# Advanced analytics and Python SSE

Source: /docs/raptor-mcp/advanced-analytics

Raptor Gateway guidance for SSE connection discovery, deployed function metadata, platform prerequisites and validation of analytics results.

## Check the analytics service separately

Qlik Server-Side Extensions, or SSE, run analytics functions outside the Qlik Engine through an analytics connection. Raptor Gateway can help discover the configured connection and prepare Qlik expressions or load scripts. The SSE service, its dependencies and the Qlik connection must also be installed and reachable.

The packaged Qlik integration includes support for the governed qlik-py-tools payload. The functions available to a particular installation depend on that deployed service and configuration. Forecasting, clustering, machine learning, language-model and Firecrawl workflows can have different dependencies, credentials, data formats and runtime costs.

## Discover the connection and function contract

Call sse_connection_hint before generating SSE code. Use the actual analytics connection name returned for the installation. A hint describes configuration; it does not establish that the service is running or that a named function exists.

On Qlik Sense on Windows, sse_list_functions can inspect metadata-only GetCapabilities for an explicitly approved profile ID. Without an approved profile, the tool reports discovery as unavailable. Ask the administrator for verified service metadata rather than treating an unavailable result as an empty function catalogue or executing functions just to discover them.

- Use the connection and function names returned for the configured environment.
- Check argument types, separators and output shape for the exact function and service version.
- Use supported EXTENSION patterns. ScriptEval is not an approved execution path.
- Keep service credentials and provider keys in administrator-managed configuration.
- For Qlik Cloud, confirm the customer's supported analytics connection and network route; installing a Windows bridge alone does not establish Cloud connectivity.

## Validate a small, authorized workload

Once discovery and access are confirmed, test a small representative dataset with an agreed output. Check field grain, dates, missing values and returned columns before scaling the workload. For forecasting and prediction, compare results with a baseline and record the assumptions and evaluation period.

External language-model and web-research functions may send inputs to the configured provider and incur usage charges. Approve the data and provider before enabling them. Long-running requests need an appropriate reload method and a completion check; saving an SSE expression does not prove that it executed successfully.

## Related Links

- [Install and configure](/docs/raptor-mcp/install-and-configure)
- [Qlik connections](/docs/raptor-mcp/qlik-connections)
- [Analytics and app development](/docs/raptor-mcp/analytics-and-app-development)
- [Troubleshooting](/docs/raptor-mcp/troubleshooting)