Cardiovascular System: Citation and Traceability

Data Characteristics

Cardiovascular system policies and Standard Operating Procedures (SOPs) originate from national health commissions, drug administration bureaus (guidelines, treatment norms), and internal hospital documents (clinical pathways, drug manuals). These documents are primarily PDFs and Word files. They are highly structured, containing extensive medical terminology, dosage units (e.g., mg/kg, IU), time units (e.g., hours, days), and specific diagnostic criteria and treatment processes. Update frequency is relatively stable; national guidelines revise every 3-5 years, while hospital SOPs update annually or semi-annually based on clinical practice and new drug releases. Documents are generally long, often exceeding a hundred pages, with numerous internal chapters and appendices.

Constraints on Citation and Traceability

The rigor and specialized nature of cardiovascular documents demand precise and traceable citations to ensure accurate medical decisions. The specificity of medical terms and units requires fine-grained text segmentation and retrieval to prevent misinterpretation due to lost context. Document update cycles dictate knowledge base synchronization strategies; overly frequent or delayed updates compromise information validity. Long documents and complex structures necessitate enhanced segmentation strategies. Each retrieved segment must carry sufficient contextual information while avoiding excessive length. Accurate identification and citation of numerical information like dosages and times are crucial to prevent medical errors. These constraints collectively emphasize high requirements for citation segment completeness, accuracy, and traceability.

Configuration Guidelines

Configuration Item	Recommended Value	Rationale
`Chunk size` (Segment Length)	`800–1200 characters`	Ensures the integrity of medical concepts, preventing truncation of critical information.
`Chunk overlap` (Segment Overlap)	`100–150 characters`	Maintains contextual coherence, especially in lengthy guidelines, improving the semantic completeness of retrieved segments.
`Recall count` (Recall Count)	`10–15 items`	Increases coverage, improving the chance of obtaining key information from multiple relevant SOPs.
`Similarity threshold` (Similarity Threshold)	Calibrate based on actual testing	Requires testing with actual corpus to balance the specificity and generality of recall.
`Rerank result count` (Rerank Return Count)	`Top 5 items`	Focuses on the most relevant core information, reducing the burden on engineers to filter large results.
`maxContext`	`3000 Tokens`	Accommodates the contextual needs of complex pathological descriptions and treatment plans in the cardiovascular field.

Common Pitfalls

The knowledge base search returns no results, or too few citation segments. The quote field in the response is empty. This occurs when the Similarity threshold (Similarity Threshold) is set too high, strictly filtering out potentially relevant but slightly less similar content.
The cited source document does not fully match the actual answer content. The source field points to a document where the key sentence of the answer cannot be found. This occurs when Chunk size (Segment Length) is too short, causing a complete medical concept to be split across multiple segments, and only some segments are retrieved.
After exporting a workflow and importing it into a new environment, referenced plugins cannot be found. The plugin node displays a red error state. This occurs when the corresponding version of the plugin is not installed or enabled in the target environment, preventing pluginId from being resolved.

Verification Steps

Select typical cardiovascular disease diagnosis and treatment questions. Observe whether the quote field in the returned results contains key diagnostic criteria, treatment plans, and drug dosage information.
Check the document pointed to by the source field. Manually verify the position and context of the cited segment in the original text to confirm its accuracy and completeness.
After simulating a knowledge base update, ask the same questions again. Verify whether the citation source has switched to the latest version of the policy or SOP document. Check the updateTime field.
For questions involving specific numerical values (e.g., LVEF < 40%), verify that the numerical values and units in the returned citation segment are correct and not truncated.

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 2026-09-21.