Dsh Vision Bridge

Key features

• The same picture (by attachment Id) caches the description in the same process and repeats the non-repetition billing.
• Allows the pure text model (DeepSeek V4/V4-Flash) to "see" pictures in a dialogue: paste them directly and send them in an input box. The plugin automatically calls for an OpenAI compatible visual model (VLM) to recognize the picture as a text description, then feed it to DeepSeek to continue processing.
• The DSH-synthesis of a pure text model rejects the picture (MODEL DOES NOT SUPPOT IMAGES) at the time of submission.
• After restarting, you can simply paste the picture in the input box and send it.
• Sub-intellectual sessions (subagent) are still subject to DSH kernel restrictions and do not support stickers.