ReplyWink documentation
Connect your AI provider to ReplyWink
ReplyWink uses your own AI API key. Add a profile in Settings → Model, choose the correct API format, enter an available model ID, and select Save and connect. One active provider handles AI scoring, drafting, translation, and AI-assisted setup.
On this page
Use an API credential, not a license key
Your provider’s API key authenticates AI requests. Your ReplyWink license key unlocks the paid extension allowance and belongs in License, not Model. Access to a provider’s chat website is not enough to configure an API profile: obtain an API credential and check that API usage is enabled for that account.
Choose a provider whose cost, availability, and data handling suit the content you plan to process. ReplyWink does not supply AI credits or add a markup to provider usage.
Choose the right provider and API format
| Provider | Selection | Address and model |
|---|---|---|
| OpenAI | OpenAI preset | The preset uses https://api.openai.com/v1. Select a model your account can use with Chat Completions. |
| DeepSeek | DeepSeek preset or Custom | The bundled preset uses https://api.deepseek.com/v1. If it no longer matches your provider’s guidance, use Custom with the current documented base address and model. |
| Anthropic | Anthropic preset | The preset uses https://api.anthropic.com/v1 and the Anthropic Messages format. |
| Another provider or gateway | Custom provider | Choose OpenAI-compatible or Anthropic to match the endpoint. Enter the gateway’s base address and exact model ID. |
Preset model names are editable starting points, not promises of current availability. Providers can rename, retire, or restrict models. Check their current model list rather than assuming the prefilled name will always work.
OpenAI-compatible profiles use Chat Completions; an endpoint that only exposes the Responses API is not interchangeable. For Anthropic profiles, use a Messages-compatible endpoint. The connection code adds the endpoint path, so do not paste the full /chat/completions or /messages URL into the base-address field.
For current provider details, see OpenAI’s Chat API reference, DeepSeek’s connection guide, and Anthropic’s API overview. These describe provider APIs; they are not endorsements or guarantees that every model works with ReplyWink.
Add, test, and activate a profile
- Open Model in the extension settings and add a provider.
- Choose the preset. For Custom, set a recognizable name, the API format, and a valid provider address.
- Paste your API key and enter the exact model ID. Keep the key hidden when sharing screenshots.
- Select Test connection to check the details. This performs a real provider request and does not save the profile.
- Select Save and connect to validate, save, and activate it. Look for the success message and Active badge.
You can keep multiple saved profiles, but only one is active. Select another profile to use it for subsequent AI work. Profiles do not assign scoring to one model and drafts to another.
Connect a custom or local endpoint
Use Custom when a provider needs a different base address or when you operate a compatible gateway. Copy the address from that service’s own configuration instructions. Prefer HTTPS for remote providers.
Local gateways can use an address on http://localhost or http://127.0.0.1, including the gateway’s port and any required base path. Keep the local service running while using the extension. The form requires a nonempty API-key field even for a local service; use the authentication value or local-only placeholder documented by that gateway. Never send an unrelated provider’s real key to a gateway merely to fill the field.
A local URL does not by itself guarantee that data stays on your computer. Check whether the gateway runs inference locally or forwards requests to another service.
Troubleshoot a connection or generation error
- “Enter the provider address, API key and model.”
- One of the required fields is empty. Complete all three; a custom name is optional.
- The provider rejects the key or reports an authentication error
- Check that the key belongs to the selected provider and account, has not been revoked, and is copied without surrounding whitespace. Verify provider-side access rather than trying your ReplyWink license key.
- Model not found or not permitted
- Use an exact model ID supported by this API format and enabled for your account. A display name from the provider’s chat app may not be a valid API ID.
- 404 or endpoint errors
- Check the base address and API format. Remove a duplicated endpoint suffix. For DeepSeek, compare the installed preset with the latest official guide and use Custom if necessary.
- Quota, billing, or rate-limit errors
- Check provider balance, account limits, and usage. A ReplyWink license cannot resolve a provider-side limit. Wait as directed by the provider before trying again.
- Network failure or timeout
- Confirm that the service is reachable and a local gateway is running. Check any proxy or gateway configuration. Avoid repeatedly regenerating while the underlying connection is failing.
- Test succeeds but scoring does not run
- Select Save and connect, confirm the profile is active, finish ReplyWink Setup, and refresh the social page. Posts can also be excluded by scoring thresholds.
Understand costs and data flow
Connection checks are real AI requests and may be billed. AI scoring can run as you browse after setup; “unlimited scoring” describes the extension’s allowance, not free AI inference. Local filtering reduces which posts reach the provider.
API profiles are stored in your browser, and the key authenticates requests to the endpoint you configure. Post content and relevant account context go to that endpoint for AI work. Review the Privacy Policy and your provider’s policies. For assistance, send support the provider name, model ID, extension version, and redacted error text—not your API key.