Trodo

You'll create a prompt, label a version production, and fetch it from your app. Five minutes.

1. Create the prompt

Open Prompts in the sidebar and click New prompt. Give it a name your app will fetch by — say refund-agent. Names are slug-ish (a-z, 0-9, -, _, .).

In the Messages editor, write a system message and a user message. Use {{double braces}} for anything filled at runtime:

system

You are a refund support agent for {{company}}.
Be concise and never promise a refund you can't confirm.

user

{{question}}

Open the Model config section and set a provider and model as JSON, e.g. { "provider": "openai", "model": "gpt-4o" }. Open the Variables section — company and question were auto-detected; optionally give company a default like your company.

Click Save version, and in the dialog apply the production label. You've shipped v1.

2. Fetch it from your app

Install the SDK if you haven't:

npm install trodo-node

pip install trodo-python

Then fetch and compile. get follows the production label by default; compile fills the variables and returns a ready-to-send payload:

import trodo from 'trodo-node';
trodo.init({ siteId: process.env.TRODO_SITE_ID! });

const prompt = await trodo.prompts.get('refund-agent');

const { messages, model } = prompt.compile({
  company: 'Acme',
  question: 'where is my order?',
});

const answer = await openai.chat.completions.create({
  model: model.model ?? 'gpt-4o',
  temperature: model.temperature ?? 0.2,
  messages,
});

import trodo
trodo.init(site_id=os.environ["TRODO_SITE_ID"])

prompt = trodo.get_prompt("refund-agent")

compiled = prompt.compile(
    company="Acme",
    question="where is my order?",
)

answer = openai.chat.completions.create(
    model=compiled.model.get("model", "gpt-4o"),
    temperature=compiled.model.get("temperature", 0.2),
    messages=compiled.messages,
)

get hands you the whole prompt with its messages still holding {{company}} and {{question}}; compile(values) fills those in and returns { messages, model, tools, response_format }, with the messages array already in the shape providers expect — you don't rebuild it. (model / tools / response_format are carried straight through from the prompt.) See Concepts for the exact split.

3. Ship a change

Edit the prompt, Save version, and move production to the new version. Your running app picks it up within one cache window (60s by default) — no redeploy.

Iterating on the wording? Do it in the Playground first — run the prompt across models side by side and score the outputs — then save the winner as a version here.

Concepts — the data model and the exact shapes get and compile return.
Initialization & core calls — everything the editor can do.
Variables & templating — defaults and the template syntax.
Caching & availability — make the fetch resilient before you go to production.

Quickstart

1. Create the prompt

2. Fetch it from your app

3. Ship a change

Next

On this page