jevbooks

← All projects

keeping-an-npc-on-quest

ghosthat-studio/keeping-an-npc-on-quest · Homepage

Keep an LLM NPC on quest with staged facts and Jev checks.

Automated README summary

Automatically summarized from the README on 13 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.

What it is

Companion code for Old Nain episode 002: a terminal text game where Boone, an NPC written by a local model, must not spoil the plot. For educational use only.

How it uses Jev

Jev classifies each player message (Choice: quest, small talk, tangent, breaking game) to set Boone's job. After the local model writes a line, four Nouls check it at once: gives away a locked fact, invents something, breaks character, steers back. A failed check gets one rewrite with the reason, then a fallback line.

Primitives:choicenoul

Technique worth stealing

Stage facts so the model only sees what the player has unlocked, then use Jev to classify player messages and check every generated line.

Try it

uv sync; ollama pull hf.co/TheDrummer/Cydonia-24B-v4.3-GGUF:Q4_K_M; uv run python -m npc_on_quest chat --show-jev

View on GitHub

Start using Jev: API key and your first request →

Model judgments from README excerpts

Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.

Models & evaluation dates

Inclusion gate
jev-1.13.0
Project signals
jev-1.13.0
Category
jev-1.13.0
Pattern tags
jev-1.13.0

Evidence

These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.

  • Jev is core to the projectunclear0.60
  • Shows a System One patternnot found in the excerpt0.34
  • Describes handling uncertaintynot found in the excerpt0.08
  • Reports measured numbersnot found in the excerpt0.06
  • Is a runnable toolunclear0.47
  • Top-10 learning shortlist signalnot found in the excerpt0.38
  • Is a replica of the modelnot found in the excerpt0.08
  • Solves a common problem (0–2)score on a 0–2 scale0.32 / 2
  • About JevFound in material0.97
Signals
Jev is core to the project0.60Shows a System One pattern0.34Describes handling uncertainty0.08Reports measured numbers0.06Is a runnable tool0.47Top-10 learning shortlist signal0.38Is a replica of the model0.0800.51Solves a common problem (0–2)0.32 / 2012

no-judgment band (0.40–0.60)

Design patterns
Funnel0.91Questionnaire0.84Ghostwriter0.46Question bank0.45Traffic light0.26Thermostat0.25Check-up0.22Blind review0.2000.51

no-judgment band (0.40–0.60)

The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 13 Oct 2026.