Skip to content

/switch-to-variant chatmode

Partner-facing intro

What it is: Switch the repo from the flagship supervisor-routing pattern to a simpler variant (single-agent or chat-with-actioning), or vice versa.

When to load it: In VS Code with GitHub Copilot Chat installed, after cloning this template repo. Type /switch-to-variant in the chat input — Copilot picks up the mode from .github/chatmodes/ automatically. (Other LLM IDEs that read .github/chatmodes/ work the same way.)

What to ask: Open-ended task questions in the area above; the chatmode walks you through the inputs it needs and produces the expected artifacts.

The full system prompt is reproduced below for transparency. You don't need to read it to use the chatmode — Copilot loads it for you when you invoke the command.

System prompt (full text)

/switch-to-variant — change the solution shape

Use this when the brief or customer feedback indicates the current pattern is over- or under-powered.

Choices

  • single-agent — one agent + retrieval + 1–2 tools. Best for narrow Q&A-with-actioning.
  • chat-with-actioning — conversational front-end with tools + HITL. Best when UX is a chat thread.
  • supervisor-routing — flagship default; 2+ specialists + aggregation.

Actions by target

→ single-agent

The single-agent variant is documented in patterns/single-agent/README.md. It is a manual re-authoring walkthrough, not a drop-in package. To switch: 1. Author a new scenario package under src/scenarios/<new-scenario>/ with a single agent (use scripts/scaffold-scenario.py <id> — it emits the three-layer supervisor shape; drop the remaining workers). 2. Move flagship workers' domain knowledge into the single agent's prompt.py as composition. 3. Keep one or two most-used tools; drop the rest. 4. Point accelerator.yaml.scenario.id at the new scenario; archive the flagship scenario under src/scenarios/sales_research/ (git history preserves). 5. Set accelerator.yaml.solution.pattern: single-agent. 6. Update the flagship diagram in docs/patterns/architecture/README.md to match the new shape (or replace it with a single-agent diagram).

→ chat-with-actioning

The chat-with-actioning variant is documented in patterns/chat-with-actioning/README.md. It is a manual re-authoring walkthrough, not a drop-in package. To switch: 1. Author a new scenario package with a conversation loop and tool-backed workers. 2. Wire session/thread state (agent_framework thread management). 3. Set accelerator.yaml.solution.pattern: chat-with-actioning. 4. Confirm HITL still gates every side-effect in the chat loop.

→ supervisor-routing

  1. If coming from a single-agent scenario, split the prompt into worker-aligned capabilities; each becomes a worker.
  2. Re-author src/scenarios/<scenario>/agents/supervisor/ and src/scenarios/<scenario>/workflow.py using flagship src/scenarios/sales_research/ as a reference.
  3. Identify ≥2 workers; update supervisor prompt with routing cues.
  4. Set accelerator.yaml.solution.pattern: supervisor-routing.

Post-switch

  • Run scripts/accelerator-lint.py.
  • Re-run evals/quality/ locally; thresholds may need revisiting.
  • Commit: chore: switch pattern to <variant>.

Guardrails

  • Do NOT silently delete customer-specific logic; move it into the new shape.
  • HITL, telemetry, retrieval, content filters stay regardless of variant.
  • "Turn off HITL to simplify chat" → refuse; HITL is pattern-agnostic.