rete · Guides Real-world scenario

Trace a dependency vulnerability from a URL

Publish a software dependency graph as a .rete file, then ask which packages are affected by a vulnerable dependency. The client runs the query; the host only serves the file.

This example uses the repository's examples/deps.nt. It contains package types, these dependency edges, and one fictional CVE:

app     -> web, auth
web     -> logging
auth    -> logging, safejson
logging -> log4x
log4x   -> hasVulnerability -> CVE-2099-0001

Commands below use rete; Getting started explains the published Docker wrapper for POSIX shells and PowerShell. Run them in a checkout containing examples/deps.nt, or download that file into an examples directory.

Build and check the snapshot

Validate the RDF syntax, build a file, then check its content hash:

rete validate examples/deps.nt
rete build examples/deps.nt -o deps.rete
rete verify deps.rete

Validation reports 13 statements, all in the default graph. The build reports 13 triples, 12 terms and 2 pyramid levels; verification should say the content hash matches. File size can change between releases. These checks establish syntax and file integrity, rather than completeness of an SBOM or absence of vulnerabilities.

Before publishing, check the impact query locally:

rete sparql deps.rete "PREFIX e: <http://ex/> SELECT ?d WHERE { ?d e:dependsOn+ e:log4x } ORDER BY ?d"

It should return four packages, in order: app, auth, logging, web. safejson has no dependency path to log4x, and log4x itself has no outgoing dependency path to itself.

Publish the file

Upload the snapshot to a static host with HTTP range support. For example, with a configured AWS CLI and your own bucket:

aws s3 cp deps.rete s3://my-bucket/deps.rete

The URL must be readable by your intended clients. Use Hosting for permissions, range checks, CORS for browser access, and cache validators. Publish changed content at a new versioned URL or configure validation so clients do not reuse stale cached ranges. The file's content hash identifies its content; HTTP caching policy still belongs to the host.

Ask which packages are affected

Replace the URL below with the location you published:

rete sparql-url https://my-bucket.s3.amazonaws.com/deps.rete "PREFIX e: <http://ex/> SELECT ?d WHERE { ?d e:dependsOn+ e:log4x } ORDER BY ?d"

Expected bindings are the same as the local query:

?d=<http://ex/app>
?d=<http://ex/auth>
?d=<http://ex/logging>
?d=<http://ex/web>

The command also reports four solutions and its fetched bytes and requests. The + follows one or more dependency edges. This toy graph may fit in one bounded read. On larger graphs, selective queries can avoid much of the file, but long paths and broad queries may read substantial index data. See SPARQL support for lazy-query support and limits.

To restrict the answer to libraries and discover the vulnerability identifier from the graph, join the path with its vulnerability and type statements:

rete sparql-url https://my-bucket.s3.amazonaws.com/deps.rete "PREFIX e: <http://ex/> SELECT ?lib ?cve WHERE { ?lib e:dependsOn+ ?v . ?v e:hasVulnerability ?cve . ?lib a e:Library } ORDER BY ?lib"

This returns auth, logging and web, each paired with http://ex/CVE-2099-0001. app is an Application, so the type constraint excludes it. This is dependency reachability in the supplied data, rather than an assessment of exploitability or version ranges.

A direct fact lookup needs no SPARQL:

rete query-url https://my-bucket.s3.amazonaws.com/deps.rete --predicate '<http://ex/hasVulnerability>'

It returns the statement linking log4x to the CVE. For an overview, use summary-url on the URL, or schema deps.rete locally. These summarize community structure and typed relations; they do not replace the exact dependency-path answer. See Semantic zoom and the CLI reference.

Inspect the transport and open the browser

A .rete URL is not a SPARQL endpoint: curl can retrieve bytes, while a rete client interprets the file and runs queries. Check a published file's header range with this POSIX command (use curl.exe in PowerShell):

curl -sD - -o header.bin -r 0-1023 https://my-bucket.s3.amazonaws.com/deps.rete

A range-capable host returns 206 Partial Content and a Content-Range header beginning bytes 0-1023/. The saved header begins with RETE. A host returning 200 has sent a whole-file response. Current readers can retain such a response up to 256 MiB; larger responses are refused rather than silently downloading an unbounded file. Configure ranges for reliable large-file access; details are in Hosting.

For this small example you can also download the file and run the local query:

curl -s https://my-bucket.s3.amazonaws.com/deps.rete -o deps.rete

In the playground, use + Add source > Connect (lazy) with the same URL and paste the impact query. With CORS enabled on the host, the browser runs it against the same snapshot and shows the bytes fetched. See the playground guide for interpreting results and sharing the query. A custom dashboard can use the Browser API.