multi-agents-custom 2.0.0 → 2.1.0

package/README.md

@@ -53,24 +53,110 @@ SKIP_MULTI_AGENTS_POSTINSTALL=1 npm install multi-agents-custom
 
 ## Manual / CLI usage
 
+The CLI now uses an **interactive wizard** instead of command-line flags. Simply run:
+
 ```bash
-# Generate for all tools (default)
 npx multi-agents-custom
+```
+
+The wizard will guide you through:
+
+1. **Select AI tools** — Choose which tools to configure (Cursor, Copilot, Qwen, AntiGravity)
+2. **Project root** — Confirm or change the target project directory
+3. **Overwrite preference** — Decide whether to overwrite existing config files
+4. **Master Agent** — Optionally run the Master Agent to orchestrate the pipeline
+
+### Non-interactive / CI mode
+
+In CI environments (or when stdin is not a TTY), the CLI automatically uses defaults:
+- Targets: all tools
+- Project root: current directory
+- Overwrite: false
+
+---
+
+## Master Agent usage
 
-# Only generate for Cursor and Copilot
-npx multi-agents-custom --targets=cursor,copilot
+The Master Agent orchestrates the five-role pipeline (PM → BA → Tech Lead → Developer → Tester) to build features end-to-end.
 
-# Overwrite existing files
-npx multi-agents-custom --overwrite
+### Via the CLI wizard
 
-# Target a specific project root
-npx multi-agents-custom --root=/path/to/project
+When running `npx multi-agents-custom`, select "Yes" at the Master Agent step and enter your feature description.
 
-# Suppress output
-npx multi-agents-custom --quiet
+The Master Agent will:
+1. **Analyze your request** and determine which agents are needed
+2. **Execute agents in sequence** (PM → BA → Tech Lead → Developer → Tester)
+3. **Pass context** between agents so each builds on previous outputs
+4. **Generate structured outputs** for each phase
+
+### Programmatically
+
+```ts
+import { MasterAgent, AgentExecutionResult } from 'multi-agents-custom';
+
+const master = MasterAgent.getInstance();
+const prompt = 'Build a user authentication system with JWT tokens';
+
+for await (const event of master.run(prompt)) {
+  const icon = event.type === 'success' ? '✅' : event.type === 'error' ? '❌' : event.type === 'warning' ? '⚠️' : 'ℹ️';
+  console.log(`${icon} [${event.agentId}] ${event.message}`);
+}
+
+// Get detailed results after completion
+const results = master.getExecutionResults();
+for (const [role, result] of results.entries()) {
+  if (result.success) {
+    console.log(`\n## ${role.toUpperCase()} Output:\n${result.output}`);
+  }
+}
+```
+
+### Smart workflow routing
+
+The Master Agent intelligently selects which agents to run based on your input:
+
+| Input keywords | Agents executed |
+|---|---|
+| `test`, `qa`, `validation` | Tester only |
+| `requirement`, `idea`, `feature` | PM → BA |
+| `implement`, `code`, `build` (default) | Full pipeline: PM → BA → Tech Lead → Developer → Tester |
+
+### Master Agent events
+
+The Master Agent yields events as it orchestrates the pipeline:
+
+| Event type | Description |
+|---|---|
+| `info` | Status updates, progress messages |
+| `success` | Agent completed successfully |
+| `warning` | Non-critical issues |
+| `error` | Agent encountered an error |
+
+Each event includes:
+- `type` — Event type (`info`, `success`, `warning`, `error`)
+- `agentId` — Which agent produced the event (e.g., `pm`, `ba`, `techlead`)
+- `message` — Human-readable message
+
+### Agent outputs
+
+Each agent generates structured markdown output:
+
+- **PM Agent**: Product requirements, problem statements, success criteria
+- **BA Agent**: User stories with acceptance criteria, business rules
+- **Tech Lead Agent**: Architecture diagrams (mermaid), component design, tech stack
+- **Developer Agent**: Implementation plan, task breakdown, code structure
+- **Tester Agent**: Test strategy, test cases, quality gates
+
+### Advanced usage
+
+```ts
+// Get current context
+const context = master.getContext();
+console.log('Session:', context.sessionId);
+console.log('Status:', context.status);
 
-# Help
-npx multi-agents-custom --help
+// Reset for new session
+master.reset();
 ```
 
 ---