Organize by tags, record text and images in cards, and feed materials to the model with @ in the chat.
This is a memory plugin for DeepSeek Harness, maintained by hello-heyongping. It provides a tag- and card-based material management solution, supporting text-and-image records, content search, and quick references for conversation context.
Core Features¶
- Tag categorization: Supports renaming, merging (renaming to an existing tag automatically merges them), and deleting tags. Deleting a tag only removes the tag itself; card content is preserved.
- Text and image cards: Card content supports block-based arrangement (text blocks / image blocks), and supports searching by title, body, image caption, source, and tag name.
- Conversation references: Type
@in the chat input box, and the “Card Materials” option will appear in the menu. Supports referencing a single card or an entire “tag group”. - Image storage mechanism: Images are content-addressed and stored using SHA-1. Transparency preservation is supported; GIFs and SVGs are not re-encoded.
- Material library structure: Managed as folders; one folder is one material library. Supports cloud drive synchronization and Git version control.
- Data protection: Includes automatic saving (800 ms debounce) and an optimistic locking mechanism. A document-level snapshot is generated every 10 minutes, and the 30 most recent snapshots are retained.
Installation¶
Add the plugin using the official installation command:
dsh plugin --profile desktop add github:hello-heyongping/dsh-helloai-works
After installation, refresh the page. A “Card Materials” panel will appear directly below “Plugins” in the left sidebar.
Typical Usage¶
- Open the panel directly below “Plugins” in the left sidebar.
- Create a new tag in the bottom input box (for example, “Character Setup”), and press Enter.
- Click “New” in the top-right corner to create a card, then add a title, text block, or image block.
- Type
@in the chat, and select “Card Materials” or “tag group” from the menu to reference it. - Data files are located in the
<DSH_HOME>/helloai-works/directory.
Notes¶
- Data persistence: After uninstalling DeepSeek Harness, data remains in the
<DSH_HOME>/helloai-works/folder. This folder belongs to the user directory and is separate from the program installation directory. - Trigger dependency: The
@reference feature requires support from the DeepSeek Harness trigger service. If the current version does not include this service, the feature will not work. - Image limits: A single image is limited to 32 MB after base64 encoding.
- Cleanup recommendation: It is not recommended to use “Clean Up Unused Images” as a routine operation, as it may cause old snapshots to lose images.
- Code changes: Changes to Host code require a full restart of DeepSeek Harness to take effect.
