Configuration
June 23, 2026 ยท View on GitHub
Co-op Translator requires one language model provider. Image translation additionally requires Azure AI Vision.
Configuration is read from environment variables. For local projects, place them in a .env file at the project root.
For Azure resource setup, see Azure AI Setup.
Local runtime setup
Use a virtual environment before running the CLI locally. Co-op Translator supports Python 3.10 through 3.12.
For normal CLI usage, install the published package inside a virtual environment:
=== "Windows"
```powershell
python -m venv .venv
.venv\Scripts\activate
pip install co-op-translator
translate --help
```
=== "macOS / Linux"
```bash
python -m venv .venv
source .venv/bin/activate
pip install co-op-translator
translate --help
```
For repository development, install dependencies from the project root instead:
poetry install
poetry run translate --help
After the CLI is available, configure one language model provider in .env.
Provider selection
The tool auto-detects providers in this order:
- Azure OpenAI
- OpenAI
If neither provider is configured, translate, evaluate, migrate-links, and run_translation fail during configuration checks. co-op-review and run_review are deterministic maintenance checks and do not require provider credentials.
Azure OpenAI
Use Azure OpenAI when your model is deployed in Azure AI Foundry or Azure OpenAI Service.
AZURE_OPENAI_API_KEY="..."
AZURE_OPENAI_ENDPOINT="https://<resource>.openai.azure.com/"
AZURE_OPENAI_MODEL_NAME="gpt-4o"
AZURE_OPENAI_CHAT_DEPLOYMENT_NAME="<deployment>"
AZURE_OPENAI_API_VERSION="2024-12-01-preview"
The connectivity check uses the endpoint, API key, API version, and deployment name before translation begins.
OpenAI
Use OpenAI when calling the OpenAI API directly.
OPENAI_API_KEY="..."
OPENAI_CHAT_MODEL_ID="gpt-4o"
OPENAI_ORG_ID="..." # optional
OPENAI_BASE_URL="..." # optional
OPENAI_CHAT_MODEL_ID is required because the translator needs an explicit chat model for API calls.
Azure AI Vision
Image translation requires Azure AI Vision so the tool can extract text from images before translating it.
AZURE_AI_SERVICE_API_KEY="..."
AZURE_AI_SERVICE_ENDPOINT="https://<resource>.cognitiveservices.azure.com/"
If image translation is selected with -img, images=True, or no content-type filter, the tool validates Vision configuration before translation starts.
Multiple credential sets
The configuration layer supports multiple credential sets by suffixing variables with the same index:
AZURE_OPENAI_API_KEY_1="..."
AZURE_OPENAI_ENDPOINT_1="https://<resource-1>.openai.azure.com/"
AZURE_OPENAI_MODEL_NAME_1="gpt-4o"
AZURE_OPENAI_CHAT_DEPLOYMENT_NAME_1="<deployment-1>"
AZURE_OPENAI_API_VERSION_1="2024-12-01-preview"
AZURE_OPENAI_API_KEY_2="..."
AZURE_OPENAI_ENDPOINT_2="https://<resource-2>.openai.azure.com/"
AZURE_OPENAI_MODEL_NAME_2="gpt-4o"
AZURE_OPENAI_CHAT_DEPLOYMENT_NAME_2="<deployment-2>"
AZURE_OPENAI_API_VERSION_2="2024-12-01-preview"
Each set must be complete. The health check selects a working set before translation proceeds.
Command requirements
| Command or API | LLM required | Vision required | Notes |
|---|---|---|---|
translate -md | Yes | No | Translates Markdown only. |
translate -nb | Yes | No | Translates notebooks only. |
translate -img | Yes | Yes | Translates images only. |
translate with no type flags | Yes | Yes | Default mode includes Markdown, notebooks, and images. |
evaluate | Yes | No | Uses LLM evaluation unless --fast is selected. |
migrate-links | Yes | No | Performs link migration, but still runs shared configuration checks. |
co-op-review | No | No | Runs deterministic translation structure, freshness, Markdown, notebook, and local link checks. |
run_translation(markdown=True) | Yes | No | Programmatic Markdown translation. |
run_translation(images=True) | Yes | Yes | Programmatic image translation. |
run_review(...) | No | No | Programmatic deterministic review. |
Output directories
Default text translation output:
translations/<language-code>/<source-relative-path>
Default translated image output:
translated_images/<language-code>/<source-relative-path>
The Python API can override these directories with translations_dir and image_dir.