Novelai Openclaw Adaptor

Verification-first workflow

If the user wants to enable runtime use, follow this order:

1. Check whether the adaptor is already present locally or already installed.
2. If it is not present, explain that the package source should be verified before any installation.
3. Ask for approval before running any install command.
4. After the user has verified the source and approved installation, use the package's documented install method.
5. Prefer interactive or local-only credential entry during configuration.

Safe examples:

pip install novelai-openclaw-adaptor

novelai-config init

Do not use examples like:

novelai-config init --api-key "YOUR_NOVELAI_API_KEY"

If the user needs help deciding whether the package is trustworthy, suggest reviewing its repository, release history, and maintainers before installation.

Supported models

Text models:

• glm-4-6
• erato
• kayra
• clio
• krake
• euterpe
• sigurd
• genji
• snek

Image models:

• nai-diffusion-4-5-full
• nai-diffusion-4-5-curated
• nai-diffusion-4-full
• nai-diffusion-4-curated
• nai-diffusion-3
• nai-diffusion-3-furry

When helping with setup, ask the user which model they want instead of assuming silently. If they do not care, recommend:

• Text: glm-4-6
• Image: nai-diffusion-4-5-full

OpenClaw configuration

When explaining how to connect OpenClaw to the adaptor:

1. Set base_url to the local adaptor endpoint, such as http://127.0.0.1:xxxx/v1 or http://localhost:xxxx/v1.
2. Set the OpenClaw model name to the adaptor-exposed model the user selected.
3. Clarify that credential handling belongs to the local adaptor configuration, not the chat.
4. If OpenClaw insists on an API key field, explain that some clients accept a placeholder value such as sk-local, but the real NovelAI credential should stay in the local adaptor config only.

Helpful local commands

If the user has already verified the package source and approved local usage, these commands are relevant:

novelai-config --help
novelai-shim --help
novelai-image --help

Use novelai-config init as the normal guided setup entry point. It should collect local configuration such as:

• UI language
• NovelAI credential through local input
• Default shim model
• Default image output directory
• Default image model

Image generation usage

Once the local adaptor is configured and running, image generation can use the normal OpenClaw or OpenAI-style prompt flow.

Example prompt:

1girl, solo, masterpiece, best quality, highly detailed

The adaptor is responsible for translating that prompt into the format expected by NovelAI.