Size of schema serialization can be optimized #181
Description
Environment details
- OS type and version:
- Python version: 3.13.3
- pip version: 25.1.1
langchain-google-spannerversion: 0.82
Steps to reproduce
See PR: #182
Right now, when the schema is serialized looks like:
"Node properties per node label": { "FiscalPeriodKeyFigures": [ { "property name": "diluted_epsexcluding_extrordinary_items", "property type": "FLOAT64" }, .... "Edge properties per edge label": { "hasThreat": [ { "property name": "company_code", "property type": "STRING" }, by shorting the to only "name" and "property", which looks like:
"Node properties per node label": { "FiscalPeriodKeyFigures": [ { "name": "diluted_epsexcluding_extrordinary_items", "type": "FLOAT64" }, .... "Edge properties per edge label": { "hasThreat": [ { "name": "company_code", "type": "STRING" }, We observe 16% reduction of the schema size and no impact on the performance of our system tested with 100 user prompts. This reduces the amount of tokens used in the generation, verification and fix prompts.