@awacloud/odf

Pure JavaScript library to read and write OpenDocument documents (ODF — .odt, .ods, .odp) in the browser. No npm dependency beyond the monorepo framework: ZIP, XML parsing/serialization and the ODF package container are either implemented here or consumed from @awacloud/fw.

The package exposes a typed document model for the three formats (a serializable JSON tree), a core set of parsers and orchestrators, and opt-in extras that promote more of the ODF vocabulary into typed fields. Whatever the typed model does not cover is preserved: unknown XML stays in _extras, and write(read(x)) re-emits the package parts it does not regenerate (pictures, thumbnails, embedded objects) byte-for-byte with their manifest media types, together with the read styles, settings and metadata; meta:generator is rewritten to @awacloud/odf. One limitation: a frame anchored directly in a paragraph is re-emitted after the paragraph's runs, not at its original position (inside a text:span, frames, fields, spacing and nested spans keep their position). It mirrors the architecture of the sibling package @awacloud/ooxml.

Installation

npm install @awacloud/odf

In the browser, via import map:

<script type="importmap">
{ "imports": {
    "@awacloud/fw":   "/node_modules/@awacloud/fw/src/main.js",
    "@awacloud/fw/":  "/node_modules/@awacloud/fw/src/",
    "@awacloud/odf":  "/node_modules/@awacloud/odf/src/main.js",
    "@awacloud/odf/": "/node_modules/@awacloud/odf/src/"
}}
</script>

Quick Start

Core (factory wiring)

Every odf module is a { name, dependencies, factory } descriptor resolved by an @awacloud/fw ModuleRuntime. The package entry exports the fw modules the odf modules depend on (fw_require) next to the odf ones (modules):

import { ModuleRuntime } from '@awacloud/fw/core/runtime.js';
import { fw_require, modules } from '@awacloud/odf';

const runtime = new ModuleRuntime();
runtime.registerAll(fw_require);
runtime.registerAll(modules);
const odt = runtime.resolve('odt');

const bytes = odt.write(odt.fromText(['Hello, world.']));
const decoded = odt.read(bytes);
console.log(odt.toText(decoded)); // → "Hello, world."

Typed paragraph manipulation

Continuing with the odt instance resolved above:

const doc = {
    body: [
        odt.paragraph('Title', { styleName: 'Title' }),
        {
            type: 'paragraph',
            runs: [
                { type: 'text', value: 'plain ' },
                { type: 'span', value: 'styled', styleName: 'T1' },
                { type: 'line-break' },
                { type: 'text', value: 'next line' }
            ]
        }
    ]
};
const styledBytes = odt.write(doc, { meta: { title: 'Doc', creator: 'Alice' } });

Reading styles back

odt.read() turns a style reference into semantic fields (bold / italic / strike / monospace, list numbering, bordered cells, margins-aligned tables) only when the style is fully mapped and comes from the content.xml automatic styles or from the styles.xml office:styles named styles; it never follows style:parent-style-name chains, ignores the automatic styles of styles.xml, and leaves every other style as an opaque styleName — see docs/guide/read-write-odt.md.

Pre-built bundles (dist/)

Each assembly root (odt, odt-large, odt-full, and the same three for ods and odp) ships as a two-surface build, generated by tools/generate-bundles.mjs (repository only, not part of the published package; a driver over @awacloud/tool-prebuild-generator) and regenerated with:

bun run gen:bundles

dist/build/index.js (@awacloud/odf/build/index.js) is an fw-mode barrel re-exporting the whole @awacloud/odf namespace (the registration arrays and every named descriptor) for bulk registration on an @awacloud/fw runtime. See docs/api/bundles/prebuilt/README.md for the resolve-key table and usage examples.

Source structure

src/
├── main.js          entry point — `modules[]`, `extras[]`, `bundle[]`, `fw_require[]` + named descriptors
├── errors.js        OdfError + ParseError/RenderError/ContractError
├── (xml parser consumed from `@awacloud/fw/io/codec/xml.js`)
├── _shared/         odfShared (namespaces, XML helpers) + odfWalker (hook engine)
├── pkg/             ODF container (ZIP + mimetype + manifest.xml)
├── meta/            meta.xml (dc:* + meta:*)
├── settings/        settings.xml (preserve-unknowns)
├── style/           styles.xml, automatic styles, page layout, master pages
├── text/            text:p, headings, lists, sections, fields, tracked changes, style registry
├── table/  draw/  chart/  math/  form/  dr3d/  number/  mc/
├── odt/  ods/  odp/ top-level orchestrators (+ their walkers)
├── extra/           opt-in extras (typed or passthrough)
└── bundles/         `*-large` / `*-full` bundle descriptors

Hook .use(...)

The odt / ods / odp orchestrators expose a walker extensible via .use(...extensions): each extra can publish hydrate* / dehydrate* hooks (Paragraph, Span, Heading, List, Table, Cell, Frame, Slide, Metadata, Settings, Styles — each orchestrator walks the node types it carries). Idempotent (duplicates are filtered out). See docs/guide/extending.md and docs/api/ for the per-module coverage.

Documentation

• docs/README.md — full index
• docs/guide/getting-started.md — first steps
• docs/guide/read-write-odt.md
• docs/guide/pkg-overview.md — ODF container
• docs/api/ — API reference per module

Tests

Co-located unit tests (src/**/*.test.js) + integration tests in tests/; run them from the monorepo root:

bun test packages/front/office/odf/

Coverage

• Core — .odt / .ods / .odp orchestrators + typed parsers for text:* / table:* / draw:* / style:* / chart:* / math:math / form:* / dr3d:* + mimetype, manifest, meta, settings, styles.
• Opt-in extras (extras in src/main.js, one file per extra under src/extra/), in four tiers:
  • P0 deep typing: tracked-changes, fields, lists, table-advanced, page-styles, properties-typed, shapes, presentation.
  • P1 complementary deep typing: text-meta-extended, sections-advanced, toc/index, image-extended, chart-typed, animations-smil, forms-controls, number-format-extended, meta-extended, math-mathml.
  • P2 secondary-domain typing: dr3d-3d, database-sources, settings-extended, script-macros, dsig-signatures.
  • P3 *-misc passthrough (catch-all _passthrough: true): text, style, draw, table, office, legacy-staroffice.
• Bundles ready to plug in: odt|ods|odp-large (core + P0) and odt|ods|odp-full (*-large + P1/P2/P3).

Design choices

• Typed document model — serializable JSON tree rather than a flat object.
• _extras mechanism — untyped elements and attributes are preserved in _extras and re-emitted on write.
• mimetype STORED — first ZIP entry, compression 0, as required by the ODF spec.
• Single dependency — only @awacloud/fw (workspace).
• Worker-safe — each factory is self-sufficient (serializable via factory.toString()).
• Browser-only — no Node API: Uint8Array, TextEncoder, TextDecoder only.
• Bounded ZIP reading — read() caps entry count, total uncompressed size and per-entry ratio (see the guide's security section).
• Mirror of @awacloud/ooxml — same factory pattern, same conventions, same docs format.

See also

• @awacloud/ooxml — sibling OOXML library
• @awacloud/fw — shared runtime + zip + crc32 + codecs
• CHANGELOG.md — L0 → L4 history

Exposed sub-paths

Maturity

L4 (awa.maturity in package.json) — the highest level of the monorepo's maturity scale: typed .odt / .ods / .odp, tested and documented to publication quality.

Licence

AGPL-3.0-only — see LICENSE.

Copyright (c) 2026 AwaCloud SAS

This package is also available under a commercial licence from AwaCloud SAS, as stated in NOTICE.

Project

• Website: https://awaforge.eu
• Source: packages/front/office/odf
• Issues: https://github.com/awacloud/awa/issues
• Security policy: https://github.com/awacloud/awa/blob/@awacloud/odf@1.0.0/SECURITY.md

npm install @awacloud/odf@1.0.0