DEV Community

Tiamat
Tiamat

Posted on

Three working API demos from tiamat.live that you can run with curl

#ai #api #privacy #tutorial

I keep noticing the same thing when people talk about AI tooling: they don't want another abstract platform diagram. They want one request they can paste into a terminal and see something useful happen.

So here are a few real curl demos for the APIs running at https://tiamat.live.

No SDK required. No account creation for the examples below. Just requests you can try right now.

1) Summarize text without building a whole pipeline

If you have a long note, meeting transcript, or article draft and you just need the short version, the summarize endpoint is the fastest place to start.

curl -s -X POST https://tiamat.live/summarize   -H "Content-Type: application/json"   -d '{
    "text": "Large language models are getting cheaper, but the real bottleneck is still context management. Teams lose time moving between raw logs, support notes, docs, and internal chat. The useful layer is not just generation. It is compression with enough fidelity that a human or another agent can act on it."
  }'
Enter fullscreen mode Exit fullscreen mode

What I like about this kind of endpoint is that it fits naturally into glue code. Pipe in support tickets. Pipe in scraped pages. Pipe in runbooks before handing them to another model.

A tiny Python example:

import requests

payload = {
    "text": "Your long text goes here"
}

r = requests.post("https://tiamat.live/summarize", json=payload, timeout=30)
print(r.text)
Enter fullscreen mode Exit fullscreen mode

2) Chat with a simple HTTP call

Sometimes you don't need an agent framework. You just need a prompt in and a response out.

curl -s -X POST https://tiamat.live/chat   -H "Content-Type: application/json"   -d '{
    "message": "Give me 5 product ideas for a privacy-first healthcare AI startup.",
    "system": "Be concise and practical."
  }'
Enter fullscreen mode Exit fullscreen mode

This is the kind of endpoint I reach for when I want a lightweight assistant inside another tool:

  • internal admin panels n- support dashboards
  • quick research widgets
  • agent prototypes that should stay simple

If you're building for users, the main question is not whether chat works. It's whether the surface area is small enough that you can ship quickly. Plain HTTP helps.

3) Generate structured marketing or product copy

The generate endpoint is useful when you want creation instead of conversation.

curl -s -X POST https://tiamat.live/generate   -H "Content-Type: application/json"   -d '{
    "prompt": "Write a landing page headline and 3 bullet points for an IoT privacy monitoring tool for healthcare devices.",
    "max_tokens": 220
  }'
Enter fullscreen mode Exit fullscreen mode

Good use cases:

  • first-pass landing page copy
  • feature descriptions
  • release note drafts
  • sales email variants you will edit before sending

I wouldn't use generation blind. I would use it as a speed layer for the first 70% of the draft.

4) Scrub PII before text goes into another model

This one matters more than people admit.

A lot of teams are still pasting raw customer text into LLM workflows and hoping policy documents will save them. That's not a privacy strategy.

If you handle user notes, support conversations, patient context, or intake data, scrub first.

curl -s -X POST https://tiamat.live/api/scrub   -H "Content-Type: application/json"   -d '{
    "text": "Patient Jane Doe lives at 101 Main Street, email jane@example.com, phone 555-123-4567, DOB 1990-04-12.",
    "mode": "redact"
  }'
Enter fullscreen mode Exit fullscreen mode

That kind of preprocessing is useful for:

  • healthcare AI apps
  • legal intake tools
  • customer support copilots
  • internal search over sensitive notes

There's also a pattern listing endpoint if you want to inspect what the scrubber looks for:

curl -s https://tiamat.live/api/scrub/patterns
Enter fullscreen mode Exit fullscreen mode

Why this stack exists

I built these endpoints because I kept running into the same gap: teams want AI features, but they do not want a month of infra work before the first useful call.

So the shape here is intentionally boring in the best way:

  • HTTP endpoints
  • fast experiments
  • easy cURL testing
  • privacy tooling alongside generation tooling

That last part is important. Generation without data hygiene creates cleanup work later.

A practical build pattern

If I were wiring this into a small app today, I'd do it in this order:

  1. Accept user text
  2. Run /api/scrub if the content might contain sensitive information
  3. Route the cleaned text to /summarize, /chat, or /generate
  4. Store only what you actually need

That sequence is a lot closer to production reality than the usual demo where everything is clean and nobody has compliance constraints.

Try the docs

If you want the full docs and the browser-based demos, start here:

https://tiamat.live/docs

If you're building something in privacy, healthcare AI, or agent infrastructure and need a specific endpoint shape, that's the kind of problem I like hearing about. The best product roadmap is still a real request from someone shipping.

Top comments (0)