Adding custom dependencies

You can use any Clojure or Java library available on Maven or Clojars. All you need to do is to provide the dependency map with the following keys:

• :coordinates - a vector of Maven artifact/version pairs
• :requires - a vector of namespaces that should be required, just like in usual :require block in ns form
• :imports - a vector of fully qualified Java classes that should be imported

Here is an example of how to add a dependency on org.clojure/data.csv:

{:name  :my-pipeline
 :deps  {:coordinates [[org.clojure/data.csv "0.2.0"]]
         :requires    [[clojure.data.csv :as csv]]
         :imports     [java.time.LocalDate java.io.File]}
 :tasks []}

After that you can use csv namespace or LocalDate class inside your pipeline actions. These dependencies will be fetched at runtime just before the pipeline execution.

Collet's optional actions use normal Maven coordinates as well. Prefer the smallest module that contains the namespaces a pipeline needs:

{:deps {:coordinates [[io.velio/collet-action-http "VERSION"]]
        :requires [[collet.actions.http]]}}

Use [io.velio/collet-actions "VERSION"] when compatibility with all legacy action namespaces is more important than classpath size. VERSION always means the version published for that specific coordinate; Collet modules are independently versioned and their versions do not need to match. The aggregate's generated POM pins the exact compatible versions of all action modules transitively, so consumers must not add or synchronize those internal versions themselves. See the module graph for every coordinate and its direct internal dependencies.

These are consumer dependencies. Inside this repository, package deps.edn files use top-level :local/root entries for other Collet packages, so a checkout uses its own source. Kmono converts them only for generated POMs, where every internal Maven coordinate receives the exact version resolved from its package tag. Do not copy those generated versions into source dependencies.