@tyvm/knowhow 0.0.158 → 0.0.160

package/src/services/conversion/index.ts

@@ -1,2 +0,0 @@
-export * from "./types";
-export * from "./ConversionService";

package/src/services/conversion/types.ts

@@ -1,87 +0,0 @@
-import { AIClient } from "../../clients";
-
-export type Modality = "text" | "html" | "image" | "audio" | "video";
-
-export interface ConvertInput {
-  filePath: string;       // source file on disk
-  inputExt: string;       // e.g. "pdf"
-  /**
-   * When a previous chain step produced multiple files (e.g. one image per
-   * PDF page), they are forwarded here so the next converter can process all
-   * of them. filePath is set to files[0] for single-file converters.
-   */
-  files?: string[];
-  startPage?: number;
-  endPage?: number;
-  startLine?: number;
-  endLine?: number;
-  startTime?: number;
-  endTime?: number;
-}
-
-export interface ConvertResult {
-  outputType: Modality;
-  text?: string;          // for text/html
-  files?: string[];       // for image/audio/video
-  usd_cost?: number;
-}
-
-export interface ConverterContext {
-  clients: AIClient;
-  cacheDir: string;
-  /**
-   * Per-converter options forwarded from ConvertOptions.converterOptions[name].
-   * Lets individual converters accept their own settings (e.g. which model to
-   * use for image->text description, DPI for pdf->image, etc).
-   */
-  options?: Record<string, any>;
-}
-
-export interface Converter {
-  name: string;                   // unique
-  inputExts?: string[];           // e.g. ["pdf"]
-  inputModality?: Modality;       // OR an input modality (e.g. "image")
-  /**
-   * If true, this converter is used as a last-resort fallback for any input
-   * type that has no other converter registered. Catch-all converters are
-   * appended after all specific matches in buildAlternatives, and are also
-   * added as BFS nodes so they can be discovered even for unknown extensions.
-   * Use sparingly — only for truly generic converters like text-passthrough.
-   */
-  catchAll?: boolean;
-  outputType: Modality;
-  /**
-   * If true, the ConversionService will cache the result of this converter to
-   * disk so repeated calls with the same input are fast. Defaults to false —
-   * caching is opt-in. Only enable for expensive converters (e.g. pdf->image,
-   * image->text via LLM) where re-running would be slow or costly.
-   */
-  cache?: boolean;
-  convert: (input: ConvertInput, ctx: ConverterContext) => Promise<ConvertResult>;
-}
-
-export interface ConvertOptions {
-  preferredConverters?: string[];
-  isGoodEnough?: (args: { filePath: string; result: ConvertResult }) => boolean;
-  force?: boolean;
-  startPage?: number;
-  endPage?: number;
-  startLine?: number;
-  endLine?: number;
-  startTime?: number;
-  endTime?: number;
-  onProgress?: (stage: string, fraction: number) => void;
-  /**
-   * Options scoped to a specific converter, keyed by converter name.
-   * e.g. { "image-to-text": { model: "gpt-4o" }, "pdf-to-img": { scale: 2 } }
-   */
-  converterOptions?: Record<string, Record<string, any>>;
-  /**
-   * Explicit intermediate modalities to route through before reaching the
-   * final targetType.  e.g. ["image"] means the chain must go through "image"
-   * first.  When provided, the ConversionService will stitch BFS sub-paths
-   * between each waypoint rather than picking the overall shortest path.
-   * CLI: --to image,text  → via=["image"], target="text"
-   */
-  via?: Modality[];
-}

package/src/services/flags.ts

@@ -1,51 +0,0 @@
-export class FlagsService {
-  flags = {};
-  log = false;
-
-  constructor(flags: string[] = [], log = false) {
-    this.register(flags);
-    this.log = log;
-  }
-
-  register(flags: string[]) {
-    flags.forEach((flag) => {
-      this.flags[flag] = false;
-    });
-  }
-
-  enable(flag: string) {
-    this.flags[flag] = true;
-    if (this.log) {
-      console.log(`${flag} is now enabled \n`);
-    }
-  }
-
-  disable(flag: string) {
-    this.flags[flag] = false;
-    if (this.log) {
-      console.log(`${flag} is now disabled \n`);
-    }
-  }
-
-  flip(flag: string) {
-    this.flags[flag] = !this.flags[flag];
-    if (this.log) {
-      const status = this.flags[flag] ? "enabled" : "disabled";
-      const message = `${flag} is now ${status}`;
-      console.log(message);
-    }
-  }
-
-  enableLogging() {
-    this.log = true;
-  }
-
-  disableLogging() {
-    this.log = false;
-  }
-
-  enabled(flag: string): boolean {
-    return this.flags[flag];
-  }
-}
-

package/src/services/index.ts

@@ -1,117 +0,0 @@
-import { AIClient, Clients } from "../clients";
-import { AgentService } from "./AgentService";
-import { EventService } from "./EventService";
-import { FlagsService } from "./flags";
-import { EmbeddingsService } from "./EmbeddingsService";
-import { KnowhowSimpleClient } from "./KnowhowClient";
-import { McpService } from "./Mcp";
-import { S3Service } from "./S3";
-import { ToolsService } from "./Tools";
-import { CliPluginService, PluginService } from "../plugins/plugins";
-import { DockerService } from "./DockerService";
-import { AgentSyncKnowhowWeb } from "./AgentSyncKnowhowWeb";
-import { AgentSyncFs } from "./AgentSyncFs";
-import { SessionManager } from "./SessionManager";
-import { TaskRegistry } from "./TaskRegistry";
-import { MediaProcessorService } from "./MediaProcessorService";
-import { BehaviorsService } from "./BehaviorsService";
-import { RuntimeReloadService } from "./RuntimeReloadService";
-import { ExtensionsService } from "./ExtensionsService";
-
-import { ConversionService } from "./conversion/ConversionService";
-import { TracingService } from "./TracingService";
-
-export * from "./AgentService";
-export * from "./EventService";
-export * from "./flags";
-export * from "./EmbeddingsService";
-export * from "./S3";
-export * from "./Tools";
-export * from "./LazyToolsService";
-export * from "./MinimalToolsService";
-export * as MCP from "./Mcp";
-export * from "./EmbeddingService";
-export * from "./DockerService";
-export * from "./MediaProcessorService";
-export * from "./AgentSyncKnowhowWeb";
-export * from "./AgentSyncFs";
-export * from "./SessionManager";
-export * from "./TaskRegistry";
-export * from "./SyncedAgentWatcher";
-export * from "./SyncerService";
-export * from "./watchers";
-export { Clients } from "../clients";
-export * from "../util/Trace";
-export { TracingService } from "./TracingService";
-export type { TracerImpl, SpanHandle } from "./TracingService";
-export * from "./conversion";
-export * from "./RuntimeReloadService";
-export * from "./ExtensionsService";
-export { BehaviorsService } from "./BehaviorsService";
-
-let Singletons = {} as {
-  Tools: ToolsService;
-  Events: EventService;
-  Agents: AgentService;
-  Tracing: typeof TracingService;
-  Embeddings: EmbeddingsService;
-  Flags: FlagsService;
-  Mcp: McpService;
-  AwsS3: S3Service;
-  Docker: DockerService;
-  knowhowApiClient: KnowhowSimpleClient;
-  Plugins: PluginService;
-  Clients: AIClient;
-  MediaProcessor: MediaProcessorService;
-  Conversion: ConversionService;
-  Behaviors: BehaviorsService;
-  RuntimeReload: RuntimeReloadService;
-  Extensions: ExtensionsService;
-};
-
-export const services = (): typeof Singletons => {
-  if (Object.keys(Singletons).length === 0) {
-    const Tools = new ToolsService();
-    const Events = new EventService();
-    const Agents = new AgentService(Tools, Events);
-    const MediaProcessor = new MediaProcessorService(Clients);
-    const Behaviors = new BehaviorsService();
-    const Conversion = new ConversionService(Clients, MediaProcessor);
-    const Plugins = new CliPluginService({
-      Agents,
-      Events,
-      Tools,
-      Clients,
-    });
-
-    Singletons = {
-      Agents,
-      AwsS3: new S3Service(),
-      Tracing: TracingService,
-      Clients,
-      Docker: new DockerService(),
-      Events,
-      Embeddings: new EmbeddingsService(),
-      Flags: new FlagsService(),
-      Mcp: new McpService(),
-      MediaProcessor,
-      Conversion,
-      Plugins,
-      Tools,
-      knowhowApiClient: new KnowhowSimpleClient(),
-      Behaviors,
-      RuntimeReload: new RuntimeReloadService(),
-      Extensions: new ExtensionsService(),
-    };
-
-    Singletons.Tools.setContext({
-      Agents,
-      Events,
-      Plugins,
-      Clients,
-      Behaviors,
-    });
-  }
-
-  return Singletons;
-};

package/src/services/modules/computerUse.ts

@@ -1,218 +0,0 @@
-/**
- * Shared, driver-agnostic interface for the computer-use service.
- *
- * The concrete implementation lives in `@tyvm/knowhow-module-computer-use`, but
- * we declare the shape here so that `ToolContext` / `ModuleContext` can carry a
- * `ComputerUse` service without the core knowhow package taking a runtime
- * dependency on the module.
- *
- * An adapter module (e.g. `@tyvm/knowhow-module-computer-use-nutjs`) can
- * register a `ComputerDriver` against the service in its `register` phase; the
- * base module then `init()`s and picks the best available driver.
- */
-
-export interface Point {
-  x: number;
-  y: number;
-}
-
-export interface Size {
-  width: number;
-  height: number;
-}
-
-export interface Region extends Point, Size {}
-
-export type MouseButton = "left" | "right" | "middle";
-
-export interface OverlayPrimitive {
-  kind: "rect" | "circle" | "line" | "point" | "text";
-  x: number;
-  y: number;
-  width?: number;
-  height?: number;
-  x2?: number;
-  y2?: number;
-  text?: string;
-  fontSize?: number;
-  color?: string;
-  lineWidth?: number;
-}
-
-export interface Display {
-  id: number;
-  bounds: Region;
-  scaleFactor: number;
-  primary: boolean;
-}
-
-export interface ScreenshotOptions {
-  region?: Region;
-  displayId?: number;
-  format?: "png" | "jpeg";
-  scale?: number;
-  /** Native capture downscale. Unlike output encoding scale, drivers may apply
-   * this before raw pixels cross the process boundary. */
-  captureScale?: number;
-}
-
-export interface DriverCapabilities {
-  input: boolean;
-  capture: boolean;
-  windows: boolean;
-  reason?: string;
-}
-
-export interface AccessibilityElement {
-  id: string;
-  role: string;
-  subrole?: string;
-  title?: string;
-  description?: string;
-  value?: string;
-  enabled?: boolean;
-  focused?: boolean;
-  bounds?: Region;
-  actions: string[];
-  childCount: number;
-}
-
-export interface AccessibilityOptions {
-  maxDepth?: number;
-  maxElements?: number;
-  interactiveOnly?: boolean;
-}
-
-/**
- * The engine seam. Concrete drivers (Rust core, CLI adapter, nut.js demo)
- * implement this. Registered with the service via `registerDriver`.
- */
-export interface ComputerDriver {
-  readonly name: string;
-  /**
-   * Lower numbers win when multiple drivers are registered. The default Rust
-   * core registers at 100; a fallback CLI adapter at 500; an experimental demo
-   * adapter can register at e.g. 10 to force selection.
-   */
-  readonly priority?: number;
-
-  capabilities(): Promise<DriverCapabilities>;
-
-  // screen
-  getDisplays(): Promise<Display[]>;
-  screenSize(): Promise<Size>;
-  screenshot(opts?: ScreenshotOptions): Promise<Buffer>;
-  pixelColor(p: Point): Promise<string>;
-
-  // mouse
-  mousePosition(): Promise<Point>;
-  moveMouse(p: Point, opts?: { duration?: number }): Promise<void>;
-  click(button?: MouseButton, opts?: { double?: boolean }): Promise<void>;
-  mouseDown(button?: MouseButton): Promise<void>;
-  mouseUp(button?: MouseButton): Promise<void>;
-  drag(
-    from: Point,
-    to: Point,
-    opts?: { button?: MouseButton; duration?: number }
-  ): Promise<void>;
-  scroll(dx: number, dy: number): Promise<void>;
-
-  // keyboard
-  typeText(text: string, opts?: { delay?: number }): Promise<void>;
-  pressKey(key: string): Promise<void>;
-  hotkey(...keys: string[]): Promise<void>;
-  keyDown(key: string): Promise<void>;
-  keyUp(key: string): Promise<void>;
-
-  // windows (optional / capability-gated)
-  getActiveWindow?(): Promise<{ title: string; app?: string; bounds?: Region } | null>;
-  listWindows?(): Promise<Array<{ title: string; app?: string; bounds?: Region }>>;
-  focusWindow?(match: string): Promise<boolean>;
-
-  accessibilityTrusted?(): Promise<boolean>;
-  accessibilityElements?(options?: AccessibilityOptions): Promise<AccessibilityElement[]>;
-  setAccessibilityValue?(id: string, value: string): Promise<void>;
-  performAccessibilityAction?(id: string, action: string): Promise<void>;
-
-  // click-through visual debugging (optional / platform capability-gated)
-  showOverlay?(primitives: OverlayPrimitive[]): Promise<void>;
-  clearOverlay?(): Promise<void>;
-
-  dispose?(): Promise<void>;
-}
-
-/**
- * The registered service (`ComputerUse`) surfaced to tools/CLI/MCP. It owns
- * driver selection and provides the normalized, platform-agnostic verbs.
- *
- * We keep this as an interface (not a class) in the core package so the module
- * owns the concrete implementation; anything reading `context.ComputerUse`
- * codes against this contract.
- */
-export interface ComputerUseService {
-  /** Register a driver (called by the base module and any adapter modules). */
-  registerDriver(driver: ComputerDriver): void;
-  /** List the names of all registered drivers. */
-  listDrivers(): string[];
-  /** Resolve/select the active driver (highest priority + capable). */
-  getDriver(): Promise<ComputerDriver>;
-
-  capabilities(): Promise<DriverCapabilities>;
-  getDisplays(): Promise<Display[]>;
-  screenSize(): Promise<Size>;
-  /** Returns encoded image bytes (PNG/JPEG) honoring service defaults. */
-  screenshot(opts?: ScreenshotOptions): Promise<Buffer>;
-  /** Convenience: base64 data (no data: prefix) for multimodal messages. */
-  screenshotBase64(opts?: ScreenshotOptions): Promise<string>;
-  pixelColor(p: Point): Promise<string>;
-
-  mousePosition(): Promise<Point>;
-  moveMouse(p: Point, opts?: { duration?: number }): Promise<void>;
-  click(button?: MouseButton, opts?: { double?: boolean }): Promise<void>;
-  drag(
-    from: Point,
-    to: Point,
-    opts?: { button?: MouseButton; duration?: number }
-  ): Promise<void>;
-  scroll(dx: number, dy: number): Promise<void>;
-  /** Smooth stepped scrolling (repeat a small delta with a pause between). */
-  scrollBy?(
-    dx: number,
-    dy: number,
-    opts?: { repeat?: number; intervalMs?: number }
-  ): Promise<void>;
-
-  typeText(text: string, opts?: { delay?: number }): Promise<void>;
-  pressKey(key: string): Promise<void>;
-  hotkey(...keys: string[]): Promise<void>;
-
-  /** Screenshot with an optional labeled grid and/or crosshair overlay. */
-  screenshotAnnotated?(
-    opts?: ScreenshotOptions & {
-      grid?: boolean;
-      gridStep?: number;
-      crosshair?: Point;
-    }
-  ): Promise<Buffer>;
-
-  getActiveWindow(): Promise<{
-    title: string;
-    app?: string;
-    bounds?: Region;
-  } | null>;
-  listWindows(): Promise<
-    Array<{ title: string; app?: string; bounds?: Region }>
-  >;
-  focusWindow(match: string): Promise<boolean>;
-
-  accessibilityTrusted(): Promise<boolean>;
-  accessibilityElements(options?: AccessibilityOptions): Promise<AccessibilityElement[]>;
-  selectAccessibilityOption(id: string, option: string): Promise<void>;
-  setAccessibilityValue(id: string, value: string): Promise<void>;
-  performAccessibilityAction(id: string, action: string): Promise<void>;
-
-  showOverlay(primitives: OverlayPrimitive[]): Promise<void>;
-  clearOverlay(): Promise<void>;
-
-  dispose(): Promise<void>;
-}