live

bayar.dev

I build the layer between your users and the model.

I'm Bayar, a solo developer in Walnut Creek, California, building AI infrastructure in public. The terminal is a real endpoint, and every number here is measured.

livea streaming AI terminal, try it

nextedge routing, caching, guardrails

try the terminal ↓read the log →see the numbers
IDLE

bayar.dev terminal — a real model behind a real endpoint. Type anything, or start with:

benchmark --latency
measured numbers, not claims
architecture --stack
what this terminal runs on
roadmap
what I'm building next
whoami
who is behind this

Up to 4,000 characters per prompt.

01 / live

Follow your request

Send a prompt in the terminal to watch it move.

  1. Your prompt

    Sent from the terminal to POST /api/chat.

  2. Rate limit

    10 requests a minute and 100 a day, per IP address.

  3. Validate

    Roles, part sizes and a 50,000-character cap are checked before any model call.

  4. Model

    Llama 3.3 70B on Cloudflare Workers AI.

  5. Stream back

    Tokens arrive as they are generated. Stop any time.

On the roadmap, not shipped

  • Provider failover
  • PII guard
  • Semantic cache
  • Private VPC and BYOK

02 / measured

Measured, not claimed

measured · the terminal model · short chat · last 7 days

Time to first token, p50
…
Output speed, p50
…
Success rate
…
Runs, all models
…

Probes run every 15 minutes. Method, every model, and raw data →

03 / log

Latest from the log

all entries →

04 / updated October 30, 2026

Now

Building on the model layer too? Say hi.

I read everything that lands in my inbox and reply to what I can.

hi@bayar.dev