Current Product Scope

This document describes the product behavior that is implemented in the repository today.

What The Product Does Today

Iris Translation v2 is a governed DOCX translation workflow built on Django, Celery, Aspose-based DOCX processing, scoped terminology, translation memory, and a TMX-based review loop.

The current repository supports:

DOCX job intake with preflight inspection
extraction into anchored translation units
translation with either approved-memory reuse or a provider call
automated segment verification and document-level QA
TMX export for review and TMX import back into the job
final DOCX reassembly
glossary, project, provider-profile, and language-pair-policy APIs
candidate-memory capture from reviewed imports and promotion into approved memory
artifact listing and controlled download links
job comparison, review coverage reporting, and audit-event search
maintenance cleanup and integrity reporting for artifacts, candidate memory, and delivery gaps
a Django-template operator console for job operations and project coverage

Supported Inputs

Current supported inputs in the public API are:

source DOCX uploads for job creation
glossary CSV uploads for import
reviewed TMX uploads for human-reviewed job updates
JSON configuration updates for projects, provider profiles, and language-pair policies

Supported Outputs

The repository can currently produce:

serialized job records and job statistics
per-unit verification results
source, intermediate, review, replay, and delivery artifacts
reviewed TMX export files
replay ZIP packages
final translated DOCX files
project review-coverage summaries
audit-event search results

Supported Workflow Shape

The implemented workflow is:

intake and preflight
extraction
translation and verification
automated QA finalization
optional TMX export/import review loop
reassembly to final DOCX

The workflow can run:

synchronously through the job action endpoints, or
asynchronously through Celery when async workflow is enabled

Knowledge And Governance Features Present Today

The repo currently includes:

domain-pack, project, and language-pair-policy scoping
provider profiles for named translation runtime configurations
glossary scope precedence across domain, project, and document family
candidate-memory capture from reviewed TMX imports
explicit promotion from candidate memory into approved memory
per-job policy snapshots and statistics snapshots
audit events for workflow and control actions
operator/admin access control on the JSON API

Current Hard Scope Boundaries

These are true of the current repo and should be treated as present-day constraints:

the public intake API is DOCX-only
review import/export is TMX-based
the checked-in runtime defaults to a local-development PostgreSQL/Redis/MinIO/OpenSearch stack, with SQLite kept as a fallback
the checked-in Docker stack is a local dev stack, not a finished production deployment shape
the current supported operator surfaces are the Django-template console and the JSON API, while Django admin remains internal-only

What Is Not In Scope In The Current Repo

The repository does not currently implement:

non-DOCX source intake
a separate SPA or standalone frontend beyond the Django-template console
built-in multi-service production deployment manifests
Django admin as a supported day-to-day operator workflow

For scaling guidance, see Operations And Scaling. For unresolved product or delivery decisions, see Open Questions.