Skip to main content

Knowledge Translation Transmutation Core - Transforming translations into gold-standard quality

Project description

KTTC

English · Русский · 中文 · हिन्दी · فارسی

KTTC - Knowledge Translation Transmutation Core

CI CodeQL PyPI Python Version License

Autonomous AI-powered translation quality assurance

KTTC uses specialized multi-agent systems to automatically detect, analyze, and fix translation quality issues following the industry-standard MQM (Multidimensional Quality Metrics) framework. Get production-ready translation quality in seconds, not hours.


Key Features

Core Analysis

  • Multi-Agent QA System - Specialized agents analyze accuracy, fluency, terminology, style, hallucination, and context
  • MQM Scoring - Industry-standard quality metrics used in WMT benchmarks
  • Language-Specific Agents - 5 native-level fluency agents for English, Chinese, Russian, Hindi, and Persian
  • 60+ Domain Glossaries - Automotive, customs, financial, legal, logistics, medical + language traps

Smart Agent Orchestration

  • Weighted Consensus - Agents with different trust weights, confidence scoring, agreement metrics
  • Multi-Agent Debate - Cross-verification between agents reduces false positives by 30-50%
  • Dynamic Agent Selection - Auto-selects 2-5 agents based on text complexity (30-50% cost savings)
  • Agent Presets - minimal (2 agents), default (3), full (5) for quick configuration
  • Self-Assessment Retry - Agents evaluate their confidence and retry on low certainty

MQM Profiles System

  • Built-in Profiles - default, strict, minimal, legal, medical, marketing, literary, technical
  • YAML Custom Profiles - Define agent selection, weights, severity multipliers, thresholds
  • Domain Adaptation - Medical: 98% threshold + hallucination agent; Legal: terminology focus
  • Quality Gates - Configurable pass/fail thresholds per domain (88-98%)

Literary Style Analysis

  • StyleFingerprint - Automatic detection of literary style patterns (Burrows Delta method)
  • Style Preservation Agent - Evaluates authorial voice preservation in literary translations
  • Country-Specific Patterns - Leskov skaz, Platanov pleonasms, Joyce stream, Hafez ghazals, Chhayavad poetry
  • Fluency Tolerance - Adjustable fluency weights for intentional stylistic deviations

Language Intelligence

  • Language Traps Detection - 60+ glossaries detect homophones, false friends, idioms, phrasal verbs, paronyms
  • Self-Check / Proofreading - Grammar, spelling, punctuation checking using school curriculum rules
  • School Curricula - FGOS (Russia), UK GPS (Britain), NCERT (India), PEP (China), Iranian grammar
  • Fast Lint Mode - Rule-based checking without LLM, ideal for CI/CD and pre-commit hooks
  • Auto-Correction - LLM-powered error fixing with iterative refinement (TEaR loop)

Enterprise Infrastructure

  • Smart Routing - Automatically selects optimal models based on text complexity (60% cost savings)
  • XLSX Export - Excel reports with Summary, Errors, Breakdown sheets for enterprise reporting
  • Translation Memory - Semantic search with quality tracking and reuse
  • Glossary Management - Custom terminology + 60+ built-in multilingual glossaries
  • Batch Processing - Process thousands of translations in parallel with CSV/JSON filtering
  • CI/CD Ready - GitHub Actions integration, exit codes, JSON/Markdown/HTML/XLSX output
  • Multi-LLM Support - OpenAI, Anthropic, GigaChat, YandexGPT with LanguageTool integration
  • Usage Analytics - Token counts, API costs, and call statistics in reports

Performance: 90% cost reduction vs manual review • 100-1000x faster • 95+ MQM quality target


Try KTTC Online

Experience KTTC without installation:

Open in Colab Streamlit Demo Open in Codespaces

  • Google Colab - Interactive tutorial with examples (5 min, no setup)
  • Streamlit Demo - Web UI to test your own translations (no code required)
  • GitHub Codespaces - Full dev environment in browser (for contributors)

Quick Start

1. Install

pip install kttc

Optional language enhancements:

pip install kttc[english]        # English: LanguageTool (5,000+ grammar rules)
pip install kttc[chinese]        # Chinese: HanLP (measure words, particles)
pip install kttc[hindi]          # Hindi: Indic NLP + Stanza + Spello
pip install kttc[persian]        # Persian: DadmaTools (spaCy-based)
pip install kttc[all-languages]  # All language helpers

2. Set API Key

export KTTC_OPENAI_API_KEY="sk-..."
# or
export KTTC_ANTHROPIC_API_KEY="sk-ant-..."

3. Check Translation Quality

kttc check source.txt translation.txt --source-lang en --target-lang es

Output:

✅ MQM Score: 96.5 (PASS - Excellent Quality)
📊 5 agents analyzed translation
⚠️  Found 2 minor issues, 0 major, 0 critical
✓ Quality threshold met (≥95.0)

That's it! KTTC works out of the box with smart defaults:

  • ✅ Smart routing (auto-selects cheaper models for simple texts)
  • ✅ Auto-glossary (uses 'base' glossary if exists)
  • ✅ Auto-format (detects output format from file extension)

Commands

KTTC provides a unified CLI with smart auto-detection:

kttc check source.txt translation.txt          # Single quality check
kttc check source.txt t1.txt t2.txt t3.txt     # Auto-compares multiple translations
kttc check translations.csv                     # Auto-detects batch mode (CSV/JSON)
kttc check source_dir/ trans_dir/              # Auto-detects directory batch mode

kttc batch --file translations.csv              # Explicit batch processing
kttc compare --source src.txt -t t1 -t t2      # Compare translations side-by-side
kttc translate --text "Hello" --source-lang en --target-lang es  # Translate with QA
kttc benchmark --source text.txt --providers openai,anthropic    # Benchmark LLMs

# Glossary management (project + user global storage)
kttc glossary list                              # List all glossaries
kttc glossary create tech --from-csv terms.csv  # Create project glossary
kttc glossary create personal --from-csv my.csv --user  # Create user glossary

# 🥚 Self-check / Proofreading (new!)
kttc check article.md --self --lang ru          # Proofread without translation
kttc proofread article.md --lang ru             # Same as above (alias)
kttc lint article.md --lang ru --fix            # Quick rule-based check (no LLM)

See full command reference: docs/en/reference/cli-commands.md


Python API

import asyncio
from kttc.agents import AgentOrchestrator
from kttc.llm import OpenAIProvider
from kttc.core import TranslationTask

async def check_quality():
    llm = OpenAIProvider(api_key="your-key")
    orchestrator = AgentOrchestrator(llm)

    task = TranslationTask(
        source_text="Hello, world!",
        translation="¡Hola, mundo!",
        source_lang="en",
        target_lang="es",
    )

    report = await orchestrator.evaluate(task)
    print(f"MQM Score: {report.mqm_score}")
    print(f"Status: {report.status}")

asyncio.run(check_quality())

See full API reference: docs/en/reference/api-reference.md


Documentation

📚 Complete documentation is available in docs/

Quick Links

Documentation Structure

Following the Diátaxis framework:

  • 📚 Tutorials - Learn by doing (step-by-step guides)
  • 📖 Guides - Solve specific problems (how-to guides)
  • 📋 Reference - Look up technical details (API, CLI)
  • 💡 Explanation - Understand concepts (architecture, design)

Translations

Languages: 🇺🇸 English (primary) · 🇷🇺 Русский · 🇨🇳 中文 · 🇮🇳 हिन्दी · 🇮🇷 فارسی

Full documentation available in:


Development

Setup

git clone https://github.com/kttc-ai/kttc.git
cd kttc
python3.11 -m venv venv
source venv/bin/activate
pip install -e ".[dev]"
pre-commit install

Quality Standards

  • Type Checking: mypy --strict
  • Formatting: black (line length: 100)
  • Linting: ruff
  • Testing: pytest with asyncio support
# Run all checks
pre-commit run --all-files
pytest --cov=kttc

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Quick start:

  1. Fork the repository
  2. Create a feature branch
  3. Make changes and add tests
  4. Run quality checks: pre-commit run --all-files && pytest
  5. Submit a pull request

Please read our Code of Conduct before contributing.


Security

For security vulnerabilities, see SECURITY.md. Do not open public issues for security concerns.


License

Licensed under the Apache License 2.0. See LICENSE for details.

Copyright 2025 KTTC AI (https://github.com/kttc-ai)


Citation

If you use KTTC in your research:

@software{kttc2025,
  title = {KTTC: Knowledge Translation Transmutation Core},
  author = {KTTC AI},
  year = {2025},
  url = {https://github.com/kttc-ai/kttc},
  version = {0.1.0}
}

Links

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

kttc-0.4.2.tar.gz (284.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

kttc-0.4.2-py3-none-any.whl (381.0 kB view details)

Uploaded Python 3

File details

Details for the file kttc-0.4.2.tar.gz.

File metadata

  • Download URL: kttc-0.4.2.tar.gz
  • Upload date:
  • Size: 284.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for kttc-0.4.2.tar.gz
Algorithm Hash digest
SHA256 3f133e6d2fca15c735d3106bdd1975f9af79d63b0b4b063a4d19af09c0829f2a
MD5 57e2fd07a574de1bae7e06105223e0d0
BLAKE2b-256 e742a356ee667afb8da11b6c907bed44b1c4a7365690b599b1314a2e718dcccb

See more details on using hashes here.

Provenance

The following attestation bundles were made for kttc-0.4.2.tar.gz:

Publisher: publish.yml on kttc-ai/kttc

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file kttc-0.4.2-py3-none-any.whl.

File metadata

  • Download URL: kttc-0.4.2-py3-none-any.whl
  • Upload date:
  • Size: 381.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for kttc-0.4.2-py3-none-any.whl
Algorithm Hash digest
SHA256 adaa0e3ff45bf551568a7b6b1c4cb3d2d59c75e0a02eccc53bf08852c8f5a505
MD5 1b382ee9e40aa33181adcb477da9c8df
BLAKE2b-256 d53697c13011dbba6888e3e662c1cb30358535a0ef22dbcc437de521d19188e7

See more details on using hashes here.

Provenance

The following attestation bundles were made for kttc-0.4.2-py3-none-any.whl:

Publisher: publish.yml on kttc-ai/kttc

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page