This vulnerability is a path traversal flaw in the /v1/files uploading endpoint of Langchain‑Chatchat 0.3.0, allowing an attacker to craft filenames that result in files being written outside the intended openai_files directory. The capability to write arbitrary files can lead to significant confidentiality, integrity, or availability impacts, such as overwriting configuration files or deploying malicious content. Based on the nature of the flaw, it is inferred that placing executable files could potentially enable remote code execution if sufficient privileges are present.

The affected system is Langchain‑Chatchat version 0.3.0, an open‑source framework that implements an OpenAI‑compatible API for file uploads. All deployments exposing the default /v1/files endpoint are vulnerable and must be assessed.

Based on the vulnerability description, it is inferred that the flaw can be exploited remotely via the /v1/files endpoint. Attackers can manipulate the uploaded filename to place files anywhere on the filesystem of the host running Langchain‑Chatchat. This provides a high potential for data breach or system compromise. The CVSS score of 9.1 indicates a critical severity, and the EPSS score of < 1% suggests low but nonzero exploitation probability. The vulnerability is not listed in the CISA KEV catalog, but the path traversal weakness itself presents a significant risk. Immediate action is recommended until an official fix is released.