@tyvm/knowhow 0.0.108 → 0.0.109-dev.86123ed

package/src/fileSync.ts

@@ -1,11 +1,14 @@
 import * as fs from "fs";
 import * as path from "path";
+import * as os from "os";
 import { KnowhowSimpleClient, KNOWHOW_API_URL } from "./services/KnowhowClient";
 import { loadJwt } from "./login";
 import { getConfig } from "./config";
 import { services } from "./services";
 import { S3Service } from "./services/S3";
-import { getHashes, hasFileChangedSinceUpload, saveUploadHash, isLocalFileMatchingRemote, isLocalFileMatchingDownloadHash, saveDownloadHash } from "./hashes";
+import { getHashes, saveHashes, hasFileChangedSinceUpload, saveUploadHash, isLocalFileMatchingRemote, isLocalFileMatchingDownloadHash, saveDownloadHash } from "./hashes";
+
+export const DEFAULT_BATCH_SIZE = 5;
 
 export interface FileSyncOptions {
   upload?: boolean;
@@ -15,6 +18,33 @@ export interface FileSyncOptions {
   dryRun?: boolean;
 }
 
+/**
+ * Run an array of async tasks in batches of `batchSize` at a time.
+ * Returns results in the same order as the input tasks.
+ */
+export async function batchRun<T>(
+  tasks: (() => Promise<T>)[],
+  batchSize: number = DEFAULT_BATCH_SIZE
+): Promise<T[]> {
+  const results: T[] = [];
+  for (let i = 0; i < tasks.length; i += batchSize) {
+    const batch = tasks.slice(i, i + batchSize);
+    const batchResults = await Promise.all(batch.map((t) => t()));
+    results.push(...batchResults);
+  }
+  return results;
+}
+
+/**
+ * Expands a leading ~ to the user's home directory
+ */
+function expandHome(p: string): string {
+  if (p === "~" || p.startsWith("~/") || p.startsWith("~\\")) {
+    return path.join(os.homedir(), p.slice(1));
+  }
+  return p;
+}
+
 /**
  * Returns true if the path looks like a directory (ends with /)
  */
@@ -25,7 +55,7 @@ function isDirectoryPath(p: string): boolean {
 /**
  * Recursively list all files in a local directory, returning relative paths
  */
-function listFilesRecursively(dir: string): string[] {
+export function listFilesRecursively(dir: string): string[] {
   const results: string[] = [];
   if (!fs.existsSync(dir)) return results;
   const entries = fs.readdirSync(dir, { withFileTypes: true });
@@ -83,7 +113,8 @@ export async function fileSync(options: FileSyncOptions = {}) {
 
   // Process each file mount
   for (const mount of config.files) {
-    const { remotePath, localPath, direction = "download" } = mount;
+    const { remotePath, localPath: rawLocalPath, direction = "download" } = mount;
+    const localPath = expandHome(rawLocalPath);
 
     // Determine actual direction based on flags and config
     let actualDirection = direction;
@@ -126,16 +157,16 @@ export async function fileSync(options: FileSyncOptions = {}) {
   }
 }
 
-
 /**
  * Download a file from Knowhow FS to local filesystem
  */
-async function downloadFile(
+export async function downloadFile(
   client: KnowhowSimpleClient,
   s3Service: S3Service,
   remotePath: string,
   localPath: string,
-  dryRun: boolean
+  dryRun: boolean,
+  hashes?: any
 ): Promise<void> {
   console.log(`⬇️  Downloading ${remotePath} → ${localPath}`);
 
@@ -146,8 +177,7 @@ async function downloadFile(
 
   try {
     // Fast-path: check stored download hash before hitting the API
-    const hashes = await getHashes();
-    if (await isLocalFileMatchingDownloadHash(localPath, hashes)) {
+    if (hashes && await isLocalFileMatchingDownloadHash(localPath, hashes)) {
       console.log(`   ✓ Skipping ${localPath} (matches stored download hash)`);
       return;
     }
@@ -159,7 +189,7 @@ async function downloadFile(
     if (isLocalFileMatchingRemote(localPath, checksumSHA256)) {
       console.log(`   ✓ Skipping ${localPath} (matches remote checksum)`);
       // Store the hash so future syncs can skip without hitting the API
-      await saveDownloadHash(localPath);
+      await saveDownloadHash(localPath, hashes);
       return;
     }
 
@@ -173,7 +203,7 @@ async function downloadFile(
     await s3Service.downloadFromPresignedUrl(downloadUrl, localPath);
 
     // Save download hash so we can skip unchanged files next time
-    await saveDownloadHash(localPath);
+    await saveDownloadHash(localPath, hashes);
 
     // Get file size for logging
     const stats = fs.statSync(localPath);
@@ -186,12 +216,13 @@ async function downloadFile(
 /**
  * Upload a file from local filesystem to Knowhow FS
  */
-async function uploadFile(
+export async function uploadFile(
   client: KnowhowSimpleClient,
   s3Service: S3Service,
   remotePath: string,
   localPath: string,
-  dryRun: boolean
+  dryRun: boolean,
+  hashes?: any
 ): Promise<void> {
   console.log(`⬆️  Uploading ${localPath} → ${remotePath}`);
 
@@ -207,15 +238,14 @@ async function uploadFile(
   }
 
   // Skip upload if file hasn't changed since last upload
-  const hashes = await getHashes();
-  const changed = await hasFileChangedSinceUpload(localPath, hashes);
+  const changed = hashes ? await hasFileChangedSinceUpload(localPath, hashes) : true;
   if (!changed) {
     console.log(`   ✓ Skipping ${localPath} (unchanged since last upload)`);
     return;
   }
 
   // Get presigned upload URL
-  const presignedUrl = await client.getOrgFilePresignedUploadUrl(remotePath);
+  const presignedUrl = await client.getOrgFilePresignedUploadUrl(remotePath, localPath);
 
   // Upload file using presigned URL
   await s3Service.uploadToPresignedUrl(presignedUrl, localPath);
@@ -224,7 +254,7 @@ async function uploadFile(
   await client.markOrgFileUploadComplete(remotePath);
 
   // Save upload hash so we can skip unchanged files next time
-  await saveUploadHash(localPath);
+  await saveUploadHash(localPath, hashes);
 
   const stats = fs.statSync(localPath);
   console.log(`   ✓ Uploaded ${stats.size} bytes`);
@@ -246,6 +276,8 @@ export async function uploadDirectory(
 
   console.log(`⬆️  Uploading directory ${localDir} → ${remoteDir}`);
 
+  const hashes = await getHashes();
+
   if (!fs.existsSync(localDir)) {
     console.warn(`   ⚠️  Local directory not found: ${localDir}`);
     return 0;
@@ -261,26 +293,30 @@ export async function uploadDirectory(
 
   console.log(`   Found ${localFiles.length} local file(s)`);
 
-  let count = 0;
-  for (const relFile of localFiles) {
+  const tasks = localFiles.map((relFile) => async () => {
     const localFilePath = localDir + relFile;
     const remoteFilePath = remoteDir + relFile;
     try {
-      await uploadFile(client, s3Service, remoteFilePath, localFilePath, dryRun);
-      count++;
+      await uploadFile(client, s3Service, remoteFilePath, localFilePath, dryRun, hashes);
+      return 1;
     } catch (error) {
       console.error(
         `   ❌ Failed to upload ${localFilePath}, skipping: ${error.message}`
       );
+      return 0;
     }
-  }
-  return count;
+  });
+
+  const counts = await batchRun(tasks);
+  await saveHashes(hashes);
+
+  return counts.reduce((sum, n) => sum + n, 0);
 }
 
 /**
  * Download all files from a remote directory path to a local directory
  */
-async function downloadDirectory(
+export async function downloadDirectory(
   client: KnowhowSimpleClient,
   s3Service: S3Service,
   remotePath: string,
@@ -293,6 +329,8 @@ async function downloadDirectory(
 
   console.log(`⬇️  Downloading directory ${remoteDir} → ${localDir}`);
 
+  const hashes = await getHashes();
+
   // List all org files and find those in the remote directory
   const response = await client.listOrgFiles();
   const allFiles = response.data;
@@ -313,16 +351,19 @@ async function downloadDirectory(
 
   console.log(`   Found ${matchingFiles.length} remote file(s)`);
 
-  let count = 0;
-  for (const f of matchingFiles) {
+  const tasks = matchingFiles.map((f) => async () => {
     const fullRemotePath = f.folderPath.endsWith("/")
       ? f.folderPath + f.fileName
       : f.folderPath + "/" + f.fileName;
     // Strip the base remote dir prefix to get relative path
     const relativePath = fullRemotePath.slice(remoteDir.length);
     const localFilePath = localDir + relativePath;
-    await downloadFile(client, s3Service, fullRemotePath, localFilePath, dryRun);
-    count++;
-  }
-  return count;
+    await downloadFile(client, s3Service, fullRemotePath, localFilePath, dryRun, hashes);
+    return 1;
+  });
+
+  const counts = await batchRun(tasks);
+  await saveHashes(hashes);
+
+  return counts.reduce((sum, n) => sum + n, 0);
 }

package/src/hashes.ts

@@ -1,16 +1,35 @@
 import fs from "fs";
 import * as crypto from "crypto";
 import { Hashes } from "./types";
-import { readFile, writeFile } from "./utils";
+import { readFile } from "./utils";
 import { convertToText } from "./conversion";
 
 export async function getHashes() {
-  const hashes = JSON.parse(await readFile(".knowhow/.hashes.json", "utf8"));
-  return hashes as Hashes;
+  try {
+    const hashes = JSON.parse(await readFile(".knowhow/.hashes.json", "utf8"));
+    return hashes as Hashes;
+  } catch (err: any) {
+    if (err.code === "ENOENT") {
+      return {} as Hashes;
+    }
+    throw err;
+  }
 }
 
+/**
+ * Atomically save hashes to disk — writes to a temp file then renames,
+ * preventing concurrent writes from producing corrupted/truncated JSON.
+ */
 export async function saveHashes(hashes: any) {
-  await writeFile(".knowhow/.hashes.json", JSON.stringify(hashes, null, 2));
+  const target = ".knowhow/.hashes.json";
+  const tmp = `${target}.tmp.${process.pid}`;
+  try {
+    fs.writeFileSync(tmp, JSON.stringify(hashes, null, 2));
+    fs.renameSync(tmp, target);
+  } catch (err) {
+    try { fs.unlinkSync(tmp); } catch (_) {}
+    throw err;
+  }
 }
 
 export async function md5Hash(str: string) {
@@ -90,17 +109,19 @@ export async function hasFileChangedSinceUpload(
 }
 
 /**
- * Saves the hash of the file at the time of a successful upload
+ * Mutates the provided hashes object with the upload hash for localPath.
+ * If no hashes object is provided, loads, mutates, and saves independently.
  */
-export async function saveUploadHash(localPath: string) {
-  const hashes = await getHashes();
+export async function saveUploadHash(localPath: string, hashes?: any) {
+  const standalone = !hashes;
+  if (standalone) hashes = await getHashes();
   const content = fs.readFileSync(localPath);
   const currentHash = crypto.createHash("md5").update(content).digest("hex");
   if (!hashes[localPath]) {
     hashes[localPath] = { fileHash: currentHash, promptHash: "" };
   }
   hashes[localPath][UPLOAD_KEY] = currentHash;
-  await saveHashes(hashes);
+  if (standalone) await saveHashes(hashes);
 }
 
 /**
@@ -120,18 +141,19 @@ export async function isLocalFileMatchingDownloadHash(
 }
 
 /**
- * Saves the SHA-256 hash of the file after a successful download so we can
- * skip unchanged files on the next sync.
+ * Mutates the provided hashes object with the download hash for localPath.
+ * If no hashes object is provided, loads, mutates, and saves independently.
  */
-export async function saveDownloadHash(localPath: string) {
-  const hashes = await getHashes();
+export async function saveDownloadHash(localPath: string, hashes?: any) {
+  const standalone = !hashes;
+  if (standalone) hashes = await getHashes();
   const content = fs.readFileSync(localPath);
   const currentHash = crypto.createHash("sha256").update(content).digest("base64");
   if (!hashes[localPath]) {
     hashes[localPath] = { fileHash: currentHash, promptHash: "" };
   }
   hashes[localPath][DOWNLOAD_KEY] = currentHash;
-  await saveHashes(hashes);
+  if (standalone) await saveHashes(hashes);
 }
 
 /**

package/src/index.ts

@@ -53,6 +53,7 @@ export * as ai from "./ai";
 
 // Export module system types for external modules
 export * from "./services/modules/types";
+export { ModulesService } from "./services/modules";
 // Export plugin types for external plugins
 export { PluginBase } from "./plugins/PluginBase";
 export { PluginMeta, Plugin, PluginContext } from "./plugins/types";
@@ -138,6 +139,23 @@ export async function upload() {
       if (!source.remoteId) {
         throw new Error("remoteId is required for knowhow uploads");
       }
+      // Warn if the local embeddingModel differs from the one stored on the backend
+      try {
+        const remoteEmbedding = await knowhowApiClient.getOrgEmbedding(source.remoteId);
+        const localModel = config.embeddingModel || EmbeddingModels.openai.EmbeddingAda2;
+        const remoteModel = remoteEmbedding?.modelName;
+        if (remoteModel && remoteModel !== localModel) {
+          console.warn(
+            `⚠️  WARNING: Embedding model mismatch for "${remoteEmbedding.name}" (remoteId: ${source.remoteId}).\n` +
+            `   Local config.embeddingModel:  ${localModel}\n` +
+            `   Backend embedding modelName:  ${remoteModel}\n` +
+            `   Vectors generated with different models are not comparable — search results will be incorrect.\n` +
+            `   Update your config.embeddingModel to "${remoteModel}" or update the backend embedding to "${localModel}".`
+          );
+        }
+      } catch (e) {
+        // Non-fatal — don't block upload if metadata fetch fails
+      }
       const url = await knowhowApiClient.getPresignedUploadUrl(source);
       console.log("Uploading to", url);
       await AwsS3.uploadToPresignedUrl(url, source.output);

package/src/logger.ts

@@ -0,0 +1,197 @@
+import type { LogLevel } from "./services/EventService";
+
+/**
+ * App-wide logger utility.
+ *
+ * Features:
+ * 1. `logger.info/warn/error(source, message)` — routes through EventService
+ * 2. `Logger.of("ClassName")` — creates a bound logger so you don't repeat the source
+ * 3. `logger.installConsoleOverload()` — replaces console.log/warn/error/info with
+ *    our closure, so ALL output (including third-party modules) goes through us
+ * 4. `logger.silence()` / `logger.unsilence()` — suppress all output, useful for
+ *    commands that need clean stdout (e.g. github-credentials)
+ *
+ * Usage (module-level):
+ *   import { logger } from "../logger";
+ *   logger.info("MyService", "Something happened");
+ *
+ * Usage (class-level):
+ *   import { Logger } from "../logger";
+ *   class MyClass {
+ *     private logger = Logger.of("MyClass");
+ *     doThing() { this.logger.info("Something happened"); }
+ *   }
+ *
+ * Silence mode (for clean-stdout commands):
+ *   logger.silence();   // suppress everything
+ *   // ... do work that must produce clean stdout ...
+ *   logger.unsilence(); // restore
+ */
+
+// ---- Internal state ---------------------------------------------------------
+
+let silenced = false;
+
+// Original console methods — saved before any overload is installed
+const _originalConsole = {
+  log: console.log.bind(console),
+  warn: console.warn.bind(console),
+  error: console.error.bind(console),
+  info: console.info.bind(console),
+};
+
+let consoleOverloadInstalled = false;
+
+// ---- EventService lazy accessor ---------------------------------------------
+
+function getEvents() {
+  try {
+    const { services } = require("./services") as typeof import("./services");
+    return services().Events;
+  } catch {
+    return null;
+  }
+}
+
+// ---- Core emit logic --------------------------------------------------------
+
+function emit(source: string, message: string, level: LogLevel): void {
+  if (silenced) return;
+
+  try {
+    const events = getEvents();
+    if (events) {
+      events.log(source, message, level);
+      return;
+    }
+  } catch {
+    // fall through to direct console output
+  }
+
+  // Fallback: use original console methods (bypasses any overload we installed)
+  const prefix = source ? `[${source}] ` : "";
+  if (level === "warn") _originalConsole.warn(`${prefix}${message}`);
+  else if (level === "error") _originalConsole.error(`${prefix}${message}`);
+  else _originalConsole.log(`${prefix}${message}`);
+}
+
+// ---- Bound logger (returned by Logger.of) -----------------------------------
+
+export interface BoundLogger {
+  log(message: string, level?: LogLevel): void;
+  info(message: string): void;
+  warn(message: string): void;
+  error(message: string): void;
+}
+
+function makeBoundLogger(source: string): BoundLogger {
+  return {
+    log(message: string, level: LogLevel = "info"): void {
+      emit(source, message, level);
+    },
+    info(message: string): void {
+      emit(source, message, "info");
+    },
+    warn(message: string): void {
+      emit(source, message, "warn");
+    },
+    error(message: string): void {
+      emit(source, message, "error");
+    },
+  };
+}
+
+// ---- Public API -------------------------------------------------------------
+
+export const logger = {
+  log(source: string, message: string, level: LogLevel = "info"): void {
+    emit(source, message, level);
+  },
+
+  info(source: string, message: string): void {
+    emit(source, message, "info");
+  },
+
+  warn(source: string, message: string): void {
+    emit(source, message, "warn");
+  },
+
+  error(source: string, message: string): void {
+    emit(source, message, "error");
+  },
+
+  /**
+   * Suppress all log output. Useful for commands that need clean stdout
+   * (e.g. git credential helpers). All logger.* calls and overloaded
+   * console.* calls become no-ops until unsilence() is called.
+   */
+  silence(): void {
+    silenced = true;
+  },
+
+  /**
+   * Restore log output after a silence() call.
+   */
+  unsilence(): void {
+    silenced = false;
+  },
+
+  /**
+   * Returns true if the logger is currently silenced.
+   */
+  isSilenced(): boolean {
+    return silenced;
+  },
+
+  /**
+   * Install console overload. After this call, console.log/warn/error/info
+   * all route through our closure (respecting silence mode).
+   * Safe to call multiple times — only installs once.
+   *
+   * Call this early in CLI startup (before any modules are loaded) to ensure
+   * third-party module logs don't bypass the silence mechanism.
+   */
+  installConsoleOverload(): void {
+    if (consoleOverloadInstalled) return;
+    consoleOverloadInstalled = true;
+
+    const route = (originalFn: (...args: any[]) => void, args: any[]) => {
+      if (silenced) return;
+      originalFn(...args);
+    };
+
+    console.log = (...args: any[]) => route(_originalConsole.log, args);
+    console.info = (...args: any[]) => route(_originalConsole.info, args);
+    console.warn = (...args: any[]) => route(_originalConsole.warn, args);
+    // Note: console.error is intentionally NOT overloaded — real errors (stack
+    // traces, crash reports) should always be visible. Only suppress via silence().
+    // If you want to suppress errors too, call logger.silence() which checks the flag
+    // before the overloaded console.warn/log routes reach here anyway.
+  },
+
+  /**
+   * Remove the console overload and restore original console methods.
+   */
+  uninstallConsoleOverload(): void {
+    if (!consoleOverloadInstalled) return;
+    console.log = _originalConsole.log;
+    console.info = _originalConsole.info;
+    console.warn = _originalConsole.warn;
+    consoleOverloadInstalled = false;
+  },
+};
+
+/**
+ * Factory for creating a bound logger with a fixed source name.
+ * Ideal for class-level loggers:
+ *
+ *   class MyClass {
+ *     private logger = Logger.of("MyClass");
+ *     doThing() { this.logger.info("hello"); }
+ *   }
+ */
+export const Logger = {
+  of(source: string): BoundLogger {
+    return makeBoundLogger(source);
+  },
+};

package/src/plugins/embedding.ts

@@ -20,9 +20,12 @@ export class EmbeddingPlugin extends PluginBase {
 
   constructor(context) {
     super(context);
-    
+
     // Subscribe to file:post-edit events
-    this.context.Events.on("file:post-edit", this.handleFilePostEdit.bind(this));
+    this.context.Events.on(
+      "file:post-edit",
+      this.handleFilePostEdit.bind(this)
+    );
   }
 
   async embed() {
@@ -68,10 +71,12 @@ export class EmbeddingPlugin extends PluginBase {
       this.log(`Reading entry ${entry.id}`);
     }
 
-    const contextLength = JSON.stringify(context).split(" ").length;
+    const ids = context.map((entry) => entry.id);
+
+    const contextLength = JSON.stringify(ids).split(" ").length;
     this.log(`Found ${context.length} entries. Loading ${contextLength} words`);
 
-    return `EMBEDDING PLUGIN: Our knowledgebase contains this information which can be used to answer the question:
-    ${JSON.stringify(context)}`;
+    return `EMBEDDING PLUGIN: Our knowledgebase indicates these embedding entries may be related to the question:
+    ${JSON.stringify(ids)}`;
   }
-}
+}

package/src/plugins/plugins.ts

@@ -1,5 +1,4 @@
 import { Plugin, PluginContext } from "./types";
-import { Config } from "../types";
 import { VimPlugin } from "./vim";
 import { LinterPlugin } from "./LinterPlugin";
 import { LanguagePlugin } from "./language";
@@ -49,26 +48,6 @@ export class PluginService {
     return instance.meta.key;
   }
 
-  /**
-   * Load plugins from config's pluginPackages map.
-   * Each entry maps a plugin key to an npm package name or file path.
-   * Errors are caught and logged as warnings without crashing.
-   */
-  async loadPluginsFromConfig(config: Config): Promise<void> {
-    const pluginPackages = config.pluginPackages || {};
-    for (const [key, spec] of Object.entries(pluginPackages)) {
-      try {
-        await this.loadPlugin(spec);
-      } catch (error) {
-        this.events?.log(
-          "PluginService",
-          `Failed to load plugin "${key}" from "${spec}": ${error instanceof Error ? error.message : error}`,
-          "warn"
-        );
-      }
-    }
-  }
-
   /** Disable a plugin by its key; returns `true` if found. */
   disablePlugin(key: string): boolean {
     const p = this.pluginMap.get(key);

package/src/plugins/vim.ts

@@ -73,22 +73,11 @@ export class VimPlugin extends PluginBase {
 
   async call() {
     const vimFiles = await this.getVimFiles();
-    const fileContents = await Promise.all(
-      vimFiles.map(async (f) => {
-        const loaded = await this.getFileContents(f);
-
-        const preview =
-          loaded.content.length > 1000
-            ? loaded.content.slice(0, 1000) +
-              "... file trimmed, read file for full content"
-            : loaded.content;
-
-        return {
-          sourceFile: loaded.filePath,
-          content: loaded.content.slice(0, 1000),
-        };
-      })
-    );
+    const fileContents = vimFiles.map((f) => {
+      return {
+        sourceFile: f,
+      };
+    });
     if (fileContents.length === 0) {
       return "VIM PLUGIN: No files open in vim";
     }