Analysis View
Inspect the exact query AskEdi ran against your source, the AI-generated version, and how long each stage took.
The Analysis view shows the exact query that ran against your data source for each response, the version the AI originally generated, and how long each stage took. Every answer is traceable to a verifiable query against your own data — this is what separates AskEdi from a black-box AI tool.
Opening the Analysis View
Each response in a chat has an action bar directly below it. The action bar contains icons for rating, copying, exporting, and inspecting the response. Click the code icon in that row to open the Analysis view for that specific response.
The panel opens inline below the response. It shows the query or queries that ran, the latency figures, and — in Private mode — both the anonymized query the AI generated and the rewritten query Edilitics executed.
The Analysis view is available to both the chat owner and shared viewers.
Query Transparency
When AskEdi produces a response, two things happen in sequence: the AI generates a query based on your question and the context your privacy mode approved, then Edilitics executes that query against your source. The Analysis view surfaces both steps.
You can use it to:
- Verify that the query actually reflects what you asked
- Confirm that only a read operation ran against your source
- Copy the query and run it independently in a database client or BI tool
- Understand why a response took as long as it did
- Debug unexpected results by comparing what the AI intended versus what executed
Query Languages
The query language shown in the Analysis view depends entirely on your data source. AskEdi generates and executes in the native language of the source.
| Source type | Query language shown |
|---|---|
| MySQL, PostgreSQL, BigQuery, Snowflake, Redshift, SQL Server, SQLite | SQL |
| MongoDB, MongoDB Atlas | MongoDB aggregation pipeline |
| CSV, Excel, Parquet, Avro, Feather, Pickle, JSON, Google Sheets | Polars |
MongoDB queries are expressed as aggregation pipelines: a JSON array of stage operators ($match, $group, $sort, etc.) that the AI constructs and Edilitics executes against your collection. For flat file sources, AskEdi reads the file into a Polars DataFrame and the AI generates a Polars expression to query it.
What the Analysis View Shows
A single query panel is shown, labelled with your AI provider name, for example "OpenAI Generated Query". This is both what the AI produced and what Edilitics executed. No rewriting occurs because the AI received real column names.
Two query panels are shown side by side.
[Provider] Generated Query: what the AI produced using anonymized column identifiers (col_1, col_2, etc.). The AI never received real column names, so the query it wrote uses the anonymized aliases.
AskEdi Processed Query: the rewritten query with real column names substituted in. This is the query Edilitics actually executed against your source. Comparing the two shows the exact column mapping AskEdi applied.
The copy button is on the AskEdi Processed Query panel. This gives you the executable query with real column names, not the anonymized version the AI generated.
Some responses involve more than one query, for example when multiple charts are generated or when a Decision Intelligence response performs separate aggregations for each analytical section. Use the pagination controls at the bottom of the Analysis view to step through each query.
Latency Telemetry
Each response shows two latency figures in the action bar. The two figures together show you where time was spent.
Prop
Type
Frequently Asked Questions
Related Docs
Privacy Modes
How Private mode anonymizes column names and why the two queries differ.
Charts
How AskEdi generates chart responses and how to inspect the query behind each chart.
Decision Intelligence
How multi-section Decision Intelligence responses generate separate queries per analytical section.
Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.
Last updated on