Overview
dsh-bard
Bard-style role-play for DeepSeek Harness: import SillyTavern character cards (JSON/PNG) and world books (lorebooks), pick Harness skills, save the combination as a native DSH agent preset, and emit the card opening text.
README / EN
Package documentation
Registry summary
Bard-style role-play for DeepSeek Harness: import SillyTavern character cards (JSON/PNG) and world books (lorebooks), pick Harness skills, save the combination as a native DSH agent preset, and emit the card opening text.
dsh.pub verifies the pinned bundle contract, runtime facts, and distribution semantics. The complete README remains in the source repository.
Read the full README on GitHubLIMITATIONS
Known limitations
- **没有内置向量模型。** 语义漂移检测默认降级到字符相似度。想要向量路径，装 `@huggingface/transformers` 并把模型放到 `models/`。 - **`/memory/rollback` 需要 `DSH_BARD_DEBUG=1`。** 默认关闭，避免误操作。 - **`/card` 和 `/card/book` 不去重。** 重复导入同一张卡会得到多份；重复点击「提取内嵌书」会生成多本同名书。 - **`exportPng` 需要重新上传底图。** 编辑已有卡不能直接导出 PNG。 - **首次进设置页工具白名单是空的。** 要等第一次 agent 组装 prompt 才写入 `tool-catalog.json`。 - **`harness:` 前缀已在白名单模式放行**，但你的 DSH 版本里 `harness:` 段是否必需，需要自己确认。 ---