What it does RedInk is a self-hosted generator that produces full Xiaohongshu (Little Red Book) posts from one sentence. It writes a multi-page outline, generates a styled cover image, then batches content pages with matching visuals. The stack is a Flask backend and Vue 3 frontend, wrapped in a Docker image for one-command deployment.

The interesting bit The project is built around the 🍌Nano Banana Pro🍌 image model and Gemini 3 for text, but the config files reveal it is really provider-agnostic glue: swap in OpenAI-compatible APIs or DALL-E 3 via YAML. The “high concurrency” toggle (up to 15 images at once) is a nice touch for impatient users, though the README repeatedly warns that GCP trial accounts will hit rate limits and should keep it off.

Key highlights

• Three-step pipeline: smart outline → cover generation → batch content pages
• Docker one-liner: docker run -d -p 12398:12398 histonemax/redink:latest
• Supports both Gemini and OpenAI-compatible providers for text and images
• Web UI or YAML config; API keys are masked in the interface
• History and output persistence via Docker volume mounts
• Regenerate individual pages without losing style consistency

Caveats

• The container ships with blank API key templates; you must bring your own keys
• Image generation is slow enough that the README literally asks you not to leave the page
• CC BY-NC-SA 4.0 license means commercial use requires a separate license (details truncated in sources)

Verdict Worth a spin if you or your marketing team posts to Xiaohongshu regularly and you already have API access. Skip it if you need a generic design tool — this is tightly scoped to one platform’s format.