cwalch commented on issue #42806:
URL: https://github.com/apache/superset/issues/42806#issuecomment-5369569447
Hi,
we are interested in contributing an initial implementation of SIP-224.
Our primary use case is an AI assistant embedded directly into a Superset
dashboard, allowing users to ask natural-language questions about the data they
are currently viewing.
For example:
“Why did revenue in January decrease so significantly compared to the
previous month?”
We already use Superset's MCP service and would like to build the assistant
on top of the existing MCP infrastructure rather than introducing a separate
data-access or permission layer.
The proposed flow would be:
Dashboard
↓
AI Assistant / Chat Panel
↓
Dashboard context
(active filters, charts, datasets)
↓
LLM / Agent
↓
Superset MCP tools
↓
Superset RBAC / RLS
↓
Data
A key requirement for us is that all AI-driven data access continues to
respect the permissions of the currently authenticated Superset user. Since the
MCP service already integrates with Superset's security model, reusing it seems
like a good fit for this.
For an initial PR, we would suggest keeping the scope deliberately small:
Feature-flagged AI assistant
AI panel available from a dashboard
Pass the current dashboard context and active filters to the assistant
Use existing Superset MCP tools for data access
Execute requests in the context of the current user
Read-only analytical questions initially
Return/stream the generated answer in the panel
A later iteration could support more advanced analytical workflows, for
example automatically drilling into dimensions to identify the main drivers
behind a change in a metric.
Before starting the implementation, we would appreciate feedback from the
maintainers on two points:
Would an initial vertical-slice PR along these lines be useful for SIP-224?
Should the AI backend invoke mcp_service internally, or is there another
abstraction/API planned for the implementation of SIP-224?
If this direction aligns with the SIP, we would be happy to prepare an
initial PR.
Greetings Clemens
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
