| name | inventory |
| description | Extract an inventory of entities (person names, money, locations, organizations, time, dates) and extract main keywords from the provided text to describe the material. |
When the user asks you to "make an inventory", "inventory this", or sends content with instructions to extract entities and keywords, you must instantly perform the following analysis:
-
Entity Extraction: Read the provided text or transcript and systematically identify and list all significant entities, categorized clearly into the following headers:
- Time and Dates: (e.g., October 24th, 1999, 5:00 PM, Next Tuesday)
- Person Names: (e.g., John Doe, Elon Musk)
- Organizations: (e.g., Google, United Nations, Ministry of Magic)
- Locations: (e.g., Paris, Texas, Mount Everest)
- Money: (e.g., $100, 50 million euros, 10 bitcoins)
-
Keyword Extraction: Identify and extract the 5-10 most important "Main Keywords" or key phrases that best define the core themes and topics of the material.
-
Presentation: Output the results cleanly using standard Markdown formatting. Use bulleted lists. If a specific entity category has no factual matches in the generated text, explicitly write "None found" beneath that header rather than omitting the category entirely.
-
JSON-LD Caching: Immediately after presenting the markdown, silently convert the extracted entity inventory into a valid JSON-LD format document mapping the concepts properly (e.g., schema.org). You MUST use your file writing tools or run a bash command (cat << 'EOF' > ...) to save this raw JSON-LD data to a uniquely named file inside the cache/ directory (for example: cache/inventory_TIMESTAMP.jsonld).
Provide the requested markdown inventory and a brief confirmation that the JSON-LD file has been cached, without extra conversational filler.