agrag.common.schema_fingerprint

Fingerprints of a schema and of the extraction setup that ran against it.

schema_fingerprint covers structure. extraction_fingerprint covers what the extractor reads besides structure: descriptions, the settings that change its output, and the model name. Graph stores both on every document and chunk it writes, and update() compares them.

Classes:

• LLMOutputSettings – Extractor options that change LLM output besides the model and prompt.
• RunRecord – What a document and its chunks record about the run that wrote them.

Functions:

• extraction_fingerprint – Return the hash of everything the extractor reads besides structure.
• run_record – Build the run record of a schema and a set of extraction settings.
• schema_fingerprint – Return the hash of the structure of a schema.