refactor: simplify ralph wizard from 9 agents to 2

Remove execution layer that never actually controlled execution:
- execution-bridge.ts (model param was ignored)
- group-scheduler.ts (task-tool mode never used)
- model-selector.ts (recommendations were display-only)
- context-manager.ts (never invoked)
- execution-limits.ts

Remove redundant agent prompts (overlapping outputs):
- requirements-analyst, architecture-planner, risk-analyst
- testing-specialist, verification (merged into planner.ts)
- execution-optimizer (output was ignored)
- final-review (scores were cosmetic)

Simplify plan-orchestrator.ts from 2400 LOC to 520 LOC:
- Before: 9 agents, 6 phases, ~40-60 minutes
- After: 2 agents (research + planner), ~18 minutes

Remove /api/execution/* endpoints and related server code.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
arkon
2026-01-30 12:15:09 +01:00
co-authored by Claude Opus 4.5
parent 0020ae5a3e
commit a92826b66d
18 changed files with 329 additions and 5468 deletions
+1 -7
View File
@@ -6,11 +6,5 @@
*/
export { RESEARCH_AGENT_PROMPT } from './research-agent.js';
export { REQUIREMENTS_ANALYST_PROMPT } from './requirements-analyst.js';
export { ARCHITECTURE_PLANNER_PROMPT } from './architecture-planner.js';
export { TESTING_SPECIALIST_PROMPT } from './testing-specialist.js';
export { RISK_ANALYST_PROMPT } from './risk-analyst.js';
export { PLANNER_PROMPT } from './planner.js';
export { CODE_REVIEWER_PROMPT } from './code-reviewer.js';
export { VERIFICATION_PROMPT } from './verification.js';
export { EXECUTION_OPTIMIZER_PROMPT } from './execution-optimizer.js';
export { FINAL_REVIEW_PROMPT } from './final-review.js';