RAGFlow versions prior to 0.27.0 contain a server-side request forgery (SSRF) vulnerability in the set_api_key endpoint. Authenticated attackers can supply an unvalidated base_url parameter to make the server issue requests to arbitrary internal or external endpoints. This can be used to probe internal services such as MySQL, Elasticsearch, Redis, MinIO, the Infinity vector database, and cloud metadata endpoints. The vulnerability also leaks raw upstream exception text in responses, revealing internal service details, open ports, and configuration fingerprints. The issue is fixed in version 0.27.0.
RAGFlow version 0.24.0 has an improper access control vulnerability in the get_dataset function within the evaluation_app API. This flaw allows an attacker to execute arbitrary code or commands if they can access the vulnerable entry point. The vulnerability is critical with a CVSS score of 9.8, indicating high impact on confidentiality, integrity, and availability.
RAGFlow through 0.27.2 contains a path traversal vulnerability in the dev_insert_chunks_from_file and dev_insert_metadata_from_file endpoints that allows authenticated attackers to read arbitrary files by supplying absolute file paths in the file_path parameter. Attackers with valid access tokens can exploit missing path validation to read any file accessible to the service, with disclosure limited to files matching expected JSON structures that are then written to datasets.
RAGFlow before 0.26.3 contains a server-side request forgery vulnerability in the agent workflow "Invoke" component (agent/component/invoke.py). The component builds an outbound request URL from canvas configuration and runtime template variables and passes it to requests.get, requests.post, or requests.put without calling the shared assert_url_is_safe validator or pinning the resolved address, unlike the crawler, SearXNG, file-upload, and RSS fetch paths. A user who can create or trigger an agent can direct the server to fetch loopback, link-local, and RFC 1918 destinations, including cloud instance metadata endpoints and services co-located on the deployment network, and the response body is returned as the component output. Where an agent is configured to interpolate the chat query into the Invoke URL, the destination is chosen by whoever can send that query.
RAGFlow before 0.26.3 stores an agent pipeline (DSL) node name without sanitization: the agent update endpoint normalizes the submitted DSL via normalize_dsl, which only performs JSON serialization validation and preserves the node name verbatim. The dataflow-result web UI then renders that name into the "Rerun from current step" confirmation modal via dangerouslySetInnerHTML, and the i18next configuration sets escapeValue:false, so the value is inserted into the DOM without HTML encoding. An authenticated workspace user who can create or edit an agent can inject arbitrary JavaScript that executes in the session of another workspace member who opens the dataflow result and clicks rerun, enabling session/token theft and account takeover across the user trust boundary.