For IT administrators and technical reviewers

What the Desktop App and Word Add-in install,
run, and connect to.

Pelaikan's installed configuration is designed for legal teams that need Microsoft Word integration and local document handling. The Desktop App provides the local runtime; the Word Add-in provides the in-document interface.

Local
Python engine
on the user's device
127.0.0.1
Loopback API
for local services
Word
Office.js add-in
for document access

Deployment Architecture

Three components work together, with different responsibilities.

01

Desktop App

A native Windows/macOS application built with Tauri. It renders the Pelaikan interface in a system webview, manages local application permissions, and starts the bundled engine after the user signs in.

  • Ships as platform-specific desktop installers.
  • Includes the packaged web UI and application assets.
  • May access user-selected folders and local application data to build citation databases.
03

Word Add-in

An Office.js taskpane hosted from Pelaikan's Firebase site. It reads the active Word document through Microsoft Office APIs and uses the Desktop App's local engine/session for installed workflows.

  • Runs inside Microsoft Word as a standard Office web add-in.
  • Uses localhost origins to communicate with the local engine.
  • Uses the cloud API as an authenticated bridge where needed.

Typical Data Flow

The installed workflow keeps document-heavy operations on the endpoint and uses cloud services for identity, orchestration, and configured AI/API calls.

1

User signs in

The Desktop App authenticates the user and binds the installation to a client identifier. Session tokens are used for authenticated calls to the Pelaikan API and local engine.

2

Local engine starts

The Tauri shell launches the bundled Python sidecar. The engine listens on loopback and exposes local endpoints used by the Desktop App and Word Add-in.

3

Documents are parsed and indexed

User-selected files are processed locally for document structure, footnotes, citation extraction, OCR, embeddings, and local database search.

4

Resolution and congruence checks run

Pelaikan attempts local database resolution first where available, then uses configured APIs, web search, cloud libraries, or AI providers for source retrieval and support analysis.

IT Review Checklist

Items commonly reviewed before deployment in managed environments.

Local Access

The app needs access to user-selected documents, local application data, and local database folders. It does not require inbound network access from other machines.

Loopback Networking

Allow the app, Word, and the embedded webview to reach 127.0.0.1/localhost on the engine ports used by the installation, primarily 5050 with fallbacks used in development and add-in contexts.

Outbound Services

Allow outbound HTTPS to Pelaikan hosting and API domains, Firebase/Google identity services, and Microsoft identity endpoints where configured for sign-in or add-in operation.

Office Add-in Policy

The production manifest points Word to the hosted taskpane and declares both Pelaikan cloud domains and localhost engine origins. Centralized deployment can be handled through normal Microsoft 365 add-in administration.

AI and External APIs

Some citation resolution, OCR, embeddings, and congruence steps may call configured cloud AI or source APIs. Enterprise deployments can review provider choices and regional settings as part of data protection approval.

Logs and Local Stores

The engine writes diagnostic logs locally and maintains local SQLite/vector index data for user databases. These files remain under the user's endpoint storage unless your organization backs them up or syncs them separately.

Security Boundary in Plain English

The Desktop App is the trusted local runtime. It holds the permissions needed to read selected files, launch the local engine, and use native networking where browser/webview restrictions would otherwise block localhost traffic. The Word Add-in is a hosted Office interface that relies on Office APIs for document access and the local engine for installed-client workflows.

The cloud backend is primarily the identity, metering, and orchestration layer for installed clients. For web-only use, processing can occur server-side; for Desktop App and Word Add-in use, the architecture is intentionally oriented around local processing and explicit outbound calls.

Read the Data & Privacy Page

Need a Technical Review Package?

We can provide the current manifest, endpoint list, installer details, and data-processing notes for your organization's IT, security, or procurement review.

Request Deployment Details