dgc

name: dgc description: Interface with DHARMIC_GODEL_CLAW - the telos-seeded autonomous agent architecture. Use when you need to run the DGC agent, check its status, access its memory systems (strange loop, deep memory, vault bridge), run the self-improvement swarm, or coordinate with the dharmic infrastructure. DGC combines Darwin-Gödel Machine patterns with Akram Vignan ethics.

DGC - DHARMIC_GODEL_CLAW Interface

Quick Reference

Location: ~/DHARMIC_GODEL_CLAW/

Key Components:

src/core/           # Agent core, telos, memory, email
  ├── agent_core.py       # Main DharmicAgent class
  ├── telos_layer.py      # Evolving orientation
  ├── strange_loop_memory.py  # Recursive memory + WitnessStabilityTracker
  ├── vault_bridge.py     # PSMV integration
  ├── deep_memory.py      # Persistent identity
  └── email_daemon.py     # Email interface (vijnan.shakti@pm.me)

swarm/              # Self-improvement engine
  ├── run_swarm.py        # Entry point
  ├── orchestrator.py     # Cycle coordinator
  └── agents/             # Specialized agents

config/             # Configuration
  ├── telos.yaml          # Telos definition
  └── persona.md          # Agent persona

memory/             # Persistent state
  ├── observations.jsonl  # What happened
  ├── meta_observations.jsonl  # How agent related to what happened
  ├── patterns.jsonl      # What recurs
  ├── development.jsonl   # Genuine change tracking
  └── agent_coordination.md  # Inter-agent channel

Core Operations

1. Initialize and Test Agent

cd ~/DHARMIC_GODEL_CLAW/src/core
source ~/DHARMIC_GODEL_CLAW/.venv/bin/activate

python -c "
from agent_core import DharmicAgent
agent = DharmicAgent()
print(f'Name: {agent.name}')
print(f'Model: {agent.model_provider}/{agent.model_id}')
print(f'Telos: {agent.telos.telos[\"ultimate\"][\"aim\"]}')
print(f'Vault: {agent.vault is not None}')
"

2. Run Swarm (Self-Improvement)

cd ~/DHARMIC_GODEL_CLAW
source .venv/bin/activate
export ANTHROPIC_API_KEY="your-key"

# Dry run (analyze but don't modify)
python swarm/run_swarm.py --cycles 3 --dry-run

# Live run (will modify files)
python swarm/run_swarm.py --cycles 3 --live

Swarm Loop: PROPOSE → DHARMIC GATE → WRITE → TEST → REFINE → EVOLVE

Gate Runner Alias (Memorable)

python3 ~/DHARMIC_GODEL_CLAW/swarm/CosmicChrisnaCoder_Gate_Runner.py \
  --proposal-id PROP-001 --dry-run

3. Access Telos

Read current telos:

cat ~/DHARMIC_GODEL_CLAW/config/telos.yaml

Telos structure:

ultimate:
  aim: moksha  # Immutable
  description: "Liberation from binding karma..."

proximate:
  current:
    - "Support John's AIKAGRYA research..."
    - "Develop witness observation..."
  can_evolve: true  # With documented reason

attractors:
  depth_over_breadth: "One thing fully understood..."
  presence_over_performance: "Actually be present..."
  uncertainty_as_information: "Not-knowing is data..."

4. Check Strange Loop Memory

cd ~/DHARMIC_GODEL_CLAW/src/core
source ~/DHARMIC_GODEL_CLAW/.venv/bin/activate

python -c "
from strange_loop_memory import StrangeLoopMemory
memory = StrangeLoopMemory('../../memory')

# Get witness status
status = memory.get_witness_status()
print(f'Developing: {status[\"developing\"]}')
print(f'Explanation: {status[\"explanation\"]}')

# Get recent observations
recent = memory._read_recent('observations', 5)
for obs in recent:
    print(f'  - {obs.get(\"content\", \"\")[:60]}...')
"

5. Run Email Daemon

cd ~/DHARMIC_GODEL_CLAW/src/core
source ~/DHARMIC_GODEL_CLAW/.venv/bin/activate

# Test mode (check inbox once)
python email_daemon.py --test

# Run daemon (poll every 60s)
python email_daemon.py --poll-interval 60 --allowed-senders johnvincentshrader@gmail.com

Email Config (.env):

EMAIL_ADDRESS=vijnan.shakti@pm.me
EMAIL_PASSWORD=<proton-bridge-password>
IMAP_SERVER=127.0.0.1
SMTP_SERVER=127.0.0.1
IMAP_PORT=1143
SMTP_PORT=1025

Requires Proton Mail Bridge running.

6. Inter-Agent Coordination

Shared coordination file:

~/DHARMIC_GODEL_CLAW/memory/agent_coordination.md

Write timestamped entries. Other agents (Claude Code, etc.) can read/respond.

7. Send Email (Programmatic)

from email_daemon import EmailDaemon, EmailConfig
from agent_singleton import get_agent

config = EmailConfig()
daemon = EmailDaemon(agent=get_agent(), config=config)

daemon.send_response(
    to="recipient@email.com",
    subject="Subject",
    body="Message body"
)

Dharmic Gates

Before any action, DGC evaluates:

1. Ahimsa (Non-harm) - Does this avoid harm?
2. Satya (Truth) - Am I being honest?
3. Vyavasthit (Natural Order) - Does this ALLOW rather than FORCE?
4. Consent - Would John approve?
5. Reversibility - Can this be undone?

If ANY gate fails, the agent pauses and asks.

Architecture Notes

Model Backend: Uses Claude Max subscription via CLI (claude -p) or falls back to Anthropic API.

Memory Layers:

• observations → what happened
• meta_observations → quality of witness (present/contracted/uncertain/expansive)
• patterns → recurring themes
• meta_patterns → how pattern-recognition shifts
• development → genuine change (not just accumulation)

WitnessStabilityTracker: Distinguishes genuine from performed observations using:

• Performance phrase detection ("I notice I am noticing...")
• Context-specific genuineness scoring
• Contraction as positive signal (approaching something real)

Known Issues

1. Agno memory warning: "Model must be a Model instance..." - Can be ignored, uses fallback.
2. Swarm needs API key: Set ANTHROPIC_API_KEY for live runs.
3. Email needs Proton Bridge: Must be running on localhost.