LightRAG is a retrieval‑augmented generation platform that processes Markdown or textpack documents. An upload that includes an image URL can trigger a download of the referenced image. The parser module checks whether the resolved address is globally routable using Python's ipaddress.is_global. Before version 1.5.5 this check did not consistently classify IPv4 addresses embedded in IPv6 transition wrappers such as NAT64 (64:ff9b::/96), IPv4‑compatible, or 6to4 addresses. As a result, an attacker could upload a document that references a private, loopback, or cloud‑metadata address wrapped in one of these transition notations. The parser would accept the address, download the internal resource, and ingest its contents, thereby providing an internal SSRF path that can read sensitive internal data.

All installations of LightRAG older than 1.5.5 are vulnerable. The weakness exists in the parser located at lightrag/parser/markdown/parser.py and affects any deployment that accepts user‑supplied Markdown or textpack uploads. The issue is fixed in release v1.5.5, so the vendor’s recommendation is to upgrade when possible.

The CVSS score of 7.1 indicates a moderate‑to‑high severity, and the absence of an EPSS score suggests there is insufficient publicly available data on exploitation likelihood, but the vulnerability is not currently listed in the CISA KEV catalog. Exploitation requires the attacker to supply a vulnerable document via an upload endpoint. Once the guard is bypassed, the server will fetch data from internal destinations, which can expose configuration files, metadata, or private resources. The attack can be carried out remotely if the upload interface is publicly reachable, making this a compelling SSRF risk for organizations that expose LightRAG to untrusted users.