The open-source vector database engineered specifically for SecOps and forensics. Standard DBs build recommenders. Hanshiro catches zero-days.
Signatures fail when code changes. Hanshiro bridges the gap with Hybrid Boolean-Vector Search and Behavioral Embeddings.
# Initiating Hanshiro Client connection to local ledger
client = hanshiro.Client(host="localhost:8080", strict=True)
# 1. VECTORIZE // Generating Behavioral Embedding
vector = client.embed(suspicious_script, model=CodeBERT_Sec)
# 2. HYBRID SEARCH // Boolean Pre-filter + ANN
results = client.search(
vector=vector,
limit=5,
filter={
"severity": "critical",
"dest_port": [443, 8080]
},
strategy="hybrid_exact"
)
print(f"MATCH >>> {results[0].threat_family}")
# OUTPUT >>> MATCH: CobaltStrike_Beacon_Variant_X (98.4%)Solving the trilemma that generic vector databases ignore.
Security logs arrive in massive streams. Hanshiro prioritizes recent data in-memory for sub-second query availability. No batching delays.
Built-in outlier detection flags vectors that drift too far from tenant baseline during ingestion, preventing model poisoning.
Threats are networks. Query a vector and instantly retrieve the connected graph of users, IPs, and assets touched by the threat.
Field-level encryption. When a TTL expires, the encryption key is deleted, rendering the data mathematically unrecoverable instantly.