Try it

Copy the code ↓
  1. Data source
  2. Decision schema
  3. Model inference
  4. Structured result
Demo · no sign-in
2 Decision schema 3 typed questions
  • actionchoiceattackguarddodgespecialretreat
  • threatscorelowmediumhighcritical
  • needs_new_plannoulprobability 0–1

3 Model inference typesafe/jev-1.13

Waiting for a run.

4 Structured result

Run the demo to see the JSON answers, the highlighted pick and the final action.

Call Jev yourselfSame data and schema, straight to Flatkey. Set FLATKEY_API_KEY and run.
curl -sS https://router.flatkey.ai/api/alpha/decisions \
  -H "Authorization: Bearer $FLATKEY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "typesafe/jev-1.13",
  "state": "Tick 212. You: HP 38%, chakra 70%, special move ready. Opponent: HP 55%, just finished a combo and is in recovery frames at close range. Planner'\''s strategy: bait combos, punish recovery with the special move.",
  "questions": {
    "action": {
      "type": "choice",
      "instructions": "Given the current plan, which action should the fighter take this tick?",
      "criteria": {
        "attack": "Close-range basic attack or combo",
        "guard": "Block and hold position",
        "dodge": "Evade or reposition away from an incoming attack",
        "special": "Use a special move or jutsu that is ready",
        "retreat": "Back off to recover resources"
      }
    },
    "threat": {
      "type": "score",
      "instructions": "How dangerous is the opponent'\''s position right now?",
      "criteria": [
        "low",
        "medium",
        "high",
        "critical"
      ]
    },
    "needs_new_plan": {
      "type": "noul",
      "instructions": "The situation has changed enough that the System 2 planner should revise the strategy."
    }
  }
}'

No Flatkey key yet? Get one for Jev 1.13 →

Run with real Jev

Sign in and connect a key

  1. Sign in with Google.
  2. Connect your Flatkey or OpenRouter key on the next screen. Get a Flatkey key for Jev 1.13.
  3. Come straight back here and press Run again.

This site is for researching, sharing and cataloguing Jev use cases. Your key is only there to make trying Jev convenient: it is stored encrypted, used only for runs you trigger, never used for anything else, and you can delete it anytime.

Sign in with Google

About this use case

Use Jev for hybrid game agents: incoming state, Jev decision, routed outcome

Who / when

Audience

  • game-AI hobbyists and researchers wiring LLMs into real-time games
  • agent builders who want a fast reflex layer under a slow planner
  • teams whose agent costs are dominated by tiny, repeated decisions

Triggers

  • the agent must act every frame or tick, faster than an LLM responds
  • most decisions are small (attack, guard, dodge) and a few are strategic
  • planner calls dominate cost even though most ticks need no new plan
  • you want an explicit rule for when to wake the expensive model

The workflow

  1. 01
    Game stateHP, resources, positions, current planInput
  2. 02
    System 1 · ChoiceAction this tickJev
  3. 03
    Score + NoulThreat level, needs-new-plan flagJev
  4. 04
    Gate → System 2Act now or wake the LLM plannerYour code

Question shape

{
  "action": {
    "type": "choice",
    "instructions": "Given the current plan, which action should the fighter take this tick?",
    "criteria": {
      "attack": "Close-range basic attack or combo",
      "guard": "Block and hold position",
      "dodge": "Evade or reposition away from an incoming attack",
      "special": "Use a special move or jutsu that is ready",
      "retreat": "Back off to recover resources"
    }
  },
  "threat": {
    "type": "score",
    "instructions": "How dangerous is the opponent's position right now?",
    "criteria": [
      "low",
      "medium",
      "high",
      "critical"
    ]
  },
  "needs_new_plan": {
    "type": "noul",
    "instructions": "The situation has changed enough that the System 2 planner should revise the strategy."
  }
}

Gate policy

≥ 0.90

Execute the action this tick; no planner call.

0.65–0.90

Take the safer of the top two actions and keep the current plan.

< 0.65

Guard and call the System 2 planner for a new plan.

Failure modes

Pre-launch checks before you trust auto-routing:

  • Stale plans: System 1 keeps executing a strategy the situation has outgrown;
  • Spatial reasoning from text-only state (distances, facing, hitboxes);
  • Oscillating between two actions on consecutive ticks;
  • Planner calls that arrive too late for the moment they were meant for;
  • Game-specific moves described too vaguely to tell apart;
Keep exact work in code: cooldowns, resource checks, frame timing and input legality should be deterministic. Jev picks among legal actions; the game engine enforces the rules.

FAQ

What does the cost split look like in practice?

Vedant had Jev play Naruto as System 1 while Claude played Shikamaru as System 2 (post below). In one fight, 330 Jev decisions cost under 1 cent and 7 Claude calls cost 18.5 cents, which he puts at roughly 900× cheaper per call. His conclusion: neither model wins alone.

When should the agent call the LLM planner?

When the needs_new_plan answer is true, when confidence drops below your low band, or on a fixed timer as a backstop. Everything else stays in the fast loop.

Is this only for fighting games?

No. Any agent with frequent small decisions under a slower strategy fits the same split: RTS unit control, trading bots in dry-run, robotics behaviors or long-running coding agents. See the RTS micro page for a harder, more honest case.

Can I bring my own API key?

Yes. Try the free demo above without signing in. To run it live on your own text, sign in with Google and connect a Flatkey or OpenRouter key (a quick test call runs when you save it). Manage or delete it anytime under Dashboard → BYOK. Keys are stored encrypted on the server, used only for runs you trigger, and never sent back to the browser. Need a Flatkey key? Get one for Jev 1.13 on Flatkey.

Is this the official TypeSafe Jev product page?

No. jev-llm.com is an independent research and practice site. Verify pricing, models and terms at the original TypeSafe sources.

Sources

Seen on X
Independent-site note: jev-llm.com is an independent research and practice site. It is not the TypeSafe AI official site. Verify models, pricing and terms at the original source.