We would like to have a full text search prototype in the archive contents.
An efficient way of doing this is to exploit the Merkle graph as follows:

• index only the file contents (each content may appear in many different places)
• use the swh-graph and/or provenance index to show the results in context

Prototype working on a 1%+ subset of the archive.

Related work:

• SourceGraph (LWN coverage)
• Debian Code Search
• Language Server Protocol and their index format