XML Tools

Offline XML utilities for Dify: convert XML to JSON, extract values with XPath, format or minify documents, and validate against an XSD schema.

All processing happens inside the plugin runtime. The plugin makes no network requests, calls no third-party APIs, and requires no credentials.

• Source repository: https://github.com/morethanjustpassion/dify-plugin-xml-tools
• Contact: hsyhushuyou@163.com

Why this plugin

Dify workflows frequently receive XML rather than JSON — SOAP responses, ERP and MES interface payloads, RSS bodies, and legacy system exports. Today the only way to handle these is a Code node with hand-written parsing, which requires editing the sandbox dependency list and gives every workflow author a fresh opportunity to write an XXE-vulnerable parser.

No existing Marketplace plugin covers XML parsing. json_process handles JSON only; md_exporter can write XML files but cannot read them.

Setup

No configuration required. Install the plugin and the four tools are immediately available in Chatflow, Workflow, and Agent applications.

• Required APIs: none
• Required credentials: none
• Network access: none
• Connection requirements: none

Tools

XML to JSON

Converts an XML document into a JSON object. Attributes are prefixed (@ by default) and text content is keyed as #text.

Parameter	Type	Default	Description
xml_content	string	—	The XML document to convert
strip_namespace	boolean	true	Remove namespace prefixes from keys
attr_prefix	string	@	Prefix used for XML attributes

Input:

<Order no="SI001"><Item code="M-001"><Qty>120</Qty></Item></Order>

Output:

{"Order": {"@no": "SI001", "Item": {"@code": "M-001", "Qty": "120"}}}

XPath Extract

Extracts nodes, attributes, or text using an XPath 1.0 expression. Use this instead of converting the whole document when only a few fields are needed.

Parameter	Type	Default	Description
xml_content	string	—	The XML document to query
xpath	string	—	XPath 1.0 expression
strip_namespace	boolean	true	Strip namespaces so XPath needs no prefixes

Returns {"count": n, "results": [...]}. Element matches include tag, text, attributes, and the raw XML subtree.

XML Format

Pretty-prints or minifies a document. Also serves as a well-formedness check.

Parameter	Type	Default	Description
xml_content	string	—	The XML document to format
mode	select	pretty	pretty or minify
recover	boolean	false	Tolerate minor syntax errors

XML Validate

Validates a document against an XSD schema.

Parameter	Type	Default	Description
xml_content	string	—	The XML document to validate
xsd_content	string	—	The XSD schema source

Returns {"valid": bool, "errors": [{"line", "column", "message"}]}.

Security boundary

XML parsing of untrusted input is a well-known attack surface. This plugin constrains it explicitly:

Risk	Mitigation
XXE (external entity injection)	resolve_entities=False (lxml), disable_entities=True (xmltodict)
SSRF via external entities	no_network=True (lxml) — the parser cannot make network requests
Billion Laughs / entity expansion	Entities disabled entirely, plus huge_tree=False
Memory exhaustion	Input capped at 10 MB (MAX_XML_BYTES in tools/_common.py)

XSD schemas are accepted as inline source only. There is no parameter that accepts a URL, so no user input can direct the plugin at a network destination.

Errors are returned as plain messages. No credentials exist to leak, and stack traces are not exposed to the caller.

Dependencies

Package	Purpose
xmltodict	XML to JSON conversion
lxml	XPath 1.0 and XSD validation

lxml is a compiled extension. It is required because Python's standard library has no XSD validation and only supports a limited XPath subset.

Limitations

• XPath 1.0 only (lxml's xpath() does not implement XPath 2.0/3.0)
• XSD schemas must be supplied inline; xs:import/xs:include from remote URLs will not resolve, by design
• Namespace declarations (@xmlns:*) are retained on the root when strip_namespace is enabled

License

MIT