Docs CLI

Project-local Word template fill with a small localhost editor.

memgrep docs setup
memgrep docs status
memgrep docs list
memgrep docs fill <template.docx> --context '{"title":"Retro"}' [--name slug]
memgrep docs fill <template.docx> --context-file ./fields.json --name sprint-retro
memgrep docs edit [slug] [--port 8791]

Alias: memgrep doc …

Layout

<cwd>/
  .memgrep/
    templates/     # source .docx with {{ field }} / {{ dotted.path }}
    docs/          # filled .docx + .context.json sidecars

• setup — create templates and docs dirs (idempotent)
• status — show paths and counts
• list — list templates in .memgrep/templates
• fill — apply JSON context; writes .memgrep/docs/<slug>.docx and .context.json
• edit — start 127.0.0.1 web UI (default port 8791); save re-fills from the original template

Placeholders

Scalar fields:

{{ title }}
{{ meeting.date }}

Rich text (Markdown → Word)

Use a | rich placeholder; it may sit alone in a paragraph or next to a label:

{{ deliberations | rich }}
TEST INFORMATION:{{ test_information | rich }}
{{ testing_scope | rich }}Out of Scope:

Sole-paragraph rich replaces that paragraph with Markdown→Word paragraphs. Mixed paragraphs keep the label text (preserving paragraph properties) and inject the Markdown paragraphs at the placeholder site.

Context value is Markdown. Supported:

• **bold** / *italic*
• # / ## / ### headings (bold; body stays Arial 12pt)
• bullet / numbered lists (nest for indentation)
• > blockquote indentation

Rich output always forces Arial 12pt (scalars/loops still inherit the template run styles).

{
  "deliberations": "## 3.1 Review\n\n**Closed:** DSTV pending…\n\n- Item one\n  - Nested detail\n\n> Note indented"
}

The docs editor shows a Markdown toolbar for | rich fields.

Table row loops

Repeat a table row (or block of rows) with Nunjucks-style tags:

{% for item in attendees %}{{ item.name }}{% endfor %}

Put {% for item in attendees %} and {% endfor %} in the same row, or on their own marker rows around the template row. Context should include an array:

{
  "title": "Retro",
  "attendees": [
    { "name": "Ada", "role": "Chair" },
    { "name": "Bob", "role": "Scribe" }
  ]
}

The localhost editor (docs edit / docs_serve) shows iterable collections with Add row / Remove controls.

Block / table loops

Repeat an entire Word table (or a contiguous block of tables ± paragraphs) once per item in a context array. Put the {% for %} / {% endfor %} markers on their own paragraphs immediately before and after the table. Nested row loops inside the block resolve against the outer item (Nunjucks scoping):

{% for case in test_cases %}
  [ entire test-case table ]
  TEST CASE #{{ case.number }}
  {{ case.title }}
  {% for p in case.steps %}
    {{ p.number }} | {{ p.test_step_input }} | …
  {% endfor %}
{% endfor %}

{
  "task_id": "TASK-123",
  "test_cases": [
    {
      "number": "1",
      "title": "Login happy path",
      "steps": [
        {
          "number": "1",
          "test_step_input": "Open login",
          "expected_results": "Form shown",
          "actual_results": "Form shown",
          "pass_fail": "PASS"
        }
      ]
    }
  ]
}

docs extract reports nested iterables (test_cases as block → steps as rows). The editor shows Add case / nested Add row for steps. An empty array removes the template table and markers.

Existing single-table row loops are unchanged.

MCP

Always registered (no credentials): docs_setup, docs_list_templates, docs_extract, docs_fill, docs_list, docs_serve. Paths resolve from the MCP server process cwd. See Optional Suites.