What you need before the first run
You need an Axolotl account with enough DQU balance for the estimate, plus storage you control and permission to read the input and write a separate output.
- Use a work email to sign in at https://app.adequata.ai. Your subscription needs enough DQU balance to cover the estimate shown before the run starts.
- Prepare a bucket or container with separate prefixes for input and output, so Axolotl never has to write results back over the files you started with.
- Grant list, read, and write on those prefixes through an IAM role, a service principal, or an access key. The connect screen includes templates you can apply.
- For Snowflake, use a TYPE=SERVICE user, an external stage, and storage grants that cover both the Snowflake credentials and the credentials the ADEQUATA pipeline will use.
- For Databricks, connect the underlying object store where the files already live. Axolotl reads those files; it does not submit work through the Databricks SQL or Jobs API.
Your first batch run
The first run stays in the app. You connect storage, pass schema and size checks, confirm the DQU estimate, and accept or decline each stage before write-back.
- Create a project, connect the source, and run the connection test before you point Axolotl at any data.
- Enter the input path and the output path. Each path can be a folder consisting of multiple partitioned files, or a single .csv or .parquet file.
- The structure and type of your data file(s) in the written back dataset folder remain consistent with the input path.
- Complete schema validation and resolve any column conflicts the interface asks you to decide.
- Confirm that the size check passes. If the estimated memory is above the cap, the run will not start.
- Choose the stage range and the options you want, including deduplication, anomaly methods, and synthetic tuning.
- Review the DQU estimate. When the balance is sufficient, start the run.
- As each stage finishes, open the review and accept or decline the sample before Axolotl continues.
- You can leave the page after the run has started. Progress remains available from the project and the dashboard.
Common job failures
Most failed starts are a credential, a path permission, a schema conflict, or an estimate the balance or memory cap will not allow.
| Symptom | Likely cause | Action |
|---|---|---|
| Connection test fails | Wrong credentials or bucket name | Re-check template policy and reconnect |
| Access denied on copy/list | Missing list/read/write on prefix | Fix cloud policy |
| Schema conflict | Incompatible files in one folder | Resolve in UI or split folder |
| Unreadable Parquet | Corrupt metadata/footer | Replace file |
| Size blocked | Estimated RAM over threshold | No charge; use smaller path |
| Insufficient DQU | Estimate exceeds balance + credits | Top up or narrow stages |
| Failure email | Job error after start | Open run in app; contact support with pipeline id |
How to contact support
For product questions, write to support@adequata.ai and include the pipeline identifier from the run if you already have one.
The security whitepaper, data processing addendum, and master services agreement are available on the Enterprise tier. Contact support for their access.
The public privacy policy is at https://www.adequata.ai/privacy-policy.