Setup Guide

From "I just received my license" to "live AI chatbot on my website" in ~10 minutes.

Audience. Anyone with admin access to an AWS account. No coding required. No DevOps experience required. If you can click "Next" three times in the AWS Console, you can deploy this.

Time required. ~10 minutes — 4 minutes of clicks + 4 minutes waiting for AWS to build everything + 2 minutes of basic config.

Before you start

You need:

You will pay AWS directly for usage on your own account. Typical bill is $5–$50/month depending on traffic. We never see, touch, or take a cut of your AWS bill.

0Enable Bedrock model access

This is the single non-obvious step. AWS requires you to explicitly enable each AI model before you can use it.

  1. Sign into your AWS Console at aws.amazon.com/console
  2. Search for Bedrock at the top → click "Amazon Bedrock"
  3. Confirm the region (top-right) matches what we sent you — typically eu-central-1 (Frankfurt)
  4. Sidebar → Model access
  5. Click Modify model access (top-right)
  6. Tick: Cohere Embed Multilingual, Amazon Nova Pro, Anthropic Claude Sonnet (any version)
  7. Click Next → Submit
  8. Wait ~30 seconds. Status changes to Access granted for each.
SCREENSHOT: Bedrock Model access page with the three required models showing green "Access granted" badges
Why this is needed. AWS requires this for compliance and billing. It's free to enable — you only pay when you use the models. Once granted, access stays granted forever.

If a model shows "Available to request" instead of "Access granted", click the request button. Most accounts get instant approval; some new accounts need 1-2 hours.

1Click the deploy URL

The email we sent contains a single link. Click it. You'll land in your AWS CloudFormation console with every parameter already filled in.

SCREENSHOT: CloudFormation Create Stack page with all params pre-filled

If you see a "Sign in" page first, sign in with your AWS account credentials and the deploy URL will reload.

2Verify the parameters

Scroll down. You'll see fields with values already entered:

  • Stack name — typically <your-company>-ai-assistant
  • ProjectName — typically <your-company>-ai
  • LicenseKey — your one-time license key (starts with LK-)
  • ValidationUrl — our license server URL (don't change this)
  • Features — the feature list you bought
  • Tier — ultra (legacy compatibility, can ignore)
  • ParsingModelId — amazon.nova-pro-v1:0 (default)
Don't change anything. The link we sent has everything correct. If you change ProjectName here, you'll need a fresh license key. If in doubt, ask us.

Scroll to the bottom. Tick the two acknowledgement boxes, then click Create stack.

SCREENSHOT: bottom of CFN form with the two acknowledgement boxes ticked + Create Stack button

3Wait ~4 minutes

AWS now provisions ~30 resources for you: Lambda functions, S3 buckets, DynamoDB tables, an HTTP API, a CloudFront distribution, a Bedrock Knowledge Base, the chat handler.

The page shows stack events updating in real time. Just watch.

SCREENSHOT: stack events page showing CREATE_IN_PROGRESS for various resources

When done, the stack status at top changes to CREATE_COMPLETE with a green check.

If you see CREATE_FAILED: check the events table for the first failed resource (highlighted in red). Most common cause is missing Bedrock model access — go back to Step 0. Email [email protected] if you're stuck — we'll help.

4Get your admin console URL

Stack CREATE_COMPLETE → click the Outputs tab. You'll see a table:

  • AdminConsoleUrl — your CloudFront domain — this is where you log in to manage the bot
  • WidgetJsUrl — the script tag URL for embedding on your website
SCREENSHOT: Outputs tab with AdminConsoleUrl + WidgetJsUrl rows

Copy AdminConsoleUrl, open it in a new tab.

5Get your admin API key

The admin console asks for an API key on first visit. Two ways to get it:

Option A — AWS Console

  1. Search for Systems Manager at the top
  2. Sidebar → Parameter Store
  3. Search box → type your project name (e.g. acme-ai)
  4. Click /<project>/admin-api-key
  5. Click Show to reveal the value, copy it
SCREENSHOT: SSM Parameter Store with /<project>/admin-api-key parameter, Show button

Option B — AWS CLI (faster if you have it set up)

aws ssm get-parameter \
    --name /<your-project>/admin-api-key \
    --with-decryption \
    --region eu-central-1 \
    --query 'Parameter.Value' \
    --output text

6Log in to the admin console

Go to your AdminConsoleUrl tab. Paste your API key into the prompt → click Continue.

You're in. The dashboard shows total queries (zero so far), bot list (one bot called default), and quick actions.

SCREENSHOT: admin console first-load Overview tab with empty stats

7Upload your first documents

Critical step — without documents, the bot has nothing to answer.

  1. Sidebar → Knowledge
  2. Drag-and-drop your files, or click Upload and pick
  3. Supported: PDF, DOCX, XLSX, CSV, TXT, HTML, Markdown, PPTX (PPTX requires Pro+)

Files appear with status:

  • ⏳ Syncing — being indexed (30–60s per document)
  • ✅ Indexed — ready to answer questions
SCREENSHOT: Knowledge tab with 3 files uploaded, all showing Indexed status
Tip. Start with 5-10 of your most-asked-about documents. You can always add more later. Quality > quantity for accuracy.

8Test the bot

  1. Click the floating chat icon bottom-right of any admin page (live preview)
  2. Ask a question that should be answerable from your documents
  3. Verify the answer is correct + source citations appear below
SCREENSHOT: chat widget open with test query + response showing source citation

If the bot answers wrong or says "I don't know":

  • Make sure the document is Indexed (not Syncing)
  • Try rephrasing the question
  • Check if the document is searchable text — scanned-image PDFs need OCR (we handle that automatically but it takes longer)

9Embed the widget on your website

Sidebar → Integrations. Copy the embed snippet:

<script src="https://<your-cloudfront>/widget.js" defer></script>

Paste it anywhere in the <body> of your website. Common places:

  • WordPress: Appearance → Theme → footer.php (just before </body>)
  • Shopify: Online Store → Themes → Edit code → theme.liquid (just before </body>)
  • Squarespace: Settings → Advanced → Code Injection → Footer
  • Webflow: Project Settings → Custom Code → Footer Code
  • Plain HTML: paste before the closing </body>
SCREENSHOT: Integrations tab with embed snippet + copy button

Save and reload your website. Widget appears bottom-right.

Tip. If the widget doesn't appear, check browser dev console (F12). Most common cause: a Content Security Policy on your site blocking the script — talk to your web dev or email us.

10Configure (optional)

The admin console has tabs for:

  • Appearance — logo, colors, welcome message, position, language defaults
  • Behavior — AI tone, language, monthly query cap
  • Bots — additional bots if your plan supports multi_bot (Pro: 3, Max: 10)
  • Integrations — Slack, Teams, Jira, webhooks (varies by plan)
  • Analytics — query stats, gap detection, lead capture (varies by plan)
  • License — what features your license includes + your live AWS bill
SCREENSHOT: Appearance tab with brand color picker + welcome message field

11Set up a budget alarm (recommended)

2-minute safety net against unexpected AWS bills.

  1. AWS Console → search Billing and Cost Management
  2. Budgets → Create budget
  3. Cost budget
  4. Set: name nohow-ai-monthly-cap, period Monthly, amount $50
  5. Add an email alert at 80%
  6. Create budget

Now if your AWS bill ever hits $40, you get an email — plenty of time to investigate.

SCREENSHOT: AWS Budgets create form filled in with $50 monthly + email alert
Tip. Set the budget on a tag Project=<your-project>. Our deployment auto-tags every resource with this. The budget then tracks ONLY this stack's spend.

You're done

Your bot is live. Visitors can chat with it. The admin console shows you everything that's happening. Your AWS bill is under control.

What to do next (in order of usefulness)

  1. Test from your customer's perspective. Open your website in incognito, ask 5-10 real questions, see how answers feel.
  2. Add more documents as you find gaps. The Knowledge tab updates the bot in real time — no redeployment.
  3. Tune the welcome message + starter questions in Appearance. First thing visitors see.
  4. Look at Analytics after a week of real traffic. The "Gaps" tab shows you questions visitors asked that the bot couldn't answer — add documents to fill those gaps.
  5. Try different AI quality levels if you have premium / max access. Behavior tab lets you change quality per-bot.

FAQ

My bot answers are wrong / generic.

My AWS bill seems high.

Can I move to a different region later?

Yes, but it requires a new deployment. Email us — there's a migration script that copies your KB content over.

Can I have multiple bots in my deployment?

Yes, if your plan supports multi_bot. Click Bots → + New Bot.

How do I update to a newer version?

Click your deploy URL again — it always points at the latest release. CloudFormation will detect what's changed and update only those resources. Takes 1-2 minutes. No data loss.

How do I cancel?

How do I get help?

Stuck somewhere?

We respond within 2 hours during EU business hours. Mention your project name and where you got stuck.

Email [email protected] →