The helm-docs tool auto-generates documentation from helm charts into markdown files. The resulting files contain metadata about their respective chart and a table with each of the chart's values, their defaults, and an optional description parsed from comments.

The markdown generation is entirely gotemplate driven. The tool parses metadata from charts and generates a number of sub-templates that can be referenced in a template file (by default README.md.gotmpl). If no template file is provided, the tool has a default internal template that will generate a reasonably formatted README.

The most useful aspect of this tool is the auto-detection of field descriptions from comments:

config:
  databasesToCreate:
    # -- default database for storage of database metadata
    - postgres

    # -- database for the [hashbash](https://github.com/norwoodj/hashbash-backend-go) project
    - hashbash

  usersToCreate:
    # -- admin user
    - {name: root, admin: true}

    # -- user with access to the database with the same name
    - {name: hashbash, readwriteDatabases: [hashbash]}

statefulset:
  image:
    # -- Image to use for deploying, must support an entrypoint which creates users/databases from appropriate config files
    repository: jnorwood/postgresql
    tag: "11"

  # -- Additional volumes to be mounted into the database container
  extraVolumes:
    - name: data
      emptyDir: {}

Resulting in a resulting README section like so:

You'll notice that some complex fields (lists and objects) are documented while others aren't, and that some simple fields like statefulset.image.tag are documented even without a description comment. The rules for what is and isn't documented in the final table will be described in detail later in this document.

helm-docs can be installed using homebrew:

brew install norwoodj/tap/helm-docs

or scoop:

scoop install helm-docs

This will download and install the latest release of the tool.

To build from source in this repository:

cd cmd/helm-docs
go build

Or install from source:

go install github.com/norwoodj/helm-docs/cmd/helm-docs@latest

If you want to automatically generate README.md files with a pre-commit hook, make sure you install the pre-commit binary, and add a .pre-commit-config.yaml file to your project. Then run:

pre-commit install
pre-commit install-hooks

Future changes to your chart's requirements.yaml, values.yaml, Chart.yaml, or README.md.gotmpl files will cause an update to documentation when you commit.

There are several variants of pre-commit hooks to choose from depending on your use case.

---
repos:
  - repo: https://github.com/norwoodj/helm-docs
    rev:  ""
    hooks:
      - id: helm-docs
        args:
          # Make the tool search for charts only under the `charts` directory
          - --chart-search-root=charts

---
repos:
  - repo: https://github.com/norwoodj/helm-docs
    rev:  ""
    hooks:
      - id: helm-docs-built
        args:
          # Make the tool search for charts only under the `charts` directory
          - --chart-search-root=charts

---
repos:
  - repo: https://github.com/norwoodj/helm-docs
    rev:  ""
    hooks:
      - id: helm-docs-container
        args:
          # Make the tool search for charts only under the `charts` directory
          - --chart-search-root=charts

---
repos:
  - repo: https://github.com/norwoodj/helm-docs
    rev:  ""
    hooks:
      - id: helm-docs-container
        entry: jnorwood/helm-docs:x.y.z
        args:
          # Make the tool search for charts only under the `charts` directory
          - --chart-search-root=charts

To run and generate documentation into READMEs for all helm charts within or recursively contained by a directory:

helm-docs
# OR
helm-docs --dry-run # prints generated documentation to stdout rather than modifying READMEs

The tool searches recursively through subdirectories of the current directory for Chart.yaml files and generates documentation for every chart that it finds.

You can mount a directory with charts under /helm-docs within the container.

Then run:

docker run --rm --volume "$(pwd):/helm-docs" -u $(id -u) jnorwood/helm-docs:latest

helm-docs supports a .helmdocsignore file, exactly like a .gitignore file in which one can specify directories to ignore when searching for charts. Directories specified need not be charts themselves, so parent directories containing potentially many charts can be ignored and none of the charts underneath them will be processed. You may also directly reference the Chart.yaml file for a chart to skip processing for it.

Umbrella Helm chart documentation can include dependency values with document-dependency-values flag. All dependency values will be merged into values of umbrella chart documentation.

If you want to include dependency values, but don't want to generate doc for each dependency:

• set chart-search-root parameter to directory that contains umbrella chart and all dependency charts.
• list all charts you want to generate doc using chart-to-generate flag
• set document-dependency-values flag to true

There are two important parameters to be aware of when running helm-docs. --chart-search-root specifies the directory under which the tool will recursively search for charts to render documentation for. --template-files specifies the list of gotemplate files that should be used in rendering the resulting markdown file for each chart found. By default --chart-search-root=. and --template-files=README.md.gotmpl.

If a template file is specified as a filename only as with the default above, the file is interpreted as being relative to each chart directory found. If however a template file is specified as a relative path, e.g. the first of --template-files=./_templates.gotmpl --template-files=README.md.gotmpl then the file is interpreted as being relative to the chart-search-root.

This repo is a good example of this in action. If you take a look at the .pre-commit-config.yaml file here, you'll see our search root is set to example-charts and the list of templates used for each chart is the _templates.gotmpl file in that directory and the README.md.gotmpl file in each chart directory.

If any of the specified template files is not found for a chart (you'll notice most of the example charts do not have a README.md.gotmpl) file, then the internal default template is used instead.

In addition to extra defined templates you specify in these template files, there are quite a few built-in templates that can be used as well:

The default internal template mentioned above uses many of these and looks like this:

{{ template "chart.header" . }}
{{ template "chart.deprecationWarning" . }}

{{ template "chart.badgesSection" . }}

{{ template "chart.description" . }}

{{ template "chart.homepageLine" . }}

{{ template "chart.maintainersSection" . }}

{{ template "chart.sourcesSection" . }}

{{ template "chart.requirementsSection" . }}

{{ template "chart.valuesSection" . }}

{{ template "helm-docs.versionFooter" . }}

The tool also includes the sprig templating library, so those functions can be used in the templates you supply.

This tool can parse descriptions and defaults of values from values.yaml files. The defaults are pulled directly from the yaml in the file.

It was formerly the case that descriptions had to be specified with the full path of the yaml field. This is no longer the case, although it is still supported. Where before you would document a values.yaml like so:

controller:
  publishService:
    # controller.publishService.enabled -- Whether to expose the ingress controller to the public world
    enabled: false

  # controller.replicas -- Number of nginx-ingress pods to load balance between.
  # Do not set this below 2.
  replicas: 2

You may now equivalently write:

controller:
  publishService:
    # -- Whether to expose the ingress controller to the public world
    enabled: false

  # -- Number of nginx-ingress pods to load balance between.
  # Do not set this below 2.
  replicas: 2

New-style comments are much the same as the old-style comments, except that while old comments for a field could appear anywhere in the file, new-style comments must appear on the line(s) immediately preceding the field being documented.

I invite you to check out the example-charts to see how this is done in practice. The but-auto-comments examples in particular document the new comment format.

Note that comments can continue on the next line. In that case leave out the double dash, and the lines will simply be appended with a space in-between, as in the controller.replicas field in the example above

The following rules are used to determine which values will be added to the values table in the README:

• By default, only leaf nodes, that is, fields of type int, string, float, bool, empty lists, and empty maps are added as rows in the values table. These fields will be added even if they do not have a description comment
• Lists and maps which contain elements will not be added as rows in the values table unless they have a description comment which refers to them
• Adding a description comment for a non-empty list or map in this way makes it so that leaf nodes underneath the described field will not be automatically added to the values table. In order to document both a non-empty list/map and a leaf node within that field, description comments must be added for both

e.g. In this case, both controller.livenessProbe and controller.livenessProbe.httpGet.path will be added as rows in the values table, but controller.livenessProbe.httpGet.port will not

controller:
  # -- Configure the healthcheck for the ingress controller
  livenessProbe:
    httpGet:
      # -- This is the liveness check endpoint
      path: /healthz
      port: http

Results in:

If we remove the comment for controller.livenessProbe however, both leaf nodes controller.livenessProbe.httpGet.path and controller.livenessProbe.httpGet.port will be added to the table, with or without description comments:

controller:
  livenessProbe:
    httpGet:
      # -- This is the liveness check endpoint
      path: /healthz
      port: http

Results in:

If you would like to define a key for a value, but leave the default empty, you can still specify a description for it as well as a type. This is possible with both the old and the new comment format:

controller:
  # -- (int) Number of nginx-ingress pods to load balance between
  replicas:

  # controller.image -- (string) Number of nginx-ingress pods to load balance between
  image:

This could be useful when wanting to enforce user-defined values for the chart, where there are no sensible defaults.

In cases where you do not want to include the default value from values.yaml, or where the real default is calculated inside the chart, you can change the contents of the column like so:

service:
  # -- Add annotations to the service, this is going to be a long comment across multiple lines
  # but that's fine, these will be concatenated and the @default will be rendered as the default for this field
  # @default -- the chart will add some internal annotations automatically
  annotations: []

The order is important. The first comment line(s) must be the one specifying the key or using the auto-detection feature and the description for the field. The @default comment must follow.

See here for an example.

In cases you would like to ignore certain values, you can mark it with @ignored tag:

# @ignored
service:
  port: 8080

In the old-style comment, if a key name contains any "." or " " characters, that section of the path must be quoted in description comments e.g.

service:
  annotations:
    # service.annotations."external-dns.alpha.kubernetes.io/hostname" -- Hostname to be assigned to the ELB for the service
    external-dns.alpha.kubernetes.io/hostname: stupidchess.jmn23.com

configMap:
  # configMap."not real config param" -- A completely fake config parameter for a useful example
  not real config param: value

Some helm chart values.yaml uses complicated structure for the key/value pairs. For example, it may uses a multiline string of Go template text instead of plain strings. Some values might also refer to a certain YAML/JSON object structure, like internal k8s value type, or an enum. For these use case, a standard markdown table format might be inadequate and you want to use HTML tags to render the table.

Some example use case on why you need advanced table rendering:

• Hyperlinking the value type to an anchor or HTML link somewhere for reference
• Collapsible value description using <summary> tags to save space
• Multiline default values as codeblocks, instead of one line JSON structure for readability
• Custom rendering, for colors, actions, bookmarking, cross-reference, etc
• Cascading the markdown file generated by helm-docs to be post-processed by Jamstack into a static HTML docs site.

In order to accomodate this, helm-docs provides an extensible and flexible way to customize rendering.

1. Use the HTML value renderer instead of the default markdown format

You can use chart.valuesSectionHtml to render the values table as HTML tags, instead of using chart.valuesSection. Using HTML tables provides more flexibility because it can be processed by markdown viewer as a nested blocks, instead of one row per line. This allows you to customize how each columns in a row are rendered.

2. Overriding built-in templates

You can always overrides or redefine built-in templates in your own _templates.gotmpl file. The built-in templates can be thought of as a template hook. For example, if you need to change the HTML table, for example to add a new column, or define maximum width/height, you can override chart.valuesTableHtml. Your overrides will then be called by chart.valuesSectionHtml.

You can add your own rendering logic for each column. For example, we have chart.valueDefaultColumnRender that is used to render "default value" column for each rows. If you want to override how helm-docs render the "type" column, just define your own rendering template and call it from chart.valuesTableHtml for each of the rows.

3. Using the metadata of each rows of values

Custom styling and rendering can be done as flexible as you want, but you still need a metadata that describes each rows of values. You can access this information from the templates.

When you override chart.valuesTableHtml, as you can see in the original definition in func getValuesTableTemplates() pkg/document/template.go, we iterates each row of values. For each "Value", it is modeled as a struct defined in valueRow struct in pkg/document/model.go. You can then use the fields in your template.

Some fields here are directly referenced from values.yaml:

• Key: the full name of the key referenced in values.yaml
• Type: the type of the value of the key in values.yaml. Can be automatically inferred from YAML structure, or annotated using # -- (mytype) where mytype can be any string that you refer as the type of the value.
• NotationType: the notation of the type used to render the default value. If Type refers to the data type of the value, then NotationType refers to how this value should be written/rendered by helm-docs. Generally helm-docs only remembers the notation type, but it was the writer's responsibility to make a template tag to render a specific notation type. Annotate the key with # @notationType -- (mynotation) where mynotation is an identifier to tell the renderer how to write the value.
• Default: this is the default value of the key, found from values.yaml. It is either inferred from the YAML structure or defined using # @default -- my default value annotation, in case you need to show other example values.
• Description: this is the description of the key/value, taken from the comments found in the values.yaml for the referred key.
• LineNumber: this is the line number associated with where the key is declared. You can use this to construct an anchor to the actual values.yaml file.

Note that helm-docs only provides these information, but the default behaviour is to always render it in plain Markdown file to be viewed locally.

4. Use markdown files generated by helm-docs as intermediary files to be processed further

Public helm charts sometimes needs to be published as static content instead of just stored in a repository. This is needed for helm users to be able to view or browse the chart options and dependencies.

It is often more than enough to just browse the chart values options on git hosting that is able to render markdown files as a nice HTML page, like GitHub or GitLab. However, for a certain use case, you may want to use your own documentation generator to host or publish the output of helm-docs.

If you use some kind of Jamstack like Gatsby or Hugo, you can use the output of helm-docs as an input for these doc generator. A typical use case is to override helm-docs built-in template so that it renders a markdown or markdownX files to be processed by Gatsby or Hugo into a static Web/Javascript page.

For a more concrete examples on how to do these custom rendering, see example here

Sometimes you might want to enforce helm-docs to fail when some values are not documented correctly.

By default, this option is turned off:

./helm-docs -c  example-charts/helm-3
INFO[2023-06-29T07:54:29-07:00] Found Chart directories [.]
INFO[2023-06-29T07:54:29-07:00] Generating README Documentation for chart example-charts/helm-3

but you can use the -x flag to turn it on:

helm-docs -x -c  example-charts/helm-3
INFO[2023-06-29T07:55:12-07:00] Found Chart directories [.]
WARN[2023-06-29T07:55:12-07:00] Error parsing information for chart ., skipping: values without documentation:
controller
controller.name
controller.image
controller.extraVolumes.[0].name
controller.extraVolumes.[0].configMap
controller.extraVolumes.[0].configMap.name
controller.livenessProbe.httpGet
controller.livenessProbe.httpGet.port
controller.publishService
controller.service
controller.service.annotations
controller.service.annotations.external-dns.alpha.kubernetes.io/hostname

The CLI also supports excluding fields by regexp using the -z argument

helm-docs -x -z="controller.*" -c  example-charts/helm-3
INFO[2023-06-29T08:18:55-07:00] Found Chart directories [.]
INFO[2023-06-29T08:18:55-07:00] Generating README Documentation for chart example-charts/helm-3

Multiple regexp can be passed, as in the following example:

helm-docs -x -z="controller.image.*" -z="controller.service.*"  -z="controller.extraVolumes.*"  -c  example-charts/helm-3
INFO[2023-06-29T08:21:04-07:00] Found Chart directories [.]
WARN[2023-06-29T08:21:04-07:00] Error parsing information for chart ., skipping: values without documentation:
controller
controller.name
controller.livenessProbe.httpGet
controller.livenessProbe.httpGet.port
controller.publishService

It is also possible to ignore specific errors using the -y argument.

helm-docs -x -y="controller.name" -y="controller.service"  -c  example-charts/helm-3
INFO[2023-06-29T08:23:40-07:00] Found Chart directories [.]
WARN[2023-06-29T08:23:40-07:00] Error parsing information for chart ., skipping: values without documentation:
controller
controller.image
controller.extraVolumes.[0].name
controller.extraVolumes.[0].configMap
controller.extraVolumes.[0].configMap.name
controller.livenessProbe.httpGet
controller.livenessProbe.httpGet.port
controller.publishService
controller.service.annotations
controller.service.annotations.external-dns.alpha.kubernetes.io/hostname