What Data Looks Like for This Category
Data for this category primarily comes from engine Profiler, crash platforms, real-device testing, and frame rate collection tools. Daily produced documents include performance test reports, device adaptation lists, optimization records, and crash analysis reports. Data fields include frame rate values (fps), memory usage (MB), crash stack information, device models, and test times. Data update frequency is typically daily or per version release cycle. Test data for different devices is archived separately. The content length of a single test report usually ranges from thousands to tens of thousands of characters.
Constraints Imposed by These Characteristics on Quote Sources and Traceability
Because data sources include engine performance analysis tools, crash reporting platforms, and multi-device real-device test results, quote traceability requires clear marking of the data source type for each quote to avoid confusing results from different test dimensions. Data updates daily or per version, so quote refresh rules matching the data update cycle must be configured to ensure traceability content aligns with the current version’s optimization status. Daily documents have high structural complexity but many subdivided fields. Traceability must link specific test links to corresponding fields, and support multi-device test data tagging to ensure accurate traceability and avoid cross-device performance data confusion.
Configuration Settings
These configurations are formulated based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Matches the typical content length of a single performance test report, to avoid truncating critical frame rate and memory data |
responseData | Return only performance-related fields | Adapts to the precise traceability needs of this category, filtering out irrelevant non-performance test content |
detail | Enable full traceability | Retains detailed fields such as crash stacks and device models, meeting the detailed traceability requirements for test cases and defect analysis |
| Quote Data Source Filter | Check engine Profiler, crash platform, real-device testing | Only include core performance-related data sources for this category, reducing invalid quotes |
| Quote Refresh Cycle | Set based on actual measurement (usually daily or per version) | Matches the update frequency of this category’s data, ensuring the timeliness of traceability content |
Three Common Misconfigurations
- Phenomenon: An error occurs when viewing performance traceability in chat responses, prompting "No matching data source found". Cause: Core data sources such as engine Profiler and crash platform are not checked in the configuration, causing traceability to fail to match corresponding data.
- Phenomenon: Reference content truncates critical frame rate or memory usage data. Cause: The
quoteMaxTokensetting is too low, unable to accommodate complete performance test report fragments. - Phenomenon: Unnecessary non-performance-related reference content appears in responses. Cause:
responseDatais not configured to return only performance-related fields, mixing in information from other unrelated test links.
How to Verify Proper Configuration
- Initiate a test query for low-end device frame rate anomalies, and check whether the returned results mark data source information such as engine Profiler and real-device testing.
- Check the length of reference fragments, confirm that critical data truncation does not occur, and match the
quoteMaxTokensetting value. - Verify that core data sources for this category are checked in the configuration, ensuring no "data source not found" errors occur.
- Confirm that the quote refresh rule matches the data update cycle, ensuring that new version optimization records can be properly associated for traceability.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.