dsh-superagent
An always-on coding agent open-sourced by DeepSeek that adopts a separated Router and Worker architecture. The Router receives tasks, determines intent, and dispatches work, while the Worker performs time-consuming actions such as code generation and file operations.
Timeline (last 90 days)
On the multi-worker coding agent path, key issues such as the router's task decomposition accuracy and worker conflict handling are not yet clear in dsh-superagent's public information.
dsh-superagent converts requests from GitHub and Feishu into worker sessions that execute in parallel.
In DeepSeek's own simulated tests, dsh-superagent's p95 task completion time dropped from 29.5 minutes to 15.7 minutes.
dsh-superagent adopts a router-worker architecture: the router understands requests, decomposes tasks, and distributes them to multiple workers for parallel processing.
dsh-superagent is positioned not as IDE completion, but as connecting requests in collaboration tools into an agent pipeline.
DeepSeek has open-sourced dsh-superagent, a research prototype under the MIT license.
In dsh-superagent, the Router is responsible for receiving tasks, determining intent and dispatching, while the Worker is responsible for actually executing time-consuming actions such as code generation and file operations.
Separating the Router from the Worker means the resident process does not have to cram all capabilities into the same context; the Router can listen in a lightweight manner, and Workers can be launched on demand or reused.
The source material does not provide the specific module division, model selection or performance data of dsh-superagent.
The core architecture of dsh-superagent is a resident design that separates the Router from the Worker.
All Facts (10)
On the multi-worker coding agent path, key issues such as the router's task decomposition accuracy and worker conflict handling are not yet clear in dsh-superagent's public information.
50%Unverifieddsh-superagent converts requests from GitHub and Feishu into worker sessions that execute in parallel.
50%UnverifiedIn DeepSeek's own simulated tests, dsh-superagent's p95 task completion time dropped from 29.5 minutes to 15.7 minutes.
50%Unverifieddsh-superagent adopts a router-worker architecture: the router understands requests, decomposes tasks, and distributes them to multiple workers for parallel processing.
50%Unverifieddsh-superagent is positioned not as IDE completion, but as connecting requests in collaboration tools into an agent pipeline.
50%UnverifiedThe core architecture of dsh-superagent is a resident design that separates the Router from the Worker.
50%UnverifiedIn dsh-superagent, the Router is responsible for receiving tasks, determining intent and dispatching, while the Worker is responsible for actually executing time-consuming actions such as code generation and file operations.
50%UnverifiedSeparating the Router from the Worker means the resident process does not have to cram all capabilities into the same context; the Router can listen in a lightweight manner, and Workers can be launched on demand or reused.
50%UnverifiedThe source material does not provide the specific module division, model selection or performance data of dsh-superagent.
50%UnverifiedDeepSeek has open-sourced dsh-superagent, a research prototype under the MIT license.
50%