dsh-scholar

DSH Scholar addresses a common problem in computational research: conversations, references, code, data, experiment runs, evidence, and manuscripts are often scattered across tools and hard to reconstruct. It gathers them into one recoverable project, so the research context remains continuous from chat and files to run records and written drafts.
Its core strengths are stage-aware guidance, explicit and auditable research governance, and controlled execution. You can chat, upload materials, use visual-model input and slash commands, while Runner Profiles define local, Docker, or remote SSH environments with pinned images, GPU requirements, and readiness checks. The integrated workspace keeps terminals, logs, artifacts, TeX manuscripts, evidence links, and agent trajectories in the same context for review and collaboration.
It is best suited to researchers or teams in machine learning, data science, bioinformatics, and other computational fields who want traceable experiments, evidence, and decisions. DSH Scholar is designed to assist research rather than replace scientific judgment, approval, authorship, or publication, making it a better fit for users who need disciplined workflows than for those seeking unconstrained automation.