When a question fails, Datarelix shows a short message with a way to recover — Retry, an editable copy of your question with Try again, Test connection, Manage connections or View plans. Run history shows the same failure by its code, such as SQL_TIMEOUT. Find the message below to see what it means and what clears it.
For many failures, Datarelix has already revised the query and retried a few times before showing a message.
The question could not be answered
| Message | In run history | What to do |
|---|
| Unable to answer | NOT_ANSWERABLE | The data in scope doesn’t contain what the question needs. Rephrase, or add the missing tables to the connection’s selection. |
| Ambiguous question | AMBIGUOUS_QUERY | The question has several reasonable readings. Name the measure, period or table you mean, then choose Try again. |
| Table not found | MISSING_TABLE | The query needed a table that isn’t selected. Add it to the selection and re-run the question. |
| Column not found | MISSING_COLUMN | The column doesn’t exist in the selected tables. Check the selection, or rephrase using columns that exist. |
| Data type mismatch | TYPE_MISMATCH | Values of different types were compared, such as text and dates. Rephrase so the comparison uses matching types. |
The question guide shows how to state the period, measure and grouping so a question has one clear reading.
The query took too long or used too much
| Message | In run history | What to do |
|---|
| Query timed out | SQL_TIMEOUT | The database query passed the 30-second limit. Add a date filter, narrow the scope, or ask the database owner about the workload. |
| Processing timed out | SANDBOX_TIMEOUT | A Python analysis step passed its 90-second limit. Aggregate in SQL first or reduce the rows it works on. |
| Out of memory | SANDBOX_OOM | A Python analysis step passed its 512 MB memory limit. Select fewer columns or aggregate before the analysis. |
Datarelix could not reach your database
| Message | In run history | What to do |
|---|
| Connection failed | CONNECTION_ERROR | Choose Test connection. If it fails, check the credentials, that the database is running, and that its firewall allows Datarelix. The connection guide for your engine lists the common causes. |
The generated query failed
| Message | In run history | What to do |
|---|
| SQL validation failed | SQL_VALIDATION | The query was refused before it ran — for example because it was not a read, or it referred to objects outside the connection’s scope. Rephrase the question; widening database permissions does not fix this. See Security & read-only controls. |
| Query execution error | SQL_EXECUTION | The database returned an error, for example a permission problem. Check that the connection’s identity can read the tables involved. |
| SQL syntax error | SQL_SYNTAX | The generated query wasn’t valid for your engine. Rephrase or simplify the question. |
| Processing error | SANDBOX_CODE_ERROR | The generated Python analysis code failed. Try a simpler analysis or rephrase. |
| Invalid AI output | LLM_INVALID_OUTPUT | The AI returned something Datarelix could not use. Choose Retry. |
The AI service had a problem
| Message | What to do |
|---|
| The AI service is busy | Try again in a moment. |
| The AI service timed out | Try again. If it keeps happening, ask a narrower question. |
| AI service error | Try again. If it persists, contact support with the reference shown. |
Plan limits
These messages mean the request reached a limit of your plan, not that something broke. Your data and history are unaffected. Limits per plan are in Billing & plans.
| Message | What to do |
|---|
| Question limit reached / Monthly question limit reached | You’ve used this month’s questions. The counter resets on the 1st (UTC), or choose View plans to upgrade; a new allowance applies straight away. On Team, the allowance is shared by the workspace. |
| Daily fair-use limit reached | Pro and Team allow 500 questions per person per day. The limit resets at midnight UTC; your monthly allowance is untouched. |
| Monthly AI schema analysis allowance reached | You’ve used this month’s AI schema analyses. Rerunning discovery and resuming an interrupted analysis remain free. Wait for the 1st (UTC) or upgrade. |
| This connection is paused | Activate it from Manage connections — this needs room under your plan’s connection limit — or use another connection. |
| Too many active connections for your plan | You have more active connections than your plan includes, for example after a downgrade. Pause the ones you don’t need from Manage connections, or upgrade. |
| All seats are taken | A Team invite needs a free seat. Add seats in Manage billing, or remove a member or pending invite. |
| Subscription inactive | Inviting teammates needs an active Team subscription. If a subscription has ended you are on the Free plan with your data intact; upgrade to restore its limits. |
Schema analysis
| Message | What to do |
|---|
| Analysis failed | Choose Retry Schema Analysis where it is offered, or run the analysis again. If it fails repeatedly, contact support with the reference shown. |
| Analysis interrupted | Choose Retry Schema Analysis. Schemas that were already analyzed are reused, and the retry does not use another AI schema analysis. |
| Couldn’t read the tables in your database | The analysis could not read the database, so nothing was analyzed, your saved schema is unchanged and the AI schema analysis is not counted against your monthly allowance. Datarelix already waits about a minute for a database that pauses when idle to wake up. Check that the database is running and reachable from Datarelix, then run the analysis again. |
| No tables or views were found | The schemas selected for this connection are empty, so there was nothing to analyze and the AI schema analysis is not counted. Check the schemas selected in the connection’s settings. |
If AI schema analysis takes a long time on a very large schema, narrow the connection’s allowed scope or skip the analysis and work from discovery alone.
Errors on an answer or a schema analysis show a Reference. If you contact support, include that reference, the message and what you asked. Never send database passwords or other credentials.