Quick Start

This walkthrough takes you from a fresh install to a real POST /api/{uuid}/chat call hitting OpenAI through PromptGate. You will:

1. Log in.
2. Add an OpenAI credential.
3. Create a project and an AI endpoint.
4. Issue a token.
5. Call the endpoint with curl / Python / Node.

If you haven’t installed PromptGate yet, do that first: Installation.

1. Log in and change the default password

Open http://localhost:8000. Log in with:

Email:    admin@promptgate.dev
Password: admin

Click your name in the top-right corner → Profile → change the password. (You only need to do this once.)

2. Add a provider credential

Top-right user menu → Credentials. Click New credential.

• Name: OpenAI Production
• Provider: OpenAI
• Secret: paste your sk-… key

Save. The secret is encrypted with AES-256-GCM at rest. You can see the prefix only after this point.

Don’t have an OpenAI key?

You can use any of the 8 built-in providers — Anthropic, Google Gemini, Mistral, Groq, Together AI, Cohere, or local Ollama. The provider in step 2 picks the adapter; the rest of the flow is identical. See Providers Overview.

3. Create a project

Sidebar → Projects → + New project.

• Name: Quickstart
• Type: AI Gateway
• Environment: prod

You’ll land on the project switcher; click your new project to enter it. The sidebar now shows project-scoped items: AI Endpoints, Playground, Live Logs, Metrics, Guardrails, API Tokens, Webhooks.

4. Create an AI endpoint

Project sidebar → AI Endpoints → + New endpoint.

The endpoint wizard has 7 tabs, but you only need to fill out two for this walkthrough:

Tab 1 — Core:

• Name: Hello World
• (Slug auto-fills as hello-world)

Tab 2 — Provider:

• Mode: Manual
• Provider: OpenAI
• Model: gpt-4o-mini
• Credential: pick the one you just added

Skip Limits, Streaming, Sessions, Prompt, Schema for now — defaults are fine. Hit Create.

You’ll land on the endpoint detail page. Click the eye icon next to the slug to see the endpoint URL — something like:

POST http://localhost:8000/api/8e3f...c2/hello-world

5. Issue an API token

Project sidebar → API Tokens → + New token.

• Name: quickstart
• Environment: live
• Scopes: tick chat

Save. The plaintext token (pg_live_…) is shown once. Copy it now.

One-time display

PromptGate stores SHA-256 hashes only — you can’t recover the plaintext later. If you lose it, rotate the token to mint a new one.

6. Call the endpoint

Replace <UUID> with your project’s UUID (visible in the endpoint detail URL) and <TOKEN> with the plaintext you just copied.

curl

curl -X POST http://localhost:8000/api/<UUID>/hello-world \
  -H "Authorization: Bearer <TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"message": "Say hi in one sentence."}'

Python

import os, requests

resp = requests.post(
    f"http://localhost:8000/api/{os.environ['PG_UUID']}/hello-world",
    headers={"Authorization": f"Bearer {os.environ['PG_TOKEN']}"},
    json={"message": "Say hi in one sentence."},
)
resp.raise_for_status()
print(resp.json())

Node.js

const uuid = process.env.PG_UUID;
const token = process.env.PG_TOKEN;

const r = await fetch(`http://localhost:8000/api/${uuid}/hello-world`, {
    method: 'POST',
    headers: {
        'Authorization': `Bearer ${token}`,
        'Content-Type': 'application/json',
    },
    body: JSON.stringify({ message: 'Say hi in one sentence.' }),
});
console.log(await r.json());

You should see something like:

{
  "ok": true,
  "id": "chatcmpl-...",
  "model": "gpt-4o-mini",
  "content": "Hi there! How can I help today?",
  "finish_reason": "stop",
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 9,
    "total_tokens": 23
  }
}

7. Watch it in the UI

While that request is firing, navigate to Live Logs in the project sidebar — you’ll see the request appear in real time, with status, latency, and token counts. Metrics rolls these up into 24h / 7d charts, and Audit Log records the create/use of the endpoint and token.

What you just did

• ✅ Encrypted a provider credential at rest
• ✅ Created a project-scoped AI endpoint with a fixed provider + model
• ✅ Issued a scoped, hashed API token
• ✅ Routed a chat completion through the gateway

Next steps

• AI Endpoints — system prompts, schemas, sessions, streaming, failover.
• Guardrails — turn on PII redaction, prompt-injection blocking, keyword blocklists.
• Rate Limits & Budgets — caps per-minute, per-hour, per-month.
• AI Wrapper — point any OpenAI SDK at the gateway directly.
• Cookbook — task-oriented walkthroughs.

---

© Akyros Labs LLC. All rights reserved.