What the Data for This Category Looks Like
Data for this category primarily comes from ad platform APIs, user profile datasets, offline budget scheduling sheets, and attribution tool logs. Daily associated documents include campaign setup documents, targeting setting configuration sheets, budget scheduling sheets, and conversion tracking configuration files. Data update frequency is either daily or per single campaign cycle. Data fields include targeting audience pack identifiers, budget amounts, conversion event types, delivery time parameters, and more. Units include yuan, person-times, hours, and others.
Constraints Imposed on Citation Sources and Traceability by These Characteristics
Since data sources include multiple heterogeneous types such as ad platform APIs and offline budget sheets, citation traceability requires adding unique identifiers to content from different sources to avoid confusion. Data update frequency is daily or per single campaign cycle. Traceability information must synchronously mark the update time of the corresponding data to ensure the timeliness of cited content. Most daily associated documents are structured spreadsheet files. Citations must accurately locate specific configuration item rows. Most fields are unique identifier parameters. Traceability must associate with the unique ID of the original data to facilitate subsequent verification. Unlike delivery effect optimization scenarios, campaign setup is account structure configuration before launch. Traceability must cover all configuration items.
How to Configure Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
quoteMaxToken | 900–1100 token | Matches the average content length of campaign setup structured documents, retains complete configuration parameters |
| Cited Content Display Length | 800–1200 characters | Prevents overly long content from disrupting reading, while covering core configuration information |
responseData | Enable source metadata return | Ensures results carry necessary traceability information such as data source type and update time |
| Maximum Number of Citations per Request | Top 5 entries | Aligns with the number of core configuration items that ad optimizers focus on daily, avoids redundancy |
detail | Enable full traceability mode | Displays original document line numbers and unique identifiers for cited content, supports accurate positioning of structured configuration sheets |
| Traceability Timestamp Toggle | Display synchronously with content | Adapts to timeliness verification requirements for daily or campaign-level data updates |
This configuration is based on FastGPT v4.17.0.
Three Common Misconfigurations
- Symptom: After configuring the maximum number of citations per request to exceed 2000 tokens, the large language model cannot receive context content. Cause: The system has a threshold limit for the total citation tokens. Exceeding this limit truncates all cited content, resulting in an empty context.
- Symptom: Cited source specific content is not displayed in returned results. Cause: The metadata return configuration corresponding to
responseDatais not enabled, only the abstract content generated by the large language model is returned. - Symptom: Format corruption occurs in cited content. Cause: Accurate line number positioning is not enabled for structured spreadsheet documents, leading to abnormal splicing of cross-format content.
How to Verify Proper Configuration
- Generate a test query for campaign setup configuration, check if the returned results include preset data source identifiers.
- Check the token length of returned content, confirm it does not exceed the configured
quoteMaxTokenvalue range. - Cross-check line numbers or unique identifiers of cited content, confirm they match the positioning in the original configuration documents.
- Adjust the maximum number of citations per request, verify that the number of cited entries in returned results matches the expected setting.
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.