0.6.0 training session: Oracle Bot, RL combat, Mind's Eye, multilingual pipeline

Major changes from this session:

Training:
- 0.6.0 training running: 9B on steel141 3090 Ti, 27B on rented H100 NVL
- 7,256 merged training examples (up from 3,183)
- New training data: failure modes (85), midloop messaging (27),
  prompt injection defense (29), personality (32), gold from quarantine
  bank (232), new tool examples (30), claude's own experience (10)
- All training data RCON-validated at 100% pass rate
- Bake-off: gemma3:27b 66%, qwen3.5:27b 61%, translategemma:27b 56%

Oracle Bot (Mind's Eye):
- Invisible spectator bot (mineflayer) streams world state via WebSocket
- HTML5 Canvas frontend at mind.mortdec.ai
- Real-time tool trace visualization with expandable entries
- Streaming model tokens during inference
- Gateway integration: fire-and-forget POST /trace on every tool call

Reinforcement Learning:
- Gymnasium environment wrapping mineflayer bot (minecraft_env.py)
- PPO training via Stable Baselines3 (10K param policy network)
- Behavioral cloning pretraining (97.5% accuracy on expert policy)
- Infinite training loop with auto-restart and checkpoint resume
- Bot learns combat, survival, navigation from raw experience

Bot Army:
- 8-soldier marching formation with autonomous combat
- Combat bots using mineflayer-pvp, pathfinder, armor-manager
- Multilingual prayer bots via translategemma:27b (18 languages)
- Frame-based AI architecture: LLM planner + reactive micro-scripts

Infrastructure:
- Fixed mattpc.sethpc.xyz billing gateway (API key + player list parser)
- Billing gateway now tracks all LAN traffic (LAN auto-auth)
- Gateway fallback for empty god-mode responses
- Updated mortdec.ai landing page

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Seth
2026-03-22 20:22:50 -04:00
parent baab24f8b1
commit 5b28002001
44 changed files with 20873 additions and 4352 deletions
+67
View File
@@ -0,0 +1,67 @@
# Oracle Bot — Mortdecai Mind's Eye
Live HTML5 viewport that renders what the Mortdecai AI model "sees" during Minecraft server interactions.
**Public URL:** `mind.mortdec.ai`
## Quick Start (local dev)
```bash
cd oracle-bot
npm install
node server.js
# Open http://localhost:3333
```
## Deploy to CT 644
```bash
# 1. Copy files to CT 644
ssh pve112 "pct exec 644 -- mkdir -p /opt/oracle-bot"
scp -r oracle-bot/* pve112:/tmp/oracle-bot/
ssh pve112 "pct push 644 /tmp/oracle-bot/ /opt/oracle-bot/ --recursive"
# 2. Install deps on CT 644
ssh pve112 "pct exec 644 -- bash -c 'cd /opt/oracle-bot && npm install --production'"
# 3. Install systemd service
ssh pve112 "pct exec 644 -- cp /opt/oracle-bot/oracle-bot.service /etc/systemd/system/"
ssh pve112 "pct exec 644 -- systemctl daemon-reload"
ssh pve112 "pct exec 644 -- systemctl enable --now oracle-bot"
# 4. Verify
ssh pve112 "pct exec 644 -- curl -s http://localhost:3333/health"
```
## Caddy Config (CT 600)
Add to `/etc/caddy/Caddyfile`:
```
mind.mortdec.ai {
@internal path /trace /command
respond @internal 404
reverse_proxy 192.168.0.244:3333
}
```
Then: `caddy reload --config /etc/caddy/Caddyfile`
## Endpoints
| Endpoint | Access | Description |
|----------|--------|-------------|
| `GET /` | Public | Frontend (index.html) |
| `WS /ws` | Public | WebSocket stream |
| `GET /health` | Public | Health check |
| `POST /trace` | Internal | Gateway tool traces |
| `POST /command` | Internal | Bot commands |
## Test with simulated traces
```bash
curl -X POST http://localhost:3333/trace \
-H "Content-Type: application/json" \
-d '{"tool":"rcon.execute","input":{"command":"give slingshooter08 diamond 16"},"ok":true,"step":0,"mode":"god","player":"slingshooter08","session_id":"test-001"}'
```