Quickstart

In five minutes you'll have a bot answering questions from your own documents on a live web page.

You need:

A ScaiGrid API key with scaibot:bots:create (any tenant admin has this).
A PDF, Markdown, or text document to feed the bot.
A web page where you can paste an HTML snippet.

bash
export SCAIGRID_HOST="https://scaigrid.scailabs.ai"
export SCAIGRID_API_KEY="sgk_..."

1. Create the bot#

bash
curl -X POST "$SCAIGRID_HOST/v1/modules/scaibot/bots" \
  -H "Authorization: Bearer $SCAIGRID_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Quickstart Bot",
    "slug": "quickstart",
    "model": "scailabs/poolnoodle-omni",
    "display_name": "Helper",
    "welcome_message": "Hi! Ask me anything about the handbook.",
    "knowledge_enabled": true,
    "knowledge_mode": "managed"
  }'

python
import httpx, os
bot = httpx.post(
    f"{os.environ['SCAIGRID_HOST']}/v1/modules/scaibot/bots",
    headers={"Authorization": f"Bearer {os.environ['SCAIGRID_API_KEY']}"},
    json={
        "name": "Quickstart Bot",
        "slug": "quickstart",
        "model": "scailabs/poolnoodle-omni",
        "display_name": "Helper",
        "welcome_message": "Hi! Ask me anything about the handbook.",
        "knowledge_enabled": True,
        "knowledge_mode": "managed",
    },
).json()["data"]
print(bot["id"])

javascript
const res = await fetch(`${process.env.SCAIGRID_HOST}/v1/modules/scaibot/bots`, {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.SCAIGRID_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    name: "Quickstart Bot",
    slug: "quickstart",
    model: "scailabs/poolnoodle-omni",
    display_name: "Helper",
    welcome_message: "Hi! Ask me anything about the handbook.",
    knowledge_enabled: true,
    knowledge_mode: "managed",
  }),
});
const { data: bot } = await res.json();
console.log(bot.id);

Save the returned bot.id — you'll need it for the embed.

2. Upload a document#

bash
curl -X POST "$SCAIGRID_HOST/v1/modules/scaibot/bots/$BOT_ID/documents" \
  -H "Authorization: Bearer $SCAIGRID_API_KEY" \
  -F "file=@handbook.pdf" \
  -F "name=Product Handbook"

Indexing is asynchronous. Poll the document until status: "indexed":

bash
curl "$SCAIGRID_HOST/v1/modules/scaibot/bots/$BOT_ID/documents/$DOC_ID" \
  -H "Authorization: Bearer $SCAIGRID_API_KEY"

For a handful of pages, this completes in seconds.

3. Generate an embed token#

bash
curl -X POST "$SCAIGRID_HOST/v1/modules/scaibot/bots/$BOT_ID/embed-token" \
  -H "Authorization: Bearer $SCAIGRID_API_KEY"

You'll get back a short-lived token (~1 hour by default). For a production embed you'll typically generate fresh tokens server-side per visitor; for this quickstart, copy it once.

Paste into any page on your site:

html
<script
  src="https://scaigrid.scailabs.ai/static/modules/scaibot/widget.js"
  data-bot-id="bot_abc123"
  data-token="sbt_short_lived_token"
  async>
</script>

Refresh. A floating chat bubble appears in the corner. Click it, ask a question from the handbook, watch the bot answer with a citation back to the document.

5. Inspect the conversation#

Back in the API:

bash
curl "$SCAIGRID_HOST/v1/modules/scaibot/conversations?bot_id=$BOT_ID&limit=5" \
  -H "Authorization: Bearer $SCAIGRID_API_KEY"

You'll see every chat session, with messages, latencies, token counts, and which knowledge chunks were retrieved.

What just happened#

The bot you created is a row in ScaiGrid's database with its config and a reference to a knowledge collection.
Each PDF you uploaded was chunked into ~500-token slices, embedded with ScaiGrid's embedding model, and stored in ScaiMatrix.
The widget's JavaScript opens an authenticated session with the bot's API, streams answers via Server-Sent Events, and persists the conversation id in a visitor cookie.
Every chat call is metered by ScaiGrid's accounting pipeline against your tenant's budget.

Next#

Configure tone — formality, verbosity, vocabulary in scope.
Add escalation rules so the bot hands off when it should.
See the full embed tutorial for production-grade token issuance, custom styling, and authenticated visitors.