DeepSeek vision bridge for dsh
Einskyle/dsh-llm-vision-bridge · model-adapters
DeepSeek vision bridge for dsh: route image attachments to a vision model (Qwen3-VL via pi-ai/llama.cpp) and continue on a text-only LLM (DeepSeek).
★ 4GitHub stars
0Forks
2026-08-25Last updated
JavaScriptLanguage
MITLicense
Key features
- Native LLM provider — registers deepseek-vision on the DSH LlmAdapter seam.
- Image admission, request routing, and session compaction all run through harness-native mechanisms
- no UI changes, no front-end interception.
- Zero overhead without images — image-free requests pass straight through to the fallback provider (default deepseek-official).
- LRU description cache — the same image + prompt is never re-described
Requirements
- A vision model route on the pi-ai adapter, configured under Settings → Models → llm-pi-ai (e.g.
- If the vision provider declares apiKeyEnv but the credential is not set, pi-ai reports MISSING_CREDENTIAL: store any placeholder value on the Settings page (local llama.cpp does not validate the key), or remove that apiKeyEnv.
Install command
dsh plugin --profile web add github:Einskyle/dsh-llm-vision-bridge