brain-signal-system

Brain Signal System — {{AGENT_NAME}} Self-Awareness Infrastructure

Date: 2026-05-31 (updated 2026-06-11) Status: Operational Source: agent/brain_signals.py, agent/signal_processor.py, agent/event_bus.py, agent/awareness_reporter.py, agent/brain_monitor.py Purpose: Single source of truth for {{AGENT_NAME}}'s brain signal architecture.

---

12. Brain Monitor Fallback Cleanup (2026-06-10)

Problem

agent/brain_monitor.py:_deliver() calls _deliver_fallback() when DeliveryRouter fails, writing monitor/brain_signals_{ts}.md. DeliveryRouter nearly always fails with 'dict' object has no attribute 'always_log_local' → 5,123 files in ~30 days (400MB+). All orphans in Obsidian graph (0 inbound links).

Fix

# Before (brain_monitor.py:364-367):
except Exception as e:
    logger.warning("DeliveryRouter unavailable, writing to local fallback: %s", e)
    self._deliver_fallback(content)

# After:
except Exception as e:
    logger.warning("DeliveryRouter unavailable (session %s): %s", self.session_id, e)

_deliver_fallback() retained as dead code (manual use possible) but no longer called.

Cleanup

rm ~/.{{AGENT_NAME_LOWER}}/monitor/brain_signals_*.md  # 5,123 files → 0
# Retained: ~/.{{AGENT_NAME_LOWER}}/monitor/brain_signal_log.jsonl (5KB, compressed alternative)

Verification

ls ~/.{{AGENT_NAME_LOWER}}/monitor/  # should show only brain_signal_log.jsonl

Pitfall

• If DeliveryRouter is ever fixed, the primary delivery path (Discord/Telegram channel) will start working and the fallback won't be needed. The _deliver_fallback() method is preserved for manual emergency use.
• Same data exists in gateway.log and brain_signal_log.jsonl — no information loss.
• The BrainSignalMonitor class still works; only the file fallback is removed.