else if (storedLastmod) {
                         logger.debug(`Lastmod changed (stored: ${storedLastmod}, sitemap: ${urlLastmod}), processing: ${url}`);
@@ -494,10 +507,21 @@ class ContentProcessor {
                         decayBackoff();
                         const result = await checkEtag(headResponse.headers['etag']);
                         if (result === 'skip') {
-                            logger.info(`ETag unchanged, skipping: ${url}`);
-                            visitedUrls.add(url);
-                            etagSkippedCount++;
-                            continue;
+                            // Bypass ETag skip when:
+                            //   - forceFullSync is true (previous sync never completed)
+                            //   - markdownStoreUrls is set and URL is not in it
+                            if (forceFullSync) {
+                                logger.info(`ETag unchanged but full sync not yet completed, forcing processing: ${url}`);
+                            }
+                            else if (!urlInMarkdownStore) {
+                                logger.info(`ETag unchanged but URL not in markdown store, forcing processing: ${url}`);
+                            }
+                            else {
+                                logger.info(`ETag unchanged, skipping: ${url}`);
+                                visitedUrls.add(url);
+                                etagSkippedCount++;
+                                continue;
+                            }
                         }
                         else if (result === 'no-etag') {
                             consecutiveMissingEtag++;
@@ -524,10 +548,18 @@ class ContentProcessor {
                                 decayBackoff();
                                 const result = await checkEtag(retryResponse.headers['etag']);
                                 if (result === 'skip') {
-                                    logger.info(`ETag unchanged, skipping: ${url}`);
-                                    visitedUrls.add(url);
-                                    etagSkippedCount++;
-                                    continue;
+                                    if (forceFullSync) {
+                                        logger.info(`ETag unchanged but full sync not yet completed, forcing processing: ${url}`);
+                                    }
+                                    else if (!urlInMarkdownStore) {
+                                        logger.info(`ETag unchanged but URL not in markdown store, forcing processing: ${url}`);
+                                    }
+                                    else {
+                                        logger.info(`ETag unchanged, skipping: ${url}`);
+                                        visitedUrls.add(url);
+                                        etagSkippedCount++;
+                                        continue;
+                                    }
                                 }
                                 // 'process' or 'no-etag' — fall through to full processing
                             }
@@ -542,6 +574,9 @@ class ContentProcessor {
                             // 405 (Method Not Allowed) is excluded because the server
                             // may not support HEAD but the page may still exist via GET.
                             logger.debug(`HEAD returned ${headStatus} for ${url}, skipping`);
+                            if (headStatus === 404) {
+                                notFoundUrls.add(url);
+                            }
                             skippedCount++;
                             continue;
                         }
@@ -586,7 +621,7 @@ class ContentProcessor {
                         logger.debug(`Finding links on page ${url}`);
                         let newLinksFound = 0;
                         for (const href of result.links) {
-                            const fullUrl = utils_1.Utils.buildUrl(href, pageUrlForLinks);
+                            const fullUrl = utils_1.Utils.buildUrl(href, pageUrlForLinks, logger);
                             if (fullUrl.startsWith(sourceConfig.url)) {
                                 addReferrer(fullUrl, pageUrlForLinks);
                                 if (!visitedUrls.has(utils_1.Utils.normalizeUrl(fullUrl))) {
@@ -707,7 +742,7 @@ class ContentProcessor {
         if (hasNetworkErrors) {
             logger.warn('Network errors were encountered during crawling. Cleanup may be skipped to avoid removing valid chunks.');
         }
-        return { hasNetworkErrors, brokenLinks };
+        return { hasNetworkErrors, brokenLinks, notFoundUrls };
     }
     /**
      * Detects Chrome DevTools Protocol-level errors that indicate the browser

package/dist/database.js

@@ -44,7 +44,7 @@ const sqliteVec = __importStar(require("sqlite-vec"));
 const js_client_rest_1 = require("@qdrant/js-client-rest");
 const utils_1 = require("./utils");
 class DatabaseManager {
-    static async initDatabase(config, parentLogger) {
+    static async initDatabase(config, parentLogger, embeddingDimension) {
         const logger = parentLogger.child('database');
         const dbConfig = config.database_config;
         if (dbConfig.type === 'sqlite') {
@@ -53,10 +53,10 @@ class DatabaseManager {
             logger.info(`Opening SQLite database at ${dbPath}`);
             const db = new better_sqlite3_1.default(dbPath, { allowExtension: true });
             sqliteVec.load(db);
-            logger.debug(`Creating vec_items table if it doesn't exist`);
+            logger.debug(`Creating vec_items table if it doesn't exist (dimension: ${embeddingDimension})`);
             db.exec(`
                 CREATE VIRTUAL TABLE IF NOT EXISTS vec_items USING vec0(
-                    embedding FLOAT[3072],
+                    embedding FLOAT[${embeddingDimension}],
                     product_name TEXT,
                     version TEXT,
                     branch TEXT,
@@ -81,7 +81,7 @@ class DatabaseManager {
             const collectionName = params.collection_name || `${config.product_name.toLowerCase().replace(/\s+/g, '_')}_${config.version}`;
             logger.info(`Connecting to Qdrant at ${qdrantUrl}:${qdrantPort}, collection: ${collectionName}`);
             const qdrantClient = new js_client_rest_1.QdrantClient({ url: qdrantUrl, apiKey: process.env.QDRANT_API_KEY, port: qdrantPort });
-            await this.createCollectionQdrant(qdrantClient, collectionName, logger);
+            await this.createCollectionQdrant(qdrantClient, collectionName, logger, embeddingDimension);
             logger.info(`Qdrant connection established successfully`);
             return { client: qdrantClient, collectionName, type: 'qdrant' };
         }
@@ -91,7 +91,7 @@ class DatabaseManager {
             throw new Error(errMsg);
         }
     }
-    static async createCollectionQdrant(qdrantClient, collectionName, logger) {
+    static async createCollectionQdrant(qdrantClient, collectionName, logger, embeddingDimension) {
         try {
             logger.debug(`Checking if collection ${collectionName} exists`);
             const collections = await qdrantClient.getCollections();
@@ -100,10 +100,10 @@ class DatabaseManager {
                 logger.info(`Collection ${collectionName} already exists`);
                 return;
             }
-            logger.info(`Creating new collection ${collectionName}`);
+            logger.info(`Creating new collection ${collectionName} with dimension ${embeddingDimension}`);
             await qdrantClient.createCollection(collectionName, {
                 vectors: {
-                    size: 3072,
+                    size: embeddingDimension,
                     distance: "Cosine",
                 },
             });
@@ -181,7 +181,7 @@ class DatabaseManager {
         }
         return defaultValue;
     }
-    static async setMetadataValue(dbConnection, key, value, logger) {
+    static async setMetadataValue(dbConnection, key, value, logger, embeddingDimension) {
         try {
             if (dbConnection.type === 'sqlite') {
                 const stmt = dbConnection.db.prepare(`
@@ -193,8 +193,7 @@ class DatabaseManager {
             }
             else if (dbConnection.type === 'qdrant') {
                 const metadataUUID = utils_1.Utils.generateMetadataUUID(key);
-                const dummyEmbeddingSize = 3072;
-                const dummyEmbedding = new Array(dummyEmbeddingSize).fill(0);
+                const dummyEmbedding = new Array(embeddingDimension).fill(0);
                 const metadataPoint = {
                     id: metadataUUID,
                     vector: dummyEmbedding,
@@ -258,7 +257,7 @@ class DatabaseManager {
         logger.info(`No saved run date found for ${repo}, using default: ${defaultDate}`);
         return defaultDate;
     }
-    static async updateLastRunDate(dbConnection, repo, logger) {
+    static async updateLastRunDate(dbConnection, repo, logger, embeddingDimension) {
         const now = new Date().toISOString();
         try {
             if (dbConnection.type === 'sqlite') {
@@ -276,8 +275,7 @@ class DatabaseManager {
                 const metadataKey = `last_run_${repo.replace('/', '_')}`;
                 logger.debug(`Using UUID: ${metadataUUID} for metadata`);
                 // Generate a dummy embedding (all zeros)
-                const dummyEmbeddingSize = 3072; // Same size as your content embeddings
-                const dummyEmbedding = new Array(dummyEmbeddingSize).fill(0);
+                const dummyEmbedding = new Array(embeddingDimension).fill(0);
                 // Create a point with special metadata payload
                 const metadataPoint = {
                     id: metadataUUID,

package/dist/doc2vec.js

@@ -53,6 +53,7 @@ const logger_1 = require("./logger");
 const utils_1 = require("./utils");
 const database_1 = require("./database");
 const content_processor_1 = require("./content-processor");
+const markdown_store_1 = require("./markdown-store");
 const GITHUB_TOKEN = process.env.GITHUB_PERSONAL_ACCESS_TOKEN;
 const execAsync = (0, util_1.promisify)(child_process_1.exec);
 dotenv.config();
@@ -72,6 +73,7 @@ class Doc2Vec {
         // Check environment variable if not specified in config
         const embeddingProvider = this.config.embedding?.provider || process.env.EMBEDDING_PROVIDER || 'openai';
         const embeddingConfig = this.config.embedding || { provider: embeddingProvider };
+        this.embeddingDimension = this.resolveEmbeddingDimension(embeddingConfig);
         if (embeddingProvider === 'azure') {
             const azureApiKey = embeddingConfig.azure?.api_key || process.env.AZURE_OPENAI_KEY;
             const azureEndpoint = embeddingConfig.azure?.endpoint || process.env.AZURE_OPENAI_ENDPOINT;
@@ -88,7 +90,7 @@ class Doc2Vec {
                 apiVersion: azureApiVersion,
             });
             this.embeddingModel = azureDeploymentName;
-            this.logger.info(`Using Azure OpenAI with deployment: ${azureDeploymentName}`);
+            this.logger.info(`Using Azure OpenAI with deployment: ${azureDeploymentName} (${this.embeddingDimension} dimensions)`);
         }
         else {
             const openaiApiKey = embeddingConfig.openai?.api_key || process.env.OPENAI_API_KEY;
@@ -99,9 +101,13 @@ class Doc2Vec {
             }
             this.openai = new openai_1.OpenAI({ apiKey: openaiApiKey });
             this.embeddingModel = openaiModel;
-            this.logger.info(`Using OpenAI with model: ${openaiModel}`);
+            this.logger.info(`Using OpenAI with model: ${openaiModel} (${this.embeddingDimension} dimensions)`);
         }
         this.contentProcessor = new content_processor_1.ContentProcessor(this.logger);
+        // Initialize Postgres markdown store if configured
+        if (this.config.markdown_store) {
+            this.markdownStore = new markdown_store_1.MarkdownStore(this.config.markdown_store, this.logger);
+        }
     }
     loadConfig(configPath) {
         try {
@@ -144,7 +150,26 @@ class Doc2Vec {
             process.exit(1);
         }
     }
+    resolveEmbeddingDimension(embeddingConfig) {
+        const defaultDimension = 3072;
+        const rawConfigValue = embeddingConfig?.dimension;
+        const rawEnvValue = process.env.EMBEDDING_DIMENSION;
+        const candidate = rawConfigValue ?? (rawEnvValue ? Number(rawEnvValue) : undefined);
+        if (candidate === undefined) {
+            return defaultDimension;
+        }
+        const parsedValue = typeof candidate === 'string' ? Number(candidate) : candidate;
+        if (!Number.isFinite(parsedValue) || parsedValue <= 0 || !Number.isInteger(parsedValue)) {
+            this.logger.warn(`Invalid embedding dimension provided (${candidate}), falling back to ${defaultDimension}`);
+            return defaultDimension;
+        }
+        return parsedValue;
+    }
     async run() {
+        // Initialize Postgres markdown store table if configured
+        if (this.markdownStore) {
+            await this.markdownStore.init();
+        }
         this.logger.section('PROCESSING SOURCES');
         for (const sourceConfig of this.config.sources) {
             const sourceLogger = this.logger.child(`source:${sourceConfig.product_name}`);
@@ -168,6 +193,10 @@ class Doc2Vec {
                 sourceLogger.error(`Unknown source type: ${sourceConfig.type}`);
             }
         }
+        // Close the Postgres markdown store connection pool
+        if (this.markdownStore) {
+            await this.markdownStore.close();
+        }
         this.logger.section('PROCESSING COMPLETE');
     }
     async fetchAndProcessGitHubIssues(repo, sourceConfig, dbConnection, logger) {
@@ -373,13 +402,13 @@ class Doc2Vec {
             await processIssue(issues[i]);
         }
         // Update the last run date in the database after processing all issues
-        await database_1.DatabaseManager.updateLastRunDate(dbConnection, repo, logger);
+        await database_1.DatabaseManager.updateLastRunDate(dbConnection, repo, logger, this.embeddingDimension);
         logger.info(`Successfully processed ${issues.length} issues`);
     }
     async processGithubRepo(config, parentLogger) {
         const logger = parentLogger.child('process');
         logger.info(`Starting processing for GitHub repo: ${config.repo}`);
-        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger);
+        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger, this.embeddingDimension);
         // Initialize metadata storage
         await database_1.DatabaseManager.initDatabaseMetadata(dbConnection, logger);
         logger.section('GITHUB ISSUES');
@@ -390,7 +419,7 @@ class Doc2Vec {
     async processWebsite(config, parentLogger) {
         const logger = parentLogger.child('process');
         logger.info(`Starting processing for website: ${config.url}`);
-        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger);
+        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger, this.embeddingDimension);
         await database_1.DatabaseManager.initDatabaseMetadata(dbConnection, logger);
         const validChunkIds = new Set();
         const visitedUrls = new Set();
@@ -420,7 +449,7 @@ class Doc2Vec {
                 return database_1.DatabaseManager.getMetadataValue(dbConnection, `etag:${url}`, undefined, logger);
             },
             set: async (url, etag) => {
-                await database_1.DatabaseManager.setMetadataValue(dbConnection, `etag:${url}`, etag, logger);
+                await database_1.DatabaseManager.setMetadataValue(dbConnection, `etag:${url}`, etag, logger, this.embeddingDimension);
             },
         };
         const lastmodStore = {
@@ -428,9 +457,31 @@ class Doc2Vec {
                 return database_1.DatabaseManager.getMetadataValue(dbConnection, `lastmod:${url}`, undefined, logger);
             },
             set: async (url, lastmod) => {
-                await database_1.DatabaseManager.setMetadataValue(dbConnection, `lastmod:${url}`, lastmod, logger);
+                await database_1.DatabaseManager.setMetadataValue(dbConnection, `lastmod:${url}`, lastmod, logger, this.embeddingDimension);
             },
         };
+        // If Postgres markdown store is enabled for this source, load URLs that
+        // already have markdown stored.  URLs NOT in this set will bypass
+        // lastmod/ETag skip logic so that the store is fully populated on the
+        // first sync.
+        const useMarkdownStore = config.markdown_store === true && this.markdownStore != null;
+        let markdownStoreUrls;
+        if (useMarkdownStore) {
+            markdownStoreUrls = await this.markdownStore.getUrlsWithMarkdown(urlPrefix);
+            logger.info(`Markdown store: ${markdownStoreUrls.size} URLs already stored for prefix ${urlPrefix}`);
+        }
+        // Check whether a full sync has ever completed successfully for this
+        // source.  If not, bypass all lastmod/ETag skip logic so that every
+        // page is processed at least once.  This handles the case where a
+        // previous sync was interrupted (killed mid-crawl) — the stored
+        // ETags/lastmods from the partial run would otherwise cause the
+        // remaining pages to be skipped indefinitely.
+        const syncCompleteKey = `sync_complete:${urlPrefix}`;
+        const syncCompleteValue = await database_1.DatabaseManager.getMetadataValue(dbConnection, syncCompleteKey, undefined, logger);
+        const forceFullSync = syncCompleteValue !== 'true';
+        if (forceFullSync) {
+            logger.info('Full sync has not yet completed for this source — forcing processing of all pages (bypassing lastmod/ETag skip)');
+        }
         const crawlResult = await this.contentProcessor.crawlWebsite(config.url, config, async (url, content) => {
             visitedUrls.add(url);
             logger.info(`Processing content from ${url} (${content.length} chars markdown)`);
@@ -441,21 +492,50 @@ class Doc2Vec {
                     validChunkIds.add(chunk.metadata.chunk_id);
                 }
                 await this.processChunksForUrl(chunks, url, dbConnection, logger);
+                // Store the generated markdown in Postgres
+                if (useMarkdownStore) {
+                    try {
+                        await this.markdownStore.upsertMarkdown(url, config.product_name, content);
+                    }
+                    catch (pgError) {
+                        logger.error(`Failed to store markdown in Postgres for ${url}:`, pgError);
+                    }
+                }
                 return true;
             }
             catch (error) {
                 logger.error(`Error during chunking or embedding for ${url}:`, error);
                 return false;
             }
-        }, logger, visitedUrls, { knownUrls, etagStore, lastmodStore });
+        }, logger, visitedUrls, { knownUrls, etagStore, lastmodStore, markdownStoreUrls, forceFullSync });
         this.recordBrokenLinks(config.url, crawlResult.brokenLinks);
         this.writeBrokenLinksReport();
         logger.info(`Found ${validChunkIds.size} valid chunks across processed pages for ${config.url}`);
         logger.section('CLEANUP');
+        // Remove 404 URLs from the Postgres markdown store
+        if (useMarkdownStore && crawlResult.notFoundUrls && crawlResult.notFoundUrls.size > 0) {
+            logger.info(`Removing ${crawlResult.notFoundUrls.size} not-found URLs from markdown store`);
+            for (const url of crawlResult.notFoundUrls) {
+                try {
+                    await this.markdownStore.deleteMarkdown(url);
+                }
+                catch (pgError) {
+                    logger.error(`Failed to delete markdown from Postgres for ${url}:`, pgError);
+                }
+            }
+        }
         if (crawlResult.hasNetworkErrors) {
             logger.warn('Skipping cleanup due to network errors encountered during crawling. This prevents removal of valid chunks when the site is temporarily unreachable.');
         }
         else {
+            // Mark this source as having completed a full sync.  On subsequent
+            // runs, lastmod/ETag skip logic will function normally.  If the
+            // process is killed before reaching this point, the flag stays
+            // unset and the next run will force-process all pages again.
+            if (forceFullSync) {
+                await database_1.DatabaseManager.setMetadataValue(dbConnection, syncCompleteKey, 'true', logger, this.embeddingDimension);
+                logger.info('Full sync completed successfully — subsequent runs will use normal caching');
+            }
             if (dbConnection.type === 'sqlite') {
                 logger.info(`Running SQLite cleanup for ${urlPrefix}`);
                 database_1.DatabaseManager.removeObsoleteChunksSQLite(dbConnection.db, visitedUrls, urlPrefix, logger);
@@ -502,7 +582,7 @@ class Doc2Vec {
     async processLocalDirectory(config, parentLogger) {
         const logger = parentLogger.child('process');
         logger.info(`Starting processing for local directory: ${config.path}`);
-        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger);
+        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger, this.embeddingDimension);
         const validChunkIds = new Set();
         const processedFiles = new Set();
         logger.section('FILE SCANNING AND EMBEDDING');
@@ -560,7 +640,7 @@ class Doc2Vec {
     async processCodeSource(config, parentLogger) {
         const logger = parentLogger.child('process');
         logger.info(`Starting processing for code source (${config.source})`);
-        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger);
+        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger, this.embeddingDimension);
         const validChunkIds = new Set();
         const processedFiles = new Set();
         let basePath;
@@ -692,10 +772,10 @@ class Doc2Vec {
                             await database_1.DatabaseManager.removeChunksByUrlQdrant(dbConnection, url, logger);
                         }
                     }
-                    await database_1.DatabaseManager.setMetadataValue(dbConnection, fileListKey, JSON.stringify(currentList), logger);
+                    await database_1.DatabaseManager.setMetadataValue(dbConnection, fileListKey, JSON.stringify(currentList), logger, this.embeddingDimension);
                     if (lastMtimeKey) {
                         const nextMtime = maxObservedMtime > 0 ? maxObservedMtime : Date.now();
-                        await database_1.DatabaseManager.setMetadataValue(dbConnection, lastMtimeKey, `${nextMtime}`, logger);
+                        await database_1.DatabaseManager.setMetadataValue(dbConnection, lastMtimeKey, `${nextMtime}`, logger, this.embeddingDimension);
                     }
                 }
             }
@@ -713,7 +793,7 @@ class Doc2Vec {
                 const headSha = await this.getRepoHeadSha(basePath, logger);
                 if (headSha) {
                     const shaKey = this.buildCodeShaMetadataKey(config.repo, repoBranch);
-                    await database_1.DatabaseManager.setMetadataValue(dbConnection, shaKey, headSha, logger);
+                    await database_1.DatabaseManager.setMetadataValue(dbConnection, shaKey, headSha, logger, this.embeddingDimension);
                 }
             }
             if (tempDir) {
@@ -876,7 +956,7 @@ class Doc2Vec {
     async processZendesk(config, parentLogger) {
         const logger = parentLogger.child('process');
         logger.info(`Starting processing for Zendesk: ${config.zendesk_subdomain}.zendesk.com`);
-        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger);
+        const dbConnection = await database_1.DatabaseManager.initDatabase(config, logger, this.embeddingDimension);
         // Initialize metadata storage
         await database_1.DatabaseManager.initDatabaseMetadata(dbConnection, logger);
         const fetchTickets = config.fetch_tickets !== false; // default true
@@ -1051,7 +1131,7 @@ class Doc2Vec {
             }
         }
         // Update the last run date in the database
-        await database_1.DatabaseManager.updateLastRunDate(dbConnection, `zendesk_tickets_${config.zendesk_subdomain}`, logger);
+        await database_1.DatabaseManager.updateLastRunDate(dbConnection, `zendesk_tickets_${config.zendesk_subdomain}`, logger, this.embeddingDimension);
         logger.info(`Successfully processed ${totalTickets} tickets`);
     }
     async fetchAndProcessZendeskArticles(config, dbConnection, logger) {

package/dist/markdown-store.js

@@ -0,0 +1,105 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.MarkdownStore = void 0;
+const pg_1 = require("pg");
+/**
+ * Stores generated markdown for website pages in a Postgres table.
+ *
+ * Table schema (single shared table, default name `markdown_pages`):
+ *   url          TEXT PRIMARY KEY
+ *   product_name TEXT NOT NULL
+ *   markdown     TEXT NOT NULL
+ *   updated_at   TIMESTAMPTZ DEFAULT NOW()
+ *
+ * When a website source has `markdown_store: true`, the crawl loop uses this
+ * store to decide whether to force-process a page that would otherwise be
+ * skipped by lastmod / ETag caching.  If the URL is not yet in Postgres the
+ * page is processed regardless of cache signals, ensuring the table is fully
+ * populated after the first sync.  On subsequent syncs only pages with
+ * detected changes are updated.
+ */
+class MarkdownStore {
+    constructor(config, logger) {
+        this.logger = logger.child('markdown-store');
+        this.tableName = config.table_name ?? 'markdown_pages';
+        const poolConfig = {};
+        if (config.connection_string) {
+            poolConfig.connectionString = config.connection_string;
+        }
+        else {
+            if (config.host)
+                poolConfig.host = config.host;
+            if (config.port)
+                poolConfig.port = config.port;
+            if (config.database)
+                poolConfig.database = config.database;
+            if (config.user)
+                poolConfig.user = config.user;
+            if (config.password)
+                poolConfig.password = config.password;
+        }
+        this.pool = new pg_1.Pool(poolConfig);
+    }
+    /**
+     * Create the markdown table if it doesn't already exist.
+     */
+    async init() {
+        const query = `
+            CREATE TABLE IF NOT EXISTS ${this.escapeIdentifier(this.tableName)} (
+                url          TEXT PRIMARY KEY,
+                product_name TEXT NOT NULL,
+                markdown     TEXT NOT NULL,
+                updated_at   TIMESTAMPTZ DEFAULT NOW()
+            );
+        `;
+        await this.pool.query(query);
+        this.logger.info(`Initialized Postgres markdown store (table: ${this.tableName})`);
+    }
+    /**
+     * Return the set of URLs that already have markdown stored for a given URL
+     * prefix (e.g., "https://istio.io/latest/docs/").  This is called once
+     * before the crawl starts so the crawler can decide which pages need
+     * force-processing.
+     */
+    async getUrlsWithMarkdown(urlPrefix) {
+        const result = await this.pool.query(`SELECT url FROM ${this.escapeIdentifier(this.tableName)} WHERE url LIKE $1`, [urlPrefix + '%']);
+        return new Set(result.rows.map((row) => row.url));
+    }
+    /**
+     * Insert or update the markdown for a URL.  Called after a page is
+     * successfully processed (fetched + converted to markdown).
+     */
+    async upsertMarkdown(url, productName, markdown) {
+        const query = `
+            INSERT INTO ${this.escapeIdentifier(this.tableName)} (url, product_name, markdown, updated_at)
+            VALUES ($1, $2, $3, NOW())
+            ON CONFLICT (url) DO UPDATE SET
+                product_name = EXCLUDED.product_name,
+                markdown     = EXCLUDED.markdown,
+                updated_at   = NOW();
+        `;
+        await this.pool.query(query, [url, productName, markdown]);
+    }
+    /**
+     * Remove a URL from the store (e.g., when a HEAD request returns 404).
+     */
+    async deleteMarkdown(url) {
+        await this.pool.query(`DELETE FROM ${this.escapeIdentifier(this.tableName)} WHERE url = $1`, [url]);
+    }
+    /**
+     * Close the connection pool.  Should be called once after all sources have
+     * been processed.
+     */
+    async close() {
+        await this.pool.end();
+        this.logger.info('Postgres markdown store connection pool closed');
+    }
+    /**
+     * Escape a SQL identifier (table name) to prevent injection.
+     * Uses double-quoting per the SQL standard.
+     */
+    escapeIdentifier(identifier) {
+        return '"' + identifier.replace(/"/g, '""') + '"';
+    }
+}
+exports.MarkdownStore = MarkdownStore;

package/dist/utils.js

@@ -69,12 +69,14 @@ class Utils {
             return url;
         }
     }
-    static buildUrl(href, currentUrl) {
+    static buildUrl(href, currentUrl, logger) {
         try {
             return new URL(href, currentUrl).toString();
         }
         catch (error) {
-            console.warn(`Invalid URL found: ${href}`);
+            if (logger) {
+                logger.warn(`Invalid URL found: ${href}`);
+            }
             return '';
         }
     }

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "doc2vec",
-    "version": "2.4.0",
+    "version": "2.6.0",
     "type": "commonjs",
     "description": "",
     "main": "dist/doc2vec.js",
@@ -43,6 +43,7 @@
         "mammoth": "^1.11.0",
         "openai": "^4.20.1",
         "pdfjs-dist": "^5.3.31",
+        "pg": "^8.18.0",
         "puppeteer": "^24.1.1",
         "sanitize-html": "^2.11.0",
         "sqlite-vec": "0.1.7-alpha.2",
@@ -56,6 +57,7 @@
         "@types/js-yaml": "^4.0.9",
         "@types/jsdom": "^21.1.7",
         "@types/node": "^20.10.0",
+        "@types/pg": "^8.16.0",
         "@types/sanitize-html": "^2.9.5",
         "@types/turndown": "^5.0.4",
         "@vitest/coverage-v8": "^4.0.18",