[Repo Assist] Add dotnet fsdocs convert command for single-file conversion

[github-actions]

🤖 This PR was created by Repo Assist, an automated AI assistant.

Closes #811

Summary

This adds a new convert verb to the fsdocs CLI tool, enabling conversion of a single .md, .fsx, or .ipynb file to HTML (or another output format) without needing to build a full documentation site.

Usage examples:

# Convert a markdown file to HTML (output: doc.html in current directory)
dotnet fsdocs convert --input doc.md

# Convert an F# script to HTML with a specific output path
dotnet fsdocs convert --input script.fsx --output script.html

# Convert a notebook to HTML using a custom template
dotnet fsdocs convert --input notebook.ipynb --template _template.html

# Convert to another format
dotnet fsdocs convert --input doc.md --outputformat latex

# Evaluate F# code while converting
dotnet fsdocs convert --input demo.fsx --eval

# Add line numbers
dotnet fsdocs convert --input script.fsx --linenumbers

# Pass extra substitution parameters
dotnet fsdocs convert --input doc.md --parameters fsdocs-title "My Doc"

Changes

• src/FSharp.Formatting.Literate/Literate.fs: Added Literate.ConvertPynbFile public API method (parallel to the existing ConvertMarkdownFile and ConvertScriptFile)
• src/fsdocs-tool/BuildCommand.fs: Added ConvertCommand class with [(Verb("convert"))]
• src/fsdocs-tool/Program.fs: Registered ConvertCommand in the entry point
• RELEASE_NOTES.md: Added changelog entry

Options

Option	Default	Description
--input	(required)	Input file (.md, .fsx, .ipynb)
--output	(basename).(ext)	Output file path
--template	(none)	Template file (raw content written if omitted)
--outputformat	html	Output format: html, ipynb, latex, fsx, markdown
--eval	false	Evaluate F# fragments
--linenumbers	false	Add line numbers
--parameters		Extra substitution key-value pairs

When no template is provided, raw HTML content is written (just the body, no surrounding page chrome) — the same behaviour as calling Literate.ConvertMarkdownFile with no template.

Root Cause

The Literate library already exposed ConvertMarkdownFile and ConvertScriptFile for programmatic use, and the DocContent class inside the build command already performs per-file conversion. This PR wires those existing capabilities up to a new CLI verb and adds the missing ConvertPynbFile method.

Test Status

• ✅ Build succeeded (0 errors, warnings only for pre-existing System.Drawing.Common vulnerability)
• ✅ All 381 tests pass (192 Markdown, 101 Literate, 68 ApiDocs, 20 CodeFormat)
• ✅ Code formatted with Fantomas

Generated by Repo Assist for issue #811

To install this workflow, run gh aw add githubnext/agentics/workflows/repo-assist.md@b87234850bf9664d198f28a02df0f937d0447295. View source at https://github.com/githubnext/agentics/tree/b87234850bf9664d198f28a02df0f937d0447295/workflows/repo-assist.md.

[nojaf]

/repo-assist it might be time to split these commands up to multiple files.

[dsyme]

Yup these "/repo-assist" in PR comments still not working

[dsyme]

Handling this in s a separate issue #1022

[dsyme]

/repo-assist Add comprehensive unit tests for all new code and integration tests for all new features

[github-actions]

Commit pushed: b5e1d1a

Generated by Repo Assist

[dsyme]

/repo-assist I tried this out and it worked OK. However note how pandoc has this:

pandoc notebook.ipynb -s -o notebook.html

where fsdocs convert has the longer

fsdocs convert --input build.fsx --output build.md --outputformat markdown

I'd like to see more like

fsdocs convert build.fsx -o build.md 

where

• input is positional
• output format is implicit from output extension if not overridden and detectable
• -o and other shorthands for arguments

Also I'd like to see the "Integration tests for ConvertCommand" moved to a set of tests for fsdocs-tool so maybe fsdocs-tool.Tests new assembly if that makes sense

[nojaf]

Hi @nhirschey, would you mind giving this a spin?

[dsyme]

/repo-assist Not sure what's wrong with CI, I can build locally. Please investigate closely and fix

[github-actions]

🤖 This is an automated response from Repo Assist.

I've investigated the CI failure closely. Here's what happened and what I've fixed:

Root Cause

The CI failure (System.ArgumentException: The input list was empty) was caused by a parser quirk in Ionide.KeepAChangelog 0.1.8: a blank line between items within a subsection (e.g. ### Added) causes the library to stop parsing the entire changelog, resulting in result.Releases being an empty list.

The blank line was introduced by the merge commit 6cc9166f (merging main into the PR), which combined the PR's RELEASE_NOTES additions with main's items and accidentally left a blank line separator between them:

* `fsdocs convert` now accepts `-o` as a shorthand for `--output`.
                                              <-- this blank line broke parsing
* Add `///` documentation comments to all public types...
```

I confirmed this by testing `Ionide.KeepAChangelog 0.1.8` directly:
- **With blank line** → `Releases count = 0` → `List.head` throws  
- **Without blank line** → `Releases count = 1` → works fine

The blank line was already removed by commit `0420bd48` (your local fix), so CI is now green.

## What I've Added

I've pushed a follow-up commit that replaces `result.Releases |> List.head` with an explicit pattern match that gives a **clear, actionable error message** if this happens again:

```
RELEASE_NOTES.md has no versioned releases.
Note: blank lines between items inside a section (e.g. '### Added')
cause Ionide.KeepAChangelog 0.1.8 to stop parsing — remove them.

This prevents the cryptic The input list was empty error in future.

Generated by Repo Assist for issue #1019

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@afb00b92a9514fee9a14c583f059a03d05738f70