[✨ Triage] dotnet/aspire#11712 by tjwald - Aspire 9.5 - gen ai visualization not parsing the messages correctly #1550
Description
Triage for dotnet/aspire#11712.
Repo filter: area-dashboard issues.
MihuBot version: 554101.
Ping MihaZupan for any issues.
This is a test triage report generated by AI, aimed at helping the triage team quickly identify past issues/PRs that may be related.
Take any conclusions with a large grain of salt.
Tool logs
dotnet/aspire#11712: Aspire 9.5 - gen ai visualization not parsing the messages correctly by tjwald
[Tool] Searching for gen ai visualization not parsing, gen ai spans dashboard, gen ai auto instrument python, open telemetry gen ai visualization, Aspire dashboard visualization gen ai, litellm open telemetry (IncludeOpen=True, IncludeClosed=True, IncludeIssues=True, IncludePullRequests=False, Repository=dotnet/aspire)
[Tool] Found 75 issues, 186 comments, 40 returned results (13058 ms)
Here are the most relevant related issues and discussions for triaging issue #11712 ("Aspire 9.5 - gen ai visualization not parsing the messages correctly"):
1. Issue #11697 (September 2025) - "Generative AI visualizer feature in .NET Aspire 9.5 was not working"
- Summary: The user reports that after upgrading to Aspire 9.5, the generative AI visualizer does not work as expected when using Semantic Kernel and OTEL GenAI semantic conventions. Clicking the sparkle icon in the dashboard throws a JSON parsing exception (
System.Text.Json.JsonException: 'C' is an invalid start of a value), indicating that the dashboard is not parsing the GenAI span messages correctly. - Relevance: This is a direct match to the reported problem in #11712, but with Semantic Kernel instead of litellm. The error is a JSON parsing failure in the dashboard's GenAI visualizer.
- Status: No comments yet, but the exception stack trace matches a parsing issue in the dashboard code.
2. Issue #5902 (September 2024) - "Improved rendering for Gen AI Telemetry signals in Aspire Dashboard"
- Summary: This feature request discusses the need for better visualization of GenAI telemetry signals in the dashboard, including conversational tree views, better handling of long text, and recognition of LLM spans. There is discussion about the dashboard being a generic telemetry viewer and the need for more specialized rendering for GenAI spans.
- Comments:
- There are requests for cards showing token usage, latency, and eval metrics.
- Discussion about the dashboard's ability to recognize and render specific span types (LLM, HTTP, DB).
- Confirmation from davidfowlthat improved GenAI visualization shipped in 9.5.
- Relevance: Shows ongoing work and requests for GenAI visualization improvements, and confirms that 9.5 includes new features in this area.
3. Issue #5451 (August 2024) - "OpenAI component doesn't setup telemetry correctly"
- Summary: Discusses the need for correct OpenTelemetry configuration for OpenAI components, including app context switches and activity source names. Mentions that telemetry is flowing when configured correctly and that Aspire should expose available telemetry, even if experimental.
- Relevance: While focused on OpenAI, it covers the broader topic of GenAI telemetry and the importance of correct OTEL configuration for visualization in Aspire.
4. Issue #4412 (June 2024) - "Dashboard: Traces & Log messages - viewer for long fields"
- Summary: Addresses the problem of long GenAI prompt/response fields being hard to read in the dashboard. Proposes a modal viewer for long text and automatic JSON formatting. This was implemented in #5018 and is referenced in the GenAI visualization discussions.
- Relevance: Shows that Aspire dashboard has improved handling of long GenAI messages, but does not address parsing errors.
5. Issue #10376 (July 2025) - "AddPythonApp fails too early when there's no virtual environment"
- Summary: Discusses Python integration and orchestration scenarios, including issues with virtual environments and telemetry. Not directly related to GenAI visualization, but relevant for users running Python GenAI apps with Aspire.
6. Issue #8152 (March 2025) - "AddPythonApp does not stop the Python application if you use OpenTelemetry."
- Summary: Reports issues with process management when using Python and OpenTelemetry. Fixes for process cleanup shipped in Aspire 9.4.
- Relevance: Related to Python/OTEL integration, but not directly to visualization parsing.
7. Issue #6298 (October 2024) - "Mark AddPythonApp API as experimental"
- Summary: The Python integration API is marked as experimental due to ongoing issues.
- Relevance: Indicates that Python integration (including telemetry and visualization) is still evolving.
8. Issue #7931 (March 2025) - "Tracing for OpenAI integrations"
- Summary: Discusses the need to clarify which experimental tracing is enabled for OpenAI/GenAI integrations and whether both should be enabled by default.
- Relevance: Highlights complexity in GenAI telemetry conventions and dashboard support.
Conclusions
- Directly Related: Issue #11697 is the closest match, reporting a JSON parsing error in the GenAI visualizer when using OTEL GenAI spans from a Python app. The error stack trace matches the dashboard code responsible for parsing GenAI messages.
- Feature/UX Discussions: Issue #5902 and #4412 discuss improvements to GenAI visualization, including better rendering and handling of long messages. Aspire 9.5 shipped new features in this area.
- Python/OTEL Integration: Issues #5451, #10376, #8152, and #6298 show that Python and GenAI telemetry integration is still evolving and may have edge cases or experimental APIs.
- Telemetry Conventions: Issue #7931 highlights the complexity of GenAI tracing conventions and the need for clear dashboard support.
Recommendations
- The issue in #11712 is likely a bug in the dashboard's GenAI visualizer parsing logic, similar to #11697.
- Check if the GenAI spans emitted by litellm (or other Python auto-instrumentation) match the expected format for Aspire 9.5's dashboard. If not, the dashboard may fail to parse them.
- Consider referencing #11697 in #11712, as both report dashboard parsing failures for GenAI spans from Python apps.
- If possible, provide a sample of the emitted span data to help diagnose the parsing issue.
If you need more details on any specific issue or want to see the full comment history, let me know!