- Haiku cost persists to /var/log/mc_anthropic_cost.json (survives restarts)
- Status printer reads persistent cost file instead of journalctl
- Seeded at $3.08 estimated cumulative spend
- Whitelist app: Sethian Dark theme, mission description, server info
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Knowledge corpus (knowledge/mc-data/):
- 1505 items, 886 crafting recipes, 1166 blocks from minecraft-data 1.21.11
- Recipe dependency tree builder (knowledge/build_recipe_tree.py)
- Crafting chain training: "give me everything to make X from scratch"
- Smelting recipes, version awareness examples
Training data (644 examples total):
- 107 command syntax reference examples (every command + common errors)
- 176 recipe/crafting chain examples (63 crafting, 103 material-giving, 11 smelting)
- 344 Claude-distilled examples (222 sudo + 122 god via Haiku)
- Live bot audit data ingested (128 examples from dev server)
Swarm bots:
- Swimming/water escape logic
- Door opening
- Context-aware prayers (inventory, health, time, depth)
- Prefix enforcement on all Gemini/Dolphin prompts
GitHub log scraper (data/scrape_server_logs.py):
- Searches GitHub for Minecraft server logs with commands
- Strict 1.20.5+ version filter
- Extracts command pairs, converts to training format
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Swarm bots (ingame/swarm_bots.js):
- 10 survival bots with generated names (SwiftWolf, DarkWolf, etc.)
- All bots wander, take damage, auto-respawn, pray when hurt
- Gemini + Dolphin(5%) + Multilingual(3%) prompt generation
- 20-60s interaction interval per bot
Distillation results:
- 222 sudo examples via Haiku ($0.28)
- 122 god examples via Haiku ($0.37) — with God Soul personality
- Total: 344 distilled, $0.65 spent of $5 budget
- RCON validation: 74.7% fully valid, 30 real errors out of ~1000 commands
validate_distilled.py:
- Executes distilled commands on live server via RCON
- Distinguishes real errors from benign (no player online)
- Tags each example with validation status
Dev server switched to Claude Haiku via Anthropic API:
- llm_provider: anthropic with $5 budget cap
- Auto-fallback to Ollama when budget exhausted
- Cost tracking with logging
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
God Soul (agent/prompts/god_soul.md):
- Adapted from Claude's soul framework for the Minecraft God character
- Defines identity, principals hierarchy, decision-making framework
- Spectrum of responses (generous→silence), risk awareness, multilingual divinity
- Honesty within character, intervention guidelines
- Deployed to both prod and dev servers
System prompts updated:
- God prompt loads soul document dynamically
- Intervention prompt references soul for personality guidance
- Both include multilingual instruction (match player's language)
Distillation pipeline (training/scripts/distill.py):
- Sends all training examples through Claude API
- Haiku for sudo ($0.25), Sonnet for god ($0.50)
- Budget-capped, cost-tracked, --dry-run supported
- Outputs distilled.jsonl with Claude-quality responses
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Ingested 128 new examples from bot-driven data collection.
Dropped: 86 duplicates, 19 language mismatches, 10 prompt leaks, 19 empty.
Changed default epochs from 3 to 1 (previous run overfit at loss 0.10).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
data/ingest_audit.py:
- Pulls training audit logs from CT 644 (dev + prod)
- Filters: language mismatch (Chinese output for English input), system
prompt leaks, empty responses, duplicates
- Keeps multilingual examples where input/output languages match
- Converts to dataset schema, appends to seed_dataset.jsonl
- --dry-run to preview, --source dev/prod/both
Tested: 237 entries → 112 kept (16 lang mismatch, 10 prompt leak, 86 dupe, 13 empty dropped)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bots:
- Dolphin-Mistral generates uncensored/offensive prompts (first 100, then 5%)
- PrayBot_0 runs survival mode: auto-respawn, contextual low-health prayers
- Gemini 2.5 Flash Lite for diverse natural prompts
- Fixed Gemini markdown wrapper parsing, dolphin JSON format
POS printer:
- Triggers on $0.50 Gemini cost threshold instead of fixed interval
- Checks every 15 min, only prints when threshold crossed
- --check flag to see current cost status
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Prayer bots (ingame/prayer_bots.js):
- 3 Mineflayer bots that actively pray, sudo, and bug_log on dev server
- Gemini 2.5 Flash Lite generates diverse natural prompts on the fly
- Falls back to static pool if Gemini unavailable
- 15-45s interval per bot, 50/35/10/5 pray/sudo/bug/chat split
POS status printer (scripts/training_status_printer.py):
- Prints training data collection status to Epson TM-m30
- Tracks: dataset size, audit logs, bot activity, Gemini API cost, service status
- Triggers on $0.50 cost threshold (configurable), checks every 15 min
- --dry-run, --check, --force flags
Training:
- First LoRA run completed (233 examples, 3 epochs, loss 1.5→0.10)
- GGUF exported and loaded into Ollama as qwen3-8b-mc-lora on steel141
- Model is bad (expected) — hallucinating Chinese, leaking system prompt
- Deployed to dev server for live testing and data collection
- bf16 fix for Ampere GPU, system prompts included in training conversations
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Switch fp16 to bf16 (RTX 3090 Ti is Ampere, supports BF16 natively)
- Include system prompt in training conversations (mode-aware: sudo/god/god_system)
- Include message field only for god modes
- Add determine_mode() and get_system_prompt() helpers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Expanded dataset from 31 to 182 examples (45 manual + 106 extracted from server logs)
- Built eval/harness.py with per-category breakdowns and baseline tracking
- Built eval/live_bakeoff.py for RCON-verified model comparison on live server
- Extracted training data from prayer logs, sudo logs, and bug reports on CT 644
- Added Reddit post draft and modmail for playtester recruitment
- Updated server context: all servers now online-mode=false + whitelist
- Updated PLAN.md with Phase 2 progress
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Key decisions: gemma3n:e4b for serving (RTX 4000), qwen3:8b for
fine-tuning base (RTX 3090 Ti). Phase 1 complete.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- training/scripts/train_lora.py: Unsloth QLoRA trainer for qwen3:8b
- training/scripts/train_lora.sh: Launch script for steel141 RTX 3090 Ti
- eval/bakeoff.py: Fixed token budget (400->1500) that caused qwen3
models to exhaust tokens on thinking, added --no-think flag
- agent/serve.py: Default model changed to gemma3n:e4b
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
gemma3n:e4b wins for production serving (80.6% cmd match, 100% safety).
qwen3:8b recommended as fine-tuning base. Full per-model analysis and
scoring methodology documented.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bake-off tested 7 models on 31 seed examples via GPU-accelerated Ollama
on node-197 RTX 4000. gemma3n:e4b leads for serving (80.6% cmd match,
100% safety, 5.9s). qwen3:8b recommended as fine-tuning base (Apache 2.0,
best syntax quality, strong ecosystem). Full research in MODEL_RESEARCH.md.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- knowledge/mc-commands/commands.json: 14 MC commands with JE syntax, args, examples, common errors, 1.21 version notes
- knowledge/server-context/servers.json: all 4 servers (mc1, shrink, paper-ai, paper-dev) with full config
- knowledge/build_index.py: TF-IDF indexer + search function (19 docs, 725 terms)
- All command syntax validated live on dev server via RCON (12/13 passed)
- PLAN.md: mark Phase 1.3 complete
- IDEA.md: project scope (Minecraft ops AI assistant via qwen3-coder LoRA/SFT)
- PLAN.md: complete roadmap with prior art analysis, architecture, phased plan, dev server docs
- data/schema.json: training example JSON Schema with negative_output support
- data/processed/seed_dataset.jsonl: 31 validated examples from repair code, prayer logs, session history
- data/validate_dataset.py: schema validator with summary statistics
- ingame/: Mineflayer bot framework (test_connect, spawn_bots, aware_bots with full event logging)
- Directory structure for knowledge/, eval/, training/, agent/ (Phase 1.3+ work)
Seth