Loading ToolBench
Search and retrieve medical imaging data from a DICOMweb-compliant server. Supports querying studies, series, and instances; reading DICOM metadata; extracting Structured Report text; and rendering instance frames as images.
This server demonstrates strong tool design with consistent naming conventions, comprehensive parameter descriptions, and clear input schemas. All 9 tools follow a verb-noun pattern (find-*, get-*, render-*) that signals intent. Parameter descriptions are detailed and include format constraints (DICOM UID regex patterns, maxLength), examples, and dependency hints. However, output schemas are not documented in the visible source code, and error handling guidance is minimal. The tools are well-composed with clear chaining (find-studies → find-series → find-instances), but there is no explicit error recovery guidance, no rate limiting hints, and no discussion of API failures.
Finds all Encapsulated PDF DICOM instances in a study by searching for DOC-modality series and filtering by EP SOP Class UIDs. Requires a Study Instance UID from find-studies.
Searches DICOM instances within a single series, sorted by Instance Number ascending. Requires Study and Series Instance UIDs from find-studies and find-series.
Searches DICOM series within a single study. Returns series sorted by series date, newest first. Does not retrieve instances or image data. Requires a Study Instance UID from find-studies.
Finds all Structured Report (SR) DICOM instances in a study by searching for SR-modality series and filtering by SR SOP Class UIDs. Requires a Study Instance UID from find-studies.
Searches DICOM studies on the configured DICOMweb server. Returns studies sorted by study date, newest first. Does not retrieve series, instances, or image data.
Output schemas not documented. Tools return results but no schema descriptions visible in source code for what fields/structure LLMs should expect. This forces LLMs to reason about response structure without guidance.
No error recovery guidance. Tool descriptions state what they do but do not explain what happens on failure (network timeout, invalid UID, server unreachable) or what the LLM should do next. Missing pattern: recovery-guide.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | <=2025-11-25 | v2 |
Fetches an Encapsulated PDF instance's metadata from the DICOMweb server and converts it to a human-readable text format.
Fetches an instance's metadata from the DICOMweb server and converts it to a human-readable text format.
Fetches a Structured Report instance's metadata from the DICOMweb server and converts it to a human-readable text format.
Fetches a rendered frame from a DICOM instance via the DICOMweb WADO-RS endpoint.
No pagination guidance in schemas. Search tools (find-studies, find-series, find-instances) accept limit and offset parameters but no documented total count or next_cursor returned. For large DICOM studies, missing pagination info risks context explosion.
No result size caps documented. The query parameter description for find-studies, find-series, and find-instances mentions limit=N but does not state a maximum limit or default. Large medical databases could return thousands of items, exhausting context.
No tool annotations (readOnlyHint/destructiveHint/idempotentHint). All tools are read-only but source shows no explicit tool annotation metadata. LLMs cannot infer safety classification without annotation.