dataforge/agents_and_robots
1
0
Fork 0
Code Issues Pull Requests 2 Actions Packages Projects Releases Wiki Activity

feat: enhance orchestrator prompts for multi-agent collaboration

Browse Source 
Se reescriben los tres prompts del orquestador (routing, quality, refinement)
para fomentar la interacción entre agentes en lugar de respuestas individuales.

- Routing: instruye al LLM a mantener confianza baja (0.3-0.6) en preguntas
  generales para disparar contribuciones de múltiples agentes.
- Quality: nuevo sistema de scoring que favorece la colaboración. Un solo agente
  respondiendo nunca supera 0.5, forzando que otros participen.
- Refinement: cambia el enfoque de "mejorar" a "enriquecer con perspectiva única",
  priorizando agentes con expertise diferente al anterior.
- Config: max_iterations 3→6, quality_threshold 0.8→0.85, nuevo campo
  repetition_threshold: 0.6 para detección de bucles.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Enmanuel
2026-03-06 17:10:52 +00:00
parent 7176afde0a
commit 485d6e86be
4 changed files with 32 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
agents/specials/orchestrator/config.yaml
+3 -2
View File
@@ -13,7 +13,8 @@ llm:
temperature: 0.2
orchestration:
max_iterations: 3
quality_threshold: 0.8
max_iterations: 6
quality_threshold: 0.85
delegation_timeout: 30s
repetition_threshold: 0.6 # similarity ratio (0-1) to detect circular conversations
rooms: [] # auto-detected: any room with ≥2 registered bots is managed automatically