gnome-jarvis/daemon/harness-bridge.js

import { voiceSettings } from './voice-settings.js';
import { EventEmitter } from 'node:events';
import os from 'node:os';
import { acquireQvac, closeQvac, releaseQvac, Agent, QVAC_MASTER } from './qvac-master.js';
import { createRuntimeTools } from '../skills/runtime-tools.js';
import { createPhase2Tools, filesystemRoots } from '../skills/phase2-tools.js';
import { createQvacTools } from '../skills/qvac-tools.js';
import { voiceSystemPrompt, parseHudSidecar } from '../skills/voice-prompt.js';
import { createComputerObserveTools } from '../skills/computer-observe.js';
import { createComputerActTools } from '../skills/computer-act.js';
import { createPhase9GatewayTool } from '../skills/phase9-tools.js';
import { createBrowserTools } from '../skills/browser-tools.js';
import { createWebcamTools } from '../skills/webcam-tools.js';
import { ensureAgentWorkspace, normalizeAssistantName } from './agent-workspace.js';

export function harnessRoots(fsAccess) {
  if (fsAccess === 'filesystem') return ['/'];
  if (fsAccess === 'home') return [os.homedir()];
  return [];
}

export class HarnessBridge extends EventEmitter {
  constructor({ cwd, model = QVAC_MASTER.model, tools = [], computer, observer, actuator, browser, camera, webcam, webcamNormalizer, permissionMode = 'ask', fsAccess } = {}) {
    super();
    const settings = voiceSettings();
    const access = fsAccess ?? settings.fsAccess;
    const assistantName = normalizeAssistantName(settings.assistantName);
    const workspace = cwd || ensureAgentWorkspace({ name: assistantName, prompt: settings.assistantPrompt });
    const roots = harnessRoots(access);
    this.assistantName = assistantName;
    this.assistantPrompt = settings.assistantPrompt;
    this.options = {
      cwd: workspace,
      roots,
      model,
      tools: [
        ...createRuntimeTools({ computer, camera }),
        ...createPhase2Tools({ cwd: workspace, computer, roots: filesystemRoots(access, workspace) }),
        ...createComputerObserveTools({ computer, observer }),
        ...createComputerActTools({ actuator }),
        ...createQvacTools(),
        ...createPhase9GatewayTool(),
        ...createBrowserTools({ browser }),
        ...createWebcamTools({ camera, capture: webcam, normalizer: webcamNormalizer }),
        ...tools,
      ],
      builtinTools: ['read_file', 'write_file', 'search_replace', 'list_dir', 'grep', 'run_terminal_cmd', 'web_fetch', 'fetch_page', 'google_search', 'web_search', 'wiki_search', 'hn_search', 'code_search', 'todo_write', 'task', 'update_goal', 'memory_search', 'memory_get', 'memory_write'],
      webFetch: true,
      browser,
      permissionMode,
      origin: 'jarvis-qvac',
      system: voiceSystemPrompt(assistantName, settings.assistantPrompt),
      voice: true,
      maxTurns: settings.maxTurns,
      maxShellCalls: settings.maxShellCalls,
      maxToolRounds: settings.maxToolRounds,
    };
    this.session = null;
  }

  async start() {
    if (this.session) return this.session;
    if (process.env.JARVIS_QVAC_MODEL && this.options.model !== process.env.JARVIS_QVAC_MODEL) {
      throw new Error(`Jarvis uses one QVAC master model (${process.env.JARVIS_QVAC_MODEL}); requested ${this.options.model}`);
    }
    await acquireQvac();
    this._acquired = true;
    try {
      this.session = await Agent.create(this.options);
    } catch (error) {
      releaseQvac();
      this._acquired = false;
      await closeQvac();
      throw error;
    }
    for (const event of ['agent_message_chunk', 'agent_thought_chunk', 'tool_call', 'tool_result', 'permission', 'ask_user', 'cap-chunk', 'error']) {
      this.session.on(event, (payload) => {
        if (event === 'agent_message_chunk' && payload?.text) {
          const parsed = parseHudSidecar(payload.text);
          if (parsed.hud) this.emit('hud_sidecar', parsed.hud);
        }
        this.emit(event, payload);
      });
    }
    return this.session;
  }

  setIdentity(name, extra) {
    this.assistantName = normalizeAssistantName(name);
    this.assistantPrompt = extra || '';
    this.refreshPrompt();
  }

  refreshPrompt() {
    if (!this.options) return;
    this.options.system = voiceSystemPrompt(this.assistantName, this.assistantPrompt);
  }

  async ask(text) {
    this.refreshPrompt();
    if (!this.session) await this.start();
    // Some QVAC runs deliver the final assistant text through the stream but
    // omit it from the final envelope after a tool call. Keep the current
    // post-tool stream as a fallback so the daemon can still speak the reply.
    let streamed = '';
    let postTool = '';
    let sawTool = false;
    const onChunk = (payload) => {
      const chunk = String(payload?.text || payload?.delta || '');
      streamed += chunk;
      if (sawTool) postTool += chunk;
    };
    const onToolCall = () => { sawTool = true; postTool = ''; streamed = ''; };
    const chunkSubscription = this.session.on('agent_message_chunk', onChunk);
    const toolSubscription = this.session.on('tool_call', onToolCall);
    const offChunk = typeof chunkSubscription === 'function' ? chunkSubscription : () => this.session.off?.('agent_message_chunk', onChunk);
    const offToolCall = typeof toolSubscription === 'function' ? toolSubscription : () => this.session.off?.('tool_call', onToolCall);
    try {
      const reply = await this.session.prompt(text);
      const spoken = String(reply?.text || '').trim() || postTool.trim() || streamed.trim();
      if (reply && spoken && spoken !== String(reply.text || '').trim()) return { ...reply, text: spoken };
      return reply;
    } finally {
      offChunk?.();
      offToolCall?.();
    }
  }

  cancel() { this.session?.cancel(); }

  async resetContext() {
    try { this.session?.cancel?.(); await this.session?.dispose?.(); }
    finally {
      this.session = null;
      if (this._acquired) { releaseQvac(); this._acquired = false; }
    }
  }

  async close() {
    try { await this.resetContext(); } finally { await closeQvac(); }
  }
}