Skip to content

README.md

Latest commit

 

History

History
87 lines (67 loc) · 3.16 KB

README.md

File metadata and controls

87 lines (67 loc) · 3.16 KB
Raw

animus-document-engine

An MCP (stdio) service that renders, parses, converts, and edits Office documents — Excel (.xlsx), PowerPoint (.pptx), Word (.docx) — from a structured JSON spec. It is identity-free and stateless: it never touches subjects, S3, or users. The animus-launchapp portal orchestrates it (storage + scoping + subjects); this engine is the reusable rendering core.

Part of REQUIREMENT-053 — Documents as subjects.

Why a spec

The spec is the source of truth. An AI authors a spec (not raw OOXML); a deterministic renderer turns it into a real file. Editing = mutate the spec and re-render — lossless, reviewable, and diffable. Binary payloads cross the MCP boundary as base64.

Tools

Tool In Out
render_document format, spec base64 OOXML bytes
parse_document format, bytes_base64 spec (xlsx only in v1)
apply_edits format, spec, ops[] edited, re-validated spec
convert_document bytes_base64, to base64 converted bytes (e.g. pdf)
preview_document bytes_base64, kind (pdf|png), page base64 PDF or PNG thumbnail

convert_document / preview_document use LibreOffice compiled to WebAssembly (@matbee/libreoffice-converter) — no native soffice needed. The ~112 MB WASM lives in this service, never in the portal image.

Spec shapes (v1)

// xlsx
{ "sheets": [ { "name": "Revenue",
                "columns": [ { "header": "Month" }, { "header": "USD" } ],
                "rows": [ ["Jul", 1200], ["Aug", 1500] ] } ] }

// pptx
{ "title": "Kickoff",
  "slides": [ { "title": "Agenda", "bullets": ["Scope", "Timeline"] },
              { "title": "Next", "body": "…", "notes": "speaker note" } ] }

// docx
{ "title": "Spec",
  "blocks": [ { "type": "heading", "level": 1, "text": "Overview" },
              { "type": "paragraph", "text": "…" },
              { "type": "bullets", "items": ["a", "b"] },
              { "type": "table", "rows": [["a","b"],["c","d"]] } ] }

apply_edits ops: {op:"set",path,value}, {op:"append",target,value}, {op:"remove",target,index}, {op:"replace",value} (target = sheets|slides|blocks).

Run

npm install
npm run build
node dist/index.js        # MCP stdio server

Wire it into an Animus workflow phase (or any MCP client) as a stdio server:

{ "mcpServers": { "document-engine": { "command": "node", "args": ["dist/index.js"] } } }

Dev

npm run dev        # tsx watch (stdio server)
npm run typecheck  # tsc --noEmit
npm run smoke      # render all 3 formats + xlsx parse/edit round-trip
tsx scripts/mcp-smoke.ts   # spawn the built server, call tools over MCP

Scope (v1) & roadmap

  • Generate all three formats — solid (pptxgenjs / docx / exceljs).
  • Parse — xlsx (data + headers). pptx/docx in-place parsing is DEFERRED to a future Python worker (python-pptx / python-docx are the only libraries that faithfully round-trip arbitrary decks/docs); the engine errors loudly rather than silently dropping content.
  • Convert/preview — via LibreOffice-WASM.