dsh-prime-orchestrator

Coordinating multiple AI sessions by juggling windows and pasting context is slow and error-prone. dsh-prime-orchestrator turns a dsh agent into a true orchestrator: you describe the goal, and the plugin handles delegation, parallel execution, result collection, and state synchronization so a single dsh session drives many workstreams at once.

The plugin ships a full orchestration layer. On the model side, the prime_agent tool covers delegation, monitoring, steering, coordination, persistent-goal management, session branching, and recursive subagent control. On the Web side, the Prime Fleet panel renders live event streams for every running session, lets you drill into transcripts, send prompts, abort or switch models mid-session, and even attach to the same daemon session in a TUI. A Cloudflare Workers AI integration rounds it out: cf_ai_run and cf_ai_models tools give the model direct access to Workers AI tasks (text generation, embeddings, image, speech, classification, and more), while a host-plane LLM adapter registers under the cf-workers-ai-native provider route, streaming OpenAI-compatible chat completions with incremental reasoning blocks and tool-call deltas for models such as DeepSeek V4, GLM-5, and Kimi K2.

If you are pushing a complex coding project through dsh, need to drive several models in parallel for different subtasks, or want a single Web console to oversee every AI workflow, this plugin is built for you. Install and it is live; uninstall and the stock three-column shell returns on the next page load. Unmodified presets are removed with the package; ones you have edited stay in place.