Personal AI Site

A template for building a chat-first professional portfolio. Instead of reading a static resume, visitors chat with an AI version of you — it answers questions about your career, checks your Google Calendar availability, books calls (with a two-phase approve/reject flow), and takes messages through a contact form. All features are optional and can be toggled independently.

⚠️ Cost warning: The AI chat makes real LLM API calls (OpenAI or via the OpenKBS proxy). Each conversation costs tokens. Monitor your usage and set billing limits on your provider account.

Prerequisites

• Node.js 24+ (LTS recommended)
• npm 10+
• An OpenAI API key or an OpenKBS account (includes proxy credits)
• (Optional) Google Cloud project with OAuth 2.0 Client ID + Service Account for calendar/booking features

Quick start

git clone https://github.com/open-kbs/askme.git my-site
cd my-site
npm install

Agent-assisted setup (recommended)

Open the repo in a coding agent (Claude Code, Cursor, Codex) and say "set up this project". The agent reads SETUP.md and walks you through profile, avatar, LLM key, and optional Google credentials — one question at a time.

Manual setup

1. Copy .env.example to .env.local and fill in your keys.
2. Edit config.json with your name, title, timezone, and branding.
3. Edit assets/career.json with your experience, skills, and projects.
4. (Optional) Place an avatar image at assets/avatar.png.

Then start the dev server:

npm run dev

This starts two processes:

• API server at http://localhost:8787 (Hono + PGlite for local Postgres)
• Frontend at http://localhost:5173 (Vite + React)

Project structure

config.json              # Owner profile, feature flags, system prompt, email templates
.env.local               # Secrets (gitignored)
assets/
  career.json            # Structured career data (fed into AI system prompt)
  avatar.png             # Profile photo
functions/
  api/                   # Main Lambda — chat, bookings, calendar, contact
  api-cleanup/           # Hourly cron — cleans expired rate-limit rows
  _shared/               # Shared modules (db, auth, emails, availability, …)
site-src/                # Vite + React + Tailwind frontend
local/
  server.mjs             # Local dev server (Hono, PGlite)
  setup.mjs              # First-time setup API (loopback only)
scripts/
  bundle-functions.sh    # Pre-deploy: copies _shared + config into each function

API

All requests go to POST /api with a JSON body. The action field selects the endpoint.

chat — AI conversation

{
  "action": "chat",
  "messages": [
    { "role": "user", "content": "Tell me about your work" }
  ]
}

Returns { "text": "...", "toolParts": [...] }. The AI can call tools during the conversation:

Tool	Description
checkAvailability	Returns 7 days of 30-min free/busy slots
createBooking	Creates a pending booking request
sendMessage	Sends a contact message to the site owner

get-availability — calendar slots

Requires Authorization: Bearer <google-id-token>.

{ "action": "get-availability" }

Returns { "days": [{ "date": "01-05-2026", "dayOfWeek": "Fri", "slots": [...] }] }.

create-booking — book a call

Requires Authorization: Bearer <google-id-token>.

{
  "action": "create-booking",
  "name": "Jane Doe",
  "email": "jane@example.com",
  "date": "01-05-2026",
  "startTime": "14:00",
  "duration": 30
}

send-contact — contact form

{
  "action": "send-contact",
  "name": "John Smith",
  "email": "john@example.com",
  "message": "I'd like to work together."
}

Configuration

config.json

Section	What it controls
features	Toggle calendar, bookings, contactForm independently
owner	Name, title, location, timezone, working hours
branding	Site name, logo text, meta description, avatar/CV URLs
social	LinkedIn, GitHub, Twitter, website links in nav
starterPrompts	Quick-action buttons shown in empty chat
systemPrompt	AI persona, guidelines, scope rules
emails	Mustache-lite templates for booking and contact emails

.env.local

See .env.example for all supported variables. Key ones:

Variable	Required	Description
OPENAI_API_KEY	One of these	Direct OpenAI access
OPENKBS_API_KEY	One of these	OpenKBS proxy (multi-vendor)
GOOGLE_OAUTH_CLIENT_ID	For calendar	Google OAuth client ID
GOOGLE_SERVICE_ACCOUNT_KEY	For calendar	Base64-encoded service account JSON
GOOGLE_CALENDAR_IDS	For calendar	Comma-separated calendar emails
BOOKING_SECRET	For bookings	HMAC key for approve/reject links

Deploying to OpenKBS

npm run deploy

This builds the frontend, bundles shared modules into each function, and deploys via the OpenKBS CLI. See openkbs.json for the deployment configuration.

Troubleshooting

Chat returns an error about the API key Check that .env.local has either OPENAI_API_KEY or OPENKBS_API_KEY set. Restart the dev server after changing env vars.

Google sign-in button doesn't appear Ensure GOOGLE_OAUTH_CLIENT_ID is set. In Google Cloud Console, verify that http://localhost:5173 (no trailing slash, http not https) is listed under "Authorized JavaScript origins". If your OAuth consent screen is in "Testing" mode, your Google account must be listed as a test user.

Calendar shows no availability The service account must be shared on the target Google Calendar (Calendar Settings → Share → add the service account email with "See all event details"). Also verify the Google Calendar API is enabled in your Cloud Console project.

Booking approval link says "expired" Approval links expire after 7 days. The booking must also still be in pending status.

npm run dev fails on first run Run npm install first — postinstall also installs site-src dependencies.

Roadmap

• AI chat with career-aware system prompt
• Google Calendar availability + booking flow
• Two-phase booking approval via email
• Contact form with rate limiting
• Dark/light theme
• Local dev with PGlite (zero external dependencies)
• Custom themes and color schemes
• Multi-language support
• Analytics dashboard for site owners

License

MIT — see LICENSE.