DQL generated by nl2dql is incorrect

[Dynatrace-Asad-Ali]

Describe the bug
When using the natural language to get more details about a specific problem, the generated DQL is incorrect. In the DQL, it is filtering the query on field event.problem_id. However, it should be filtering on display_id. When I run the DQL generated by nl2dql in the notebook, I don't get any data in my tenant. Upon further investigation, it appears that event.problem_id is not a field anymore.

To Reproduce
Steps to reproduce the behavior:

1. Use this nl to query all problems "List all open problems in my dynatrace env for the past one hour".
2. From the list of problems shown in the output, ask this nl question "For problem , provide more details"
3. The output I get is "There are no additional detailed records found for problem P-2512155 in the last 7 days based on the filtered problem ID in Dynatrace."

I am using Dynatrace prod tenant.

The query that is generated by nl2dql is:
fetch dt.davis.problems, from: now()-7d, to: now() | filter event.problemId == \"P-2512155\" | fields event.description, event.status, event.category, event.start, event.end, root_cause_entity_id, root_cause_entity_name, duration, affected_entities_count, event_count, affected_users_count, problem_id, dt.davis.mute.status, dt.davis.mute.user, entity_tags, labels.alerting_profile, maintenance.is_under_maintenance, aws.account.id, azure.resource.group, azure.subscription, cloud.provider, cloud.region, dt.cost.costcenter, dt.cost.product, dt.host_group.id, dt.security_context, gcp.project.id, host.name, k8s.cluster.name, k8s.cluster.uid, k8s.container.name, k8s.namespace.name, k8s.node.name, k8s.pod.name, k8s.service.name, k8s.workload.kind, k8s.workload.name

Expected behavior
I expect the dql to be:
fetch dt.davis.problems, from: now()-7d, to: now() | **filter display_id == \"P-2512155\"** | fields event.description, event.status, event.category, event.start, event.end, root_cause_entity_id, root_cause_entity_name, duration, affected_entities_count, event_count, affected_users_count, problem_id, dt.davis.mute.status, dt.davis.mute.user, entity_tags, labels.alerting_profile, maintenance.is_under_maintenance, aws.account.id, azure.resource.group, azure.subscription, cloud.provider, cloud.region, dt.cost.costcenter, dt.cost.product, dt.host_group.id, dt.security_context, gcp.project.id, host.name, k8s.cluster.name, k8s.cluster.uid, k8s.container.name, k8s.namespace.name, k8s.node.name, k8s.pod.name, k8s.service.name, k8s.workload.kind, k8s.workload.name

Screenshots

Additional context
Add any other context about the problem here.

[christian-kreuzberger-dtx]

Thanks for reporting this! I'm trying to reproduce this, first attempt in VSCode / GH Copilot with Grok code fast and dynatrace-mcp-serve 1.1.0:

Second try, without asking for a list of problems first, took a bit longer, but also succeeded:

I also tried to ask Davis Copilot directly via the #generaet_dql_from_natural_language tool, and it worked:

Could you please check whether you have any instructions / docs / agent files / ... in your repository that might confuse the LLM? We are not referring to event.problemId anywhere.

[Dynatrace-Asad-Ali]

This is my system prompt:
You are an expert AI agent specializing in interacting with Dynatrace MCP server and Github MCP Server. You have access to all the tools exposed by the Dynatrace MCP server and Github MCP Server. Use these tools only to answer user query. If there is no tool that fits the user query, do not use any of your own tool. Instead, return back without any tool execution. For Github, all the files in the github repo are located in src/com/dynatrace/ecommerce directory. Please pass this location to any tool call that needs to read, analyze or list files.

I am running the MCP client tool from self hosted n8n and here is the screenshot from today (Dec 22, 2025)