In cases where 'Lookup table-driven extraction' RAG technique is used, the lists provided in the prompts directly or the JSON structure should be generated dynamically in the future by querying available instances in TWA. Currently, this is hard-coded (e.g., "enum": ["N2", "Ar", "Air", "N/A"])
Examples of this are:
|
Try to assign a vessel type from one of the following: Teflon-lined stainless-steel vessel, glass vial, quartz tube, round bottom flask, glass scintillation vial, pyrex tube, or schlenk flask. If it is impossible to assign a vessel insert N/A. |
|
Try to assign the vessel to one of the following: Teflon-lined stainless-steel vessel, glass vial, quartz tube, round bottom flask, glass scintillation vial, pyrex tube, or schlenk flask. |
|
"properties": { # Defines the allowed properties and their types |
- All uses of
enum keyword in https://github.com/TheWorldAvatar/MOPTools/blob/00f489688de044d6a3d6ab407285796febbfa4a1/MOP_Literature_Extraction/json_schemas.py
