Inspectable analysis notebooks

Turn a data question into work your team can inspect

Combine SQL, Python, written context, tables, and charts in one ordered notebook, with review gates for consequential agent-proposed actions.

Blocks: SQL + Python + Markdown
Outputs: Tables + charts
Agent: Kole
Review: Approval for proposed actions

Quantum Lab is available to authenticated SyneHQ team members. Approval applies to configured agent-proposed consequential actions, not every manual notebook action.

Illustrative interface

Notebook workspace

Weekly account review

No live execution

Notebook3 ordered blocks

Preview only

SQL query

Account activity by week

Connection: warehouse

SELECT
  date_trunc('week', created_at) AS week,
  COUNT(*) AS active_accounts
FROM public.accounts
WHERE created_at >= CURRENT_DATE - INTERVAL '12 weeks'
GROUP BY 1
ORDER BY 1;

Python

Compare the SQL DataFrame

Input: SQL DataFrame

weekly = sql_result.assign(
    change=lambda frame: frame.active_accounts.diff()
)
weekly[["week", "active_accounts", "change"]]

Markdown

Decision note

Written context

Review the weekly change from the SQL and Python DataFrames before deciding whether to prepare a follow-up for the account team.

Decision

Keep the source query and comparison visible with the note.

Quantum Lab is an ordered notebook for working through a data question with SQL, Python, written context, tables, and charts. Put the investigation beside the result so a teammate can inspect the queries, transformations, and decisions that shaped it.

Open Quantum Lab to create and work in a lab for your team. Labs are scoped to authenticated team members and retain their ordered blocks as the analysis develops.

Build the notebook in order

Add the blocks that make the question understandable:

SQL Query blocks run against their selected connection and retain that connection context.
Python blocks start the Python runtime when the notebook needs it.
Markdown/Text blocks record the question, assumptions, and decision context alongside executable work.
Table and Chart blocks present a selected DataFrame as an output.

This sequence lets an investigation move from source data to a visible result without separating the explanation from the work. Reorder blocks as the question changes, and keep the useful intermediate steps when another person needs to revisit the analysis.

Quantum Lab / ordered notebook

Keep SQL, Python, written reasoning, and outputs in one visible order.

Illustrative notebook

A lab can combine SQL Query, Python, Markdown/Text, Table, and Chart blocks. Python starts its runtime on demand when a Python block is present.

Sample configuration - no live execution

01
SQL Query
Revenue by operating region
Connection: warehouse / analytics
```
SELECT region, SUM(amount)
FROM orders
GROUP BY region;
```

Python

Inspect the query DataFrame

Python runtime starts when this block is needed.

regional = sql_result.copy()
regional["share"] = regional.amount / regional.amount.sum()

03
Markdown
Decision context
Record what the sample analysis is meant to answer before reading the output.
```
Which regions need a follow-up before the next planning review?
```
04
Table
Regional summary
Input: Python DataFrame regional
```
region | amount | share
```
05
Chart
Revenue share
Input: Python DataFrame regional
```
x: region   y: share
```

Ask Kole to investigate with the available context

Kole can inspect the team's connections, schemas, tables, saved queries, and change tickets before proposing work. Its read-only discovery tools can run automatically so it can identify available data and inspect structure without asking for approval at each lookup.

That discovery is a starting point, not a substitute for checking the data. Use the notebook to keep the question specific: identify the connection, the tables or saved query involved, the time window, and the result needed before broadening the analysis.

When Kole proposes a configured consequential action, Quantum Lab presents it for review. Depending on the action, that can include running arbitrary SQL, creating or updating notebook blocks, saving a query, creating a database, or submitting a change. You can approve the proposal with edited arguments or reject it with feedback.

Approval does not apply to every manual notebook action. It applies to configured agent-proposed consequential actions; the regular notebook remains the place to write, arrange, and inspect analysis directly.

Quantum Lab / proposed action review

Review a configured consequential action before accepting the proposal.

Kole proposal / sample

Add a SQL Query block

The agent can propose a configured action after it has inspected available connection and schema context. This preview is a sample, not a live agent request.

Proposed purpose

Compare current revenue by operating region for the planning review.

Approval card

Configured consequential action

Preview

connection: warehouse / analytics
block type: SQL Query
placement: After current Python block

Edit arguments or leave feedback

Illustrative review card - controls disabled

Turn SQL and Python into outputs

SQL and Python blocks can produce DataFrames for later blocks. A Table block selects a SQL/Python DataFrame and displays its rows and columns. A Chart block selects a SQL/Python DataFrame, then uses its available fields to configure the chart output.

This keeps the output tied to a named source inside the notebook. Run the source SQL or Python block first, inspect the DataFrame, then choose the fields that belong in the table or chart. If the source changes, rerun it before relying on the output.

A practical analysis loop

Start with a SQL block that uses the intended connection and a bounded question.
Use Python when a transformation or calculation belongs between the query and the result.
Add Markdown/Text to explain assumptions, definitions, and the decision being considered.
Select the resulting SQL/Python DataFrame in a Table or Chart block and inspect the output.

The result is a notebook that keeps query context, calculations, explanatory notes, and visual outputs together. Go to the workflow when you are ready to structure a new investigation.

Quantum Lab / DataFrame outputs

Let the table and chart make their input visible.

Tables and charts consume SQL or Python DataFrames, so the output stays connected to the block that prepared it.

Illustrative values - no live execution

Table block

Regional summary

Input: SQL DataFrame

Illustrative regional summary from SQL and Python DataFrames
Region	Revenue share	Prepared by
North	Sample 42%	SQL DataFrame
West	Sample 31%	Python DataFrame
South	Sample 18%	Python DataFrame

Sample table output

Chart block

Weekly active accounts

Input: Python DataFrame

Sample chart / y: active accounts+48 accounts / 8 weeks

Illustrative weekly active accounts from a Python DataFrame
Week	Active accounts
Apr 01	128
Apr 08	136
Apr 15	133
Apr 22	149
Apr 29	158
May 06	154
May 13	171
May 20	176

Start

128

Latest

176

Change

+37.5%

Source

weekly

Sample chart output - no live data connection

Review consequential proposals before they run

Use the approval card to read the proposed action and its arguments before accepting it. Edit the arguments when the proposal needs a narrower target or clearer intent, or reject it and explain what should change. For SQL or database work, confirm the selected connection and statement before approval.

Keep destructive or broad changes out of an exploratory step until the target and purpose are clear. A notebook can help make the work inspectable, but it does not remove the need to review the database operation and its consequences.

Keep the investigation with the team

Quantum Lab lists and stores labs within the authenticated team context. Create a lab for a distinct question, edit its blocks as the investigation changes, and arrange the block order to match the reasoning a reviewer needs to follow.

Use clear block titles and Markdown/Text notes when an output depends on a business definition, a chosen filter, or a source-specific caveat. That context makes a later review more useful than a table or chart without its input assumptions.

When an exploratory path is no longer useful, remove the lab instead of leaving a misleading draft in the team's list. Keep the notebooks that still explain an active question or a result the team needs to revisit.

Keep the connection context, DataFrame source, and review decision close to the result so each next step has a clear starting point.

Start an investigation in Quantum Lab, or return to the workflow to plan the blocks before asking Kole for help.