docs/METADATA_SCHEMA.md

BioFlow Metadata Schema (Phase 3)

All ingested items are stored in Qdrant with a payload that includes core provenance fields plus source‑specific metadata.

Core Fields (all modalities)

Field	Type	Description
source	string	Source name (pubmed, uniprot, chembl)
source_id	string	Source identifier (e.g., pubmed:12345)
indexed_at	string	ISO timestamp when ingested
content	string	Stored raw content (text, SMILES, or sequence)
modality	string	text, molecule, or protein

PubMed (text)

Field	Type	Description
pmid	string	PubMed ID
title	string	Article title
authors	list[string]	Authors
journal	string	Journal name
pub_date	string	Publication date
year	number	Publication year
mesh_terms	list[string]	MeSH terms
url	string	PubMed URL

UniProt (protein)

Field	Type	Description
accession	string	UniProt accession
entry_name	string	UniProt entry name
protein_name	string	Protein name
gene_names	list[string]	Gene names
organism	string	Scientific name
organism_id	string	Taxon ID
function	string	Function text (truncated)
sequence_length	number	Sequence length
pdb_ids	list[string]	PDB references
url	string	UniProt URL

ChEMBL (molecule)

Field	Type	Description
chembl_id	string	ChEMBL molecule ID
name	string	Preferred name
synonyms	list[string]	Synonyms (limited)
smiles	string	Canonical SMILES
inchi_key	string	InChIKey
molecular_weight	number	Full molecular weight
alogp	number	ALogP
hba	number	H‑bond acceptors
hbd	number	H‑bond donors
psa	number	Polar surface area
ro5_violations	number	Rule‑of‑5 violations
target_chembl_id	string	Target ID (if available)
activity_type	string	Activity type (e.g., IC50)
activity_value	number	Activity value
activity_units	string	Activity units
url	string	ChEMBL URL