A FastMCP tool to search and retrieve Polars API documentation.
Three tools present with schemas and descriptions. Tool naming is good (verb-starting), but descriptions lack LLM-optimization guidance, parameter descriptions are generic, and output schemas are not formally documented. The search_polars_api tool is the most complete, but all three lack specifics about what the LLM should do with results. No error handling guidance. All tools are READ_ONLY which is positive for safety, but compositions and chaining data are absent.
Get the currently installed Polars version information
List all available high‑level Polars API components
Search Polars API methods and get their signatures and documentation. Finds exact matches first, then falls back to fuzzy search.
Output schemas not formally documented in tool definitions. For search_polars_api, the docstring mentions 'JSON array of method information with name, signature, and description' but does not specify exact field names, types, or nested structure. LLMs cannot reliably plan downstream calls without formal output schemas.
Parameter descriptions in search_polars_api are example-heavy ('Examples: "dataframe", "DataFrame", "series,lazyframe"') rather than constraint-focused. Should instead use formal constraints (enums, patterns, length limits).
No error handling guidance. The tools do not document what happens on failure (e.g. if Polars is not installed, if a component name is invalid, if search yields zero results).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 55 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Tool descriptions are generic and do not answer 'When should the LLM call this instead of a similar tool?' For example, list_polars_components says 'List all available high‑level Polars API components' but does not explain when an LLM should call this vs search_polars_api with no arguments. Discovery intent is unclear.
search_polars_api returns results as a JSON string, but the structure is not defined. The docstring says 'JSON array of method information' but does not specify fields like name, signature, docstring, module, etc. LLMs cannot extract data reliably without a formal output schema.
No pagination or result limiting documentation. search_polars_api has a max_results parameter (default 1000), but the tool description does not state that results are capped or explain how pagination works if a query matches more than the limit.
Tool composition is weak. list_polars_components returns a list of names, but there is no guidance on how to use that output to inform subsequent search_polars_api calls. The relationship between the two tools is implicit, not documented.