Two tools with clear naming and reasonable descriptions, but significant gaps in schema documentation and parameter descriptions. extract_places has a well-structured docstring and clear purpose, but the input schema lacks descriptions for both parameters. list_place_types is minimal but functional. Output schemas are not formally documented in the code. No error handling guidance, no validation rules, and no recovery paths documented. Naming follows verb_noun convention (extract_*, list_*) which is good, but parameter documentation is sparse.

• Add formal parameter descriptions to extract_places: 'text' should specify max length, encoding (UTF-8), and note that it accepts any of 13 supported languages. 'language' should document that it is optional and used only for bookkeeping, with a note that the model auto-detects.
• Document the output schema explicitly in the tool description or as a structured comment. Specify that 'places' is an array of objects with fields: text (string), type (enum: CITY|COUNTRY|REGION|ENTITY), score (float 0-1), start (int, char offset), end (int, char offset).
• Add error handling guidance: document what happens if the model fails to load (e.g., 'Model download failed. Ensure internet connectivity and retry.'), if text is empty (already handled, but should be documented), or if the pipeline times out.
• Enhance list_place_types description to explain its purpose: 'Call this first to understand the entity types extract_places can detect. Returns a list of type names and descriptions.'
• Add input validation: reject text longer than a reasonable limit (e.g., 10,000 chars) with a clear error message. Document this limit in the parameter description.