DQaaS beta walk-through
Thank you for testing DQaaS. It profiles a data file, checks it against quality rules and gives you a score and a report. This takes about 30 minutes, plus 15 for the optional command line and MCP parts. Please work through the steps in order, tick each box, and note anything that is slow, confusing or wrong.
Before you start. Your invitation email comes from the DQaaS no-reply address; check your spam folder if you cannot find it. Use a private browser window and sign in with the GitHub account the invitation names.
Please do not upload personal data about real people, health, financial or other sensitive records. Use a spreadsheet export with 10 or more rows of harmless data, such as product lists, orders or made-up customers. A CSV, Excel or JSON file works.
Part 1. Sign in
Status: Shipped
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | Open the sign-in page and choose sign in with GitHub. | A GitHub page asks you to authorise DQaaS. | |
| 2 | Approve it. Read, then accept the terms of use and privacy notice. | You land in the console with your own empty organisation. | |
| 3 | Open Settings. Look at the organisation name and your role. | You are the owner of an organisation named after you. |
Part 2. Upload and profile a file
Status: Shipped but untested in the wild
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | Open Upload and drop in your file. Leave "Keep file" ticked. | A progress bar; the job finishes without an error. | |
| 2 | Open the new dataset from the Catalogue. | A dataset page with an overview, a score and column details. | |
| 3 | Check three things against your own file: row count, column names, and one column's missing values. | The numbers match what you know to be true. | |
| 4 | Look at the Columns and Overview tabs. | Types, distinct counts and basic statistics for each column. |
Part 3. Rules and re-assessment
Status: Shipped but untested in the wild
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | Open the Rules tab and add one rule, for example "not null" on a column that has gaps. | The rule appears in the list. | |
| 2 | Press Scan again. | A new score appears that reflects your rule. | |
| 3 | Download the report as PDF or HTML. | A readable report; it shows counts and scores, not a copy of your data. |
Part 4. Account and invites
Status: Shipped; team invites are Shipped but untested in the wild
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | In Settings find "Invite someone to the beta". Do not send one unless we ask you to. | You can see how many invites you have. | |
| 2 | Create an access token with the smallest scope, then revoke it. | The token is shown once and disappears from the list after revoking. | |
| 3 | Open the Data handling page from the footer. | A plain page describing what we process, keep, for how long, and where. |
Part 5. The command line (optional)
Status: Shipped but untested in the wild
The command-line program profiles a file on your own machine without uploading it, and can also talk to your account. Install it from the Docs page first. Replace the names in capitals with your own.
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | Run dqaas --version. | A version number is printed. | |
| 2 | Run dqaas run yourfile.csv -o report.html. For an Excel file with several sheets add --sheet "Sheet name". | A report file is written. Nothing was uploaded. | |
| 3 | Create a token with write scope under Tokens, set DQAAS_URL and DQAAS_API_KEY, then run dqaas whoami. | It names your organisation and role. | |
| 4 | Run dqaas run yourfile.csv --publish. | A new dataset appears in the console, labelled as computed by the command line and not verified by the server. Only counts and column structure were sent, never rows. | |
| 5 | Run dqaas gate DATASET_ID --min-score 50 for the dataset from Part 2, then dqaas gate DATASET_ID --require-verified for the published one. Check the exit code with echo $? (or $LASTEXITCODE in PowerShell). | The first prints a verdict and exits 0 if the score is at least 50. The published dataset fails with exit code 5 because it was computed outside the server. |
Part 6. An AI assistant through MCP (optional)
Status: Shipped but untested in the wild; hosted sign-in has been completed with Claude Code and Claude in the browser
MCP lets an AI assistant use DQaaS with the permissions of your token. You need an assistant that supports MCP servers.
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | Local, no account: add the dqaas-mcp program to your assistant as a local server, then ask it to profile a file on your machine. | You get a profile of the file. Nothing was uploaded. | |
| 2 | Local, with your account: run dqaas login, restart the assistant and ask it to list your datasets. | The datasets from Part 2 are listed with their scores. | |
| 3 | Hosted: add https://api.dqaas.io/mcp as a remote server. Either give it a token as a bearer credential or, if your assistant offers it, sign in when it asks and choose read or write access. | Asking for the latest score of your dataset returns the same number as the console. | |
| 4 | With a read-only token or read access, ask the assistant to create a glossary term. | It is refused. A read token cannot change anything. |
Part 7. Finish
Status: Shipped
| Step | What to do | What you should see | OK? |
|---|---|---|---|
| 1 | Delete the dataset you uploaded. | It disappears from the catalogue. | |
| 2 | Sign out, then sign back in with GitHub. | You return to your organisation without re-accepting the terms. |
What to send us
Email [email protected] with: (1) the part and step numbers where anything failed or felt wrong, (2) anything that confused you or took more than a few seconds, (3) a screenshot of any error, with data blurred, and (4) your overall impression in a sentence or two. Do not send us the file you uploaded.
What has and has not been tested
We would rather tell you than have you find out. Shipped means it works and has been checked on the live service. Shipped but untested in the wild means it works on prepared data but no outside person has used it yet. Not built means it does not exist.
| Part | Status | Note |
|---|---|---|
| GitHub sign-in, terms and your own organisation | Shipped | Two invited people have signed in and created their organisations. |
| Upload, profile, score and report | Shipped but untested in the wild | Checked on prepared data. No outside tester has finished these steps yet, so this walk-through is the first real run. |
| Rules and re-assessment | Shipped but untested in the wild | As above. |
| Tokens, settings, data handling page | Shipped | Tokens are shown once and can be revoked. |
| Team invites | Shipped but untested in the wild | No second real account has accepted one yet. |
| Command-line program (Linux, macOS, Windows) | Shipped but untested in the wild | The Windows build is tested automatically on Windows, not yet used on a person's own PC. Publishing a local result is tested but not yet used with real accounts. |
| MCP, local and hosted | Shipped but untested in the wild | Hosted sign-in has been completed with Claude Code and Claude in the browser. Other assistants have not been tried. |
| Paid plans and single sign-on beyond GitHub | Not built | Billing is built but switched off, so every account is free during the beta. |
DQaaS is in beta. Features may change, and the terms are drafts for this test period. See Data handling for what we keep and Docs for installing the command line and connecting MCP.