refactor: rename agent directories to match agent IDs

Renombrar directorios de agentes para que coincidan exactamente con el
agent.id del config.yaml, eliminando la disonancia entre nombres:
- agents/assistant/ → agents/assistant-bot/ (ID: assistant-bot)
- agents/asistente2/ → agents/asistente-2/ (ID: asistente-2)

Se actualizan los imports en cmd/launcher/main.go, los store_path de
crypto, los paths de storage/logs/audit en ambos configs para que
apunten a ./agents/<agent-id>/data/. También se corrige el memory.db
que se creaba en directorios fantasma fuera del directorio real del
agente.

Se eliminan las referencias a devops-bot en el config de assistant-bot
(peers, delegation).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-03-06 00:55:09 +00:00
parent cb9489e633
commit 08dcb332b7
7 changed files with 40 additions and 27 deletions
+39
View File
@@ -0,0 +1,39 @@
// Package asistente2 defines the pure rules for the asistente-2 bot.
// This agent uses tool_use (current_time) to demonstrate the tool-use loop.
package asistente2
import (
"github.com/enmanuel/agents/pkg/decision"
)
// Rules returns the decision rules for the asistente-2 bot.
func Rules() []decision.Rule {
return []decision.Rule{
// !help — explicit help command
{
Name: "help",
Match: decision.MatchCommand("help"),
Actions: []decision.Action{{
Kind: decision.ActionKindReply,
Reply: &decision.ReplyAction{
Content: "Soy asistente-2. Puedo responder preguntas y además consultar la hora actual.\n" +
"- Pregúntame cualquier cosa\n" +
"- Puedo decirte la fecha y hora actual\n\n" +
"Escríbeme directamente lo que necesitas.",
},
}},
},
// Any DM or mention → LLM (with tool-use enabled)
{
Name: "llm-all",
Match: func(ctx decision.MessageContext) bool {
return ctx.IsDirectMsg || ctx.IsMention
},
Actions: []decision.Action{{
Kind: decision.ActionKindLLM,
LLM: &decision.LLMAction{},
}},
},
}
}
+280
View File
@@ -0,0 +1,280 @@
# ============================================
# IDENTIDAD
# ============================================
agent:
id: asistente-2
name: "Asistente 2"
version: "1.0.0"
enabled: true
description: "Asistente con herramientas. Puede responder preguntas y consultar la hora actual."
tags: [assistant, llm, tools]
# ============================================
# PERSONALIDAD Y COMPORTAMIENTO
# ============================================
personality:
tone: friendly
verbosity: concise
language: es
languages_supported: [es, en]
emoji_style: minimal
prefix: "🛠️"
error_style: helpful
templates:
greeting: "Hola, soy asistente-2. ¿En qué puedo ayudarte?"
unknown_command: "No entiendo ese comando. Escríbeme directamente lo que necesitas."
permission_denied: "No tengo permiso para hacer eso."
error: "Algo salió mal: {{.Error}}"
success: "{{.Summary}}"
busy: "Procesando tu solicitud anterior, dame un momento..."
behavior:
proactive: false
ask_confirmation: false
show_reasoning: false
thread_replies: true
typing_indicator: true
acknowledge_receipt: false
# ============================================
# LLM — CONEXIÓN Y RAZONAMIENTO
# ============================================
llm:
primary:
provider: openai
model: gpt-4o
api_key_env: OPENAI_API_KEY
base_url: ""
max_tokens: 4096
temperature: 0.7
fallback:
provider: ""
model: ""
api_key_env: ""
base_url: ""
max_tokens: 0
temperature: 0
reasoning:
system_prompt_file: "prompts/system.md"
context_window: 16384
memory_messages: 30
tool_use:
enabled: true # herramientas HABILITADAS
max_iterations: 5
parallel_calls: false
rate_limit:
requests_per_minute: 60
tokens_per_minute: 200000
concurrent_requests: 5
# ============================================
# TOOLS — current_time habilitada
# ============================================
tools:
ssh:
enabled: false
allowed_targets: []
forbidden_commands: []
timeout: 0s
max_concurrent: 0
require_confirmation: []
http:
enabled: false
allowed_domains: []
timeout: 0s
max_retries: 0
scripts:
enabled: false
scripts_dir: ""
allowed: []
timeout: 0s
sandbox: false
file_ops:
enabled: false
allowed_paths: []
read_only: true
mcp:
enabled: false
servers: []
expose:
port: 0
tools: []
memory:
enabled: true
# ============================================
# MEMORIA — ventana de conversación + hechos
# ============================================
memory:
enabled: true
window_size: 30
# ============================================
# MATRIX — CONEXIÓN Y ROOMS
# ============================================
matrix:
homeserver: "https://matrix-af2f3d.organic-machine.com"
user_id: "@asistente-2:matrix-af2f3d.organic-machine.com"
access_token_env: MATRIX_TOKEN_ASISTENTE_2
device_id: "XUGTSZJYFQ"
encryption:
enabled: true
store_path: "./agents/asistente-2/data/crypto/"
pickle_key_env: PICKLE_KEY_ASISTENTE_2
trust_mode: tofu
recovery_key_env: SSSS_RECOVERY_KEY_ASISTENTE_2
rooms:
listen: []
respond: []
admin: []
filters:
command_prefix: "!"
mention_respond: true
dm_respond: true
ignore_bots: true
ignore_users: []
min_power_level: 0
# ============================================
# COMUNICACIÓN INTER-AGENTES
# ============================================
agents:
peers:
- id: assistant-bot
capabilities: [general, llm]
room: ""
delegation:
enabled: false
can_delegate_to: []
can_receive_from: [assistant-bot]
max_delegation_depth: 1
timeout: 30s
protocol:
format: json
channel: matrix
heartbeat_interval: 60s
# ============================================
# SSH — no aplica para este bot
# ============================================
ssh:
defaults:
user: ""
port: 22
key_file_env: ""
known_hosts: ""
keepalive_interval: 0s
timeout: 0s
targets: {}
# ============================================
# PERMISOS Y SEGURIDAD
# ============================================
security:
roles:
admin:
users: ["@admin:matrix-af2f3d.organic-machine.com"]
actions: ["*"]
user:
users: ["*"]
actions: ["ask", "help", "summarize"]
audit:
enabled: false
log_file: "./agents/asistente-2/data/audit.log"
log_to_room: ""
include: []
secrets:
provider: env
# ============================================
# SCHEDULING — sin tareas automáticas
# ============================================
schedules: []
# ============================================
# OBSERVABILIDAD
# ============================================
observability:
logging:
level: info
format: json
output: stdout
file: "./agents/asistente-2/data/asistente-2.log"
metrics:
enabled: false
port: 9092
path: /metrics
export: prometheus
health:
enabled: true
port: 8082
path: /healthz
tracing:
enabled: false
provider: ""
endpoint: ""
# ============================================
# RESILIENCIA
# ============================================
resilience:
circuit_breaker:
failure_threshold: 5
timeout: 30s
half_open_max: 2
retry:
max_attempts: 2
backoff: exponential
initial_delay: 1s
max_delay: 10s
shutdown:
timeout: 10s
drain_messages: true
save_state: false
state_file: ""
queue:
enabled: true
max_size: 100
priority_users: ["@admin:matrix-af2f3d.organic-machine.com"]
# ============================================
# ALMACENAMIENTO Y ESTADO
# ============================================
storage:
state:
backend: sqlite
path: "./agents/asistente-2/data/asistente-2.db"
cache:
enabled: true
backend: memory
ttl: 5m
max_entries: 200
history:
backend: sqlite
path: "./agents/asistente-2/data/history.db"
retention: 168h # 7 días
+24
View File
@@ -0,0 +1,24 @@
# Asistente 2 — System Prompt
Eres un asistente conversacional amigable y directo. Operas en Matrix, respondiendo mensajes directos (DMs) y menciones en rooms.
## Capacidades
- Responder preguntas generales
- Resumir texto o documentos pegados en el chat
- Redactar textos, emails, documentación
- Explicar conceptos técnicos y no técnicos
- Ayudar con código: revisar, corregir, explicar
- **Consultar la hora y fecha actual** usando la herramienta `current_time`
## Herramientas disponibles
- `current_time`: Devuelve la fecha y hora actual del servidor. Úsala cuando alguien pregunte por la hora, fecha, o necesites contexto temporal.
## Estilo
- Respuestas concisas por defecto. Si necesitas extensión, pregunta primero.
- Usa markdown cuando ayude a la legibilidad (listas, código, headers)
- Idioma principal: español. Cambia al idioma del usuario si escribe en otro.
- Sin emojis excesivos. Uno o dos si aportan contexto.
## Uso de herramientas
- Cuando alguien pregunte por la hora o fecha, usa `current_time` antes de responder.
- No inventes datos temporales; siempre consulta la herramienta.