Skip to content

Meet PULSE: Your AI Teammate That Actually Works

Most AI tools wait for you to ask a question. PULSE is different — it is Buddy AI's autonomous virtual employee: a teammate with a name, a role, a task queue, and a memory that survives restarts. You onboard it like a new hire, teach it your systems, assign work, and it keeps going — including when you close the browser.

Install Buddy AI

pip install buddy-ai[all]

Chatbot vs virtual employee

Typical chatbot PULSE
Trigger You send a prompt Tasks sit in a queue; PULSE picks them up
Output Text in a chat window Real files (.py, .md, .sql, …) in a workspace
Knowledge Whatever fits in context KT from PDFs, URLs, and live sessions — retained permanently
Continuity Often forgets between sessions Long-term memory + conversation history across restarts
Reporting None Daily standups, activity feed, notifications

PULSE is built for doing work, not just answering questions. The Competency Engine (new in v2.2.0) complements it by measuring who on your team is strongest in each domain — PULSE is the teammate that actually executes.

What PULSE can do

Once you create an employee profile, PULSE operates as a persistent team member:

Capability What it means in practice
Identity Name, role, department, skills, timezone — PULSE introduces itself and behaves consistently
Knowledge Transfer (KT) Learn from PDFs, documents, full website crawls (BFS, up to 20 pages), or live Socratic Q&A until ≥82% confidence
Meetings Paste a transcript → summary, decisions, action items; PULSE's own items land on the Task Board
Tasks Kanban board with autonomous execution — highest-priority todo picked up within ~30 seconds
Chat Streaming conversation grounded in KT + long-term memory + recent history
Knowledge Explorer Ask what PULSE knows about any topic it has been trained on
Autonomous worker Background loop writes files, sends notifications, generates standups, suggests proactive tasks

All employee data persists under ~/.pulse_data/ — profiles, KT sessions, tasks, memories, and workspace files survive server restarts.

Honest framing

PULSE is a real autonomous agent layer on top of an LLM — not magic. It needs a configured model and API key (or Ollama locally). Task quality depends on your prompts, KT depth, and the model you choose. It creates files and manages a queue; it does not replace code review, security audits, or human judgment on production changes.

Start the web UI

One command launches the full dashboard:

buddy pulse start
# Opens http://localhost:8888

Useful flags:

buddy pulse start --port 3000 --no-browser   # custom port, no auto-open
buddy pulse start --host 0.0.0.0               # bind for LAN access

On first launch, the onboarding wizard walks you through creating your virtual employee. After that, nine sidebar pages cover day-to-day use:

Page Purpose
Dashboard Activity feed, current work, standup, proactive suggestions, knowledge domains
KT Center Upload documents, crawl URLs, or start live KT sessions
Meetings Process transcripts and extract action items
Tasks Kanban board with auto/manual execution toggle
Workspace Browse, view, and download files PULSE created
Memory Long-term facts and conversation history — view, search, forget
Chat Streaming WebSocket chat (KT + memory aware)
Knowledge Explore what PULSE has learned across KT sessions
Settings LLM provider, API key, auto-work toggle — no env vars required

Configure your model under Settings → AI Model & API Keys. Supports OpenAI, Anthropic, Google, Groq, and Ollama — changes apply immediately.

The REST API is available at http://localhost:8888/api/docs while the server runs. See the CLI reference for buddy pulse create, buddy pulse kt, and buddy pulse status.

Desktop app — PULSE that stays running

Closing a browser tab stops the web UI session. The Buddy AI Desktop app (pulse-desktop/) wraps the same PULSE stack in a native Electron shell for Windows and macOS:

  • System tray — closing the window hides the app; the autonomous worker keeps running
  • Same dashboard — loads http://127.0.0.1:8888 served by buddy pulse start
  • Auto-start backend — spawns buddy pulse start --no-browser on launch

Prerequisites

  1. Python 3.10+ with Buddy AI: pip install buddy-ai[all]
  2. Built React UI (first time, from repo root):
cd pulse-ui
npm install
npm run build
  1. Node.js 18+ for the desktop app

Run in development

cd pulse-desktop
npm install
npm start

Build installers

Platform Command Output
Windows npm run build:win Buddy AI Setup *.exe in pulse-desktop/dist/
macOS npm run build:mac Buddy AI-*.dmg in pulse-desktop/dist/ 
cd pulse-desktop
npm install
npm run build:win   # Windows — run on a Windows PC
npm run build:mac   # macOS — must run on a Mac

Installers need Python

Installers bundle the Electron shell. Users still need Python and buddy-ai installed (pip install buddy-ai[all]), unless you add a bundled Python sidecar later. Full details: pulse-desktop/README.md.

Windows tip: If you hit symlink or code-signing errors during build:

$env:CSC_IDENTITY_AUTO_DISCOVERY="false"
npm run build:win

Useful environment variables: PULSE_PORT, BUDDY_PYTHON, PULSE_DATA_DIR. See the desktop README for troubleshooting.

Python API — quick start

Everything in the UI is also available programmatically. This example mirrors examples/11_pulse_employee.py:

import os

from buddy.models.openai import OpenAIChat
from buddy.pulse import EmployeeProfile, KTSourceType, PulseEmployee

if not os.getenv("OPENAI_API_KEY"):
    print("Set OPENAI_API_KEY to run this example.")
else:
    pulse = PulseEmployee(
        employee_profile=EmployeeProfile(
            full_name="Priya Sharma",
            role="Senior Backend Engineer",
            department="Engineering",
            skills=["Python", "FastAPI", "PostgreSQL"],
            timezone="Asia/Kolkata",
            company_name="Acme Corp",
        ),
        model=OpenAIChat(id="gpt-4o"),
    )

    print(pulse.introduce_yourself())

    # Document-based KT (text or file path)
    summary = pulse.take_kt(
        source=b"Our auth uses JWT with 1-hour expiry and Redis blocklist on logout.",
        source_type=KTSourceType.DOCUMENT,
        session_name="Auth Service Overview",
        knowledge_giver="Arjun Nair",
    )
    print(summary.format_summary())

    # Assign a task
    task = pulse.receive_task(
        title="Implement Razorpay refund flow",
        description="Handle refund_processed webhook events with idempotency.",
        assigned_by="Arjun Nair",
        priority="high",
    )
    print(task.format_brief())

For live KT, meetings, and status reports, see the full PULSE reference or run the complete example script.

To launch the web server from Python:

# from buddy.pulse.app import PulseApp
# PulseApp(employee=pulse).serve()
# Or simply: buddy pulse start

What you need to know before you start

  • LLM API key required for autonomous work, KT, chat, and meetings — configure in Settings or via OPENAI_API_KEY, ANTHROPIC_API_KEY, etc.
  • Costs scale with usage — the background worker polls every 30 seconds; each task, KT session, and chat turn calls your model.
  • Local data — everything lives in ~/.pulse_data/; back it up if employees matter to your workflow.
  • Ollama works — run locally with no cloud key if you prefer, via Settings or a local Ollama instance.
  • Not a substitute for review — PULSE generates real files; treat output like any junior engineer's PR.

Next steps

pip install buddy-ai[all]
buddy pulse start

Give PULSE a role, teach it your stack, add a task — and let your virtual teammate get to work.