Duplicate Advisory: PraisonAI: PGVector and Cassandra knowledge stores interpolate vector dimensions into DDL

PraisonAI before version 4.6.78 fails to validate the caller-controlled dimension argument in the PGVector and Cassandra knowledge-store create_collection() backends. Although other identifiers such as schema, keyspace, and collection names are validated, the dimension value is declared as an integer but not enforced at runtime. This dimension value is directly interpolated into the vector column of the generated CREATE TABLE DDL statement. An attacker able to influence the collection-creation dimension can inject SQL or CQL tokens by passing malicious strings (e.g., '3); DROP TABLE tenant_secrets; --'), leading to SQL injection.

Successful exploitation allows an unauthenticated attacker to execute arbitrary SQL or CQL commands on the database, potentially resulting in complete compromise of confidentiality, integrity, and availability of the affected database. This includes unauthorized data access, modification, or deletion.

No official patch or fix is currently documented in the advisory. Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is available, avoid passing untrusted input to the dimension parameter in create_collection() calls or implement strict input validation and sanitization on the dimension argument to ensure it is a valid integer.