1.0.3

• The install-time REKA_API_KEY prompt now hides the key as you type or paste it, and links to https://developer.reka.ai/keys. Supported since Hermes 0.21.6, the plugin's minimum version. No change to models or provider behaviour.

To update an installed copy:

hermes plugins update reka-provider
hermes model --refresh

1.0.2

• New model: deepseek41-flash-20261017 (DeepSeek V4.1 Flash 1017), with a 1,048,576-token context window, tool calling, reasoning and image input. It passed a live chat and a tool call through Hermes before release.

To update an installed copy:

hermes plugins update reka-provider
hermes model --refresh

1.0.1

• glm5.3-flash now declares its 1,048,576-token context window (was 262,144), matching Reka's /v1/models.
• Setup docs explain the reasoning-effort choice instead of recommending one setting.
• The plugin doesn't update itself. To update an installed copy:

hermes plugins update reka-provider
hermes model --refresh

1.0.0

First release: Reka as a Hermes model provider with five models (qwen3.8-27b, glm5.3-flash, deepseek4-flash, deepseek-v4-pro, qwen3.6-27b), their context windows and image input, and reasoning effort sent as Reka's reasoning_effort.