Stephen Mallette created TINKERPOP-3310:
-------------------------------------------
Summary: gremlin-mcp tool cannot follow anchor-only references
Key: TINKERPOP-3310
URL: https://issues.apache.org/jira/browse/TINKERPOP-3310
Project: TinkerPop
Issue Type: Bug
Components: mcp
Affects Versions: 4.0.0
Reporter: Stephen Mallette
The Gremlin Semantics pages served by gremlin-mcp link into the reference docs
with HTML anchor URLs such as
https://tinkerpop.apache.org/docs/3.8.2/reference/#addedge-step. The installed
3.8.2 docs contain 232 of them. The reference book is split into multiple
Markdown pages, so the URL has no page path. get_gremlin_doc_page resolves it
to reference/.md, which doesn't exist, so an agent can't follow these links
even though the tool says it accepts site URLs.
These URLs should resolve to the page that owns the anchor. For example,
parsePage (or the page lookup in get_gremlin_doc_page) could rewrite
<site>/<version>/reference/#<anchor> to <page-id>#<anchor> using the section
anchors already parsed for each served page. An anchor no served page owns
should keep its site URL.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
