How-To Guides

This section has task-focused procedures for changing backends, wiring fields, tuning segmentation, and adjusting FAITH.

For copy-paste prompts and habits when you work with a coding agent, read Tips for Translation With Agents first. Start with Getting Started With Translation if you have not run the step yet.

Focused procedures for nemotron steps run translate/nemo_curator.

Run Translation

LLM backend

OpenAI-compatible servers, hosted or on-prem.

backend=llm

Run LLM Translation

NMT HTTP service

Self-hosted POST /translate microservices.

backend=nmt

Run NMT Translation

Google or AWS

Managed cloud translation APIs.

backend=google|aws

Run Google or AWS Translation

Configure and Tune

Fields and outputs

Wildcards, output_mode, chat reconstruction.

schema

Configure Fields and Output

Segmentation

Switch segmentation_mode deliberately.

segmentation

Use Fine Segmentation

FAITH Quality Gates

FAITH evaluation

Thresholds, filtering, model overrides.

faith

Run FAITH Evaluation

        graph LR
    A[Prepare YAML + env] --> B[nemotron steps run translate/nemo_curator]
    B --> C{Need FAITH?}
    C -->|yes| D[Tune faith_eval]
    C -->|no| E[Disable faith_eval]
    D --> B
    E --> B