Files

T

Mortdecai 9c2c9a2310 1200+ distilled gold examples, journal system, redstone mastery, safety awareness

Distilled Training Data (1,203 examples):
- 341 initial gold (plugins, enchantments, builds, effects, god, errors)
- 165 buildings + pipeline (100 structures built on dev, 65 request→query→act)
- 24 safety-aware (worldborder, safe tp, intentional harm, gamemode checks)
- 17 advanced logic (decanonized items, redstone gates, iterative builds)
- 12 redstone mastery (NOT/OR/AND/XOR/RS-latch/T-flip-flop/comparator/clock)
- 7 circuit verification and diagnosis
- 1 compact comparator gates
- 10 redstone methodology (build→test→save→recall→learn from mistakes)
- 8 player journal usage
- 29 creative+uncommon+pipeline+god with full tool chains

Player Journal System:
- agent/tools/player_journal.py — per-player text files (1-10 lines)
- journal.read + journal.write tool schemas added
- Cross-contaminated: God and Sudo share same journal per player
- Includes sentiment, relationship, builds, preferences, skill level

Redstone Engineering:
- agent/prompts/redstone_rules.md — baked-in wall torch, dedicated lead, repeater rules
- Learned from 4 iterations of 8-switch circuit: wall_torch on back face, not top
- T-junction bypass prevention: dedicated lead wire between merge and NOT block
- RCON limitation: can build circuits but cannot test them (lever toggle doesn't propagate)

Training Data Cleaning:
- 466 @s→@p fixes, 10 template commands removed
- 12 outdated refusals replaced with correct plugin commands
- Data de-duped across all sources

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-03-21 20:50:52 -04:00

1.5 KiB

Raw Blame History

[x] Let's add prompt templating and store/load the prompts from either 1 file or different files for each template. We will do this as a first step before implementing @CHAT-IDEA.md

Idea for interactive "chat app" that lets me manipulate the full gateway call and send a call to the gateway, and monitor the results. The end would go to the gateway would operate on dev-server.

Mode switch- We can change between manipulating the call that gets sent to the gateway and the first gateway call that gets sent to the model.

UI: Built for a single call, tooling model, just like what is implemented on prod and dev now. Top of screen: select mode(Gateway call vs Model call), independant template selector(based on mode). Top half: text box with template preloaded, editable, send button. Bottom half, scollable running log of gateway send, rcon, thinking(if available), gateway recieve, basically everything useful including what happens on the server.

Log is created in a separate file from training. It can be accessed later for training audits and to create training data. Training log records results as if the model were operating normally.

There would be different templates for god mode or sudo or self play, each a gateway call with the corresponding first model call.

Take design liberties if you can achieve the same result and make a better ui.

Implement existing applications if they fit the use case.

caddy is pct 600 on pve241, this would be chat.api.mortdec.ai hidden behind google oauth. It would run on pct 650 which is on pve112

1.5 KiB Raw Blame History