kamilkrzywda
5abbebc58f
- llm.ts: LlmClient.chatWithTools — bounded OpenAI function-calling loop (default 6 tool rounds, then one forced no-tools request); shares the retry/backoff/reasoning-effort machinery with chat(); tools are only added to the request body when non-empty, so plain chat() bodies and their tests are unchanged; a tool-call response is a success (never a truncation); malformed tool arguments become an error tool-message. - engine.ts: GET_NEIGHBOURS_TOOL + makeNeighboursExecutor (degrades unknown tools/args/keys without throwing); playerFrontier (sorted, deduped, other-nation-owned only); buildSystemPrompt gains a tool variant (contiguity rules, frontier hint) used only when the tool is actually sent; buildUserPrompt emits a frontier field whenever a resolver is given; runTurn takes getNeighbours and uses chatWithTools when both the resolver and the client's capability are present, falling back to plain chat() otherwise. - applyTurnResult: snapshot-first sole-ownership algorithm — a region key transfers only with exactly one claimant and exactly one previous owner (different from the gainer); double claims and pre-existing double ownership are left untouched; lone losses are no-ops; every key has 0 or 2+ -> 1 owner post-turn, never orphaned or duplicated. - stub.ts: deterministic one-region contiguous conquest per turn from the prompt's frontier field (bypasses the tool loop; empty frontier -> no transfer, all existing assertions stay stable). - regions.ts: optional region-adjacency.json load; hasAdjacency() and neighboursOf(keys) (union, deduped, queried excluded, sorted). - index.ts: jump route passes the adjacency resolver when loaded. - tests: chatWithTools protocol (echo, executor args, bounded rounds, malformed args, retry), tool-path runTurn wiring + fallback, playerFrontier, executor, applyTurnResult invariants, neighboursOf/ hasAdjacency (105 unit tests, 14 e2e — all green).
862 lines
33 KiB
TypeScript
862 lines
33 KiB
TypeScript
import { describe, expect, it } from 'vitest';
|
|
import type { GameState, Preset, RoundItem } from '../shared/types';
|
|
import { makeEvent, type TurnResult } from '../shared/protocol';
|
|
import type { LlmClient, ToolDefinition, WireMessage } from './llm';
|
|
import { makeStubLlmClient } from './stub';
|
|
import {
|
|
applyTurnResult,
|
|
buildSystemPrompt,
|
|
buildUserPrompt,
|
|
GET_NEIGHBOURS_TOOL,
|
|
isMeaningful,
|
|
makeNeighboursExecutor,
|
|
parseTurnResult,
|
|
playerFrontier,
|
|
queueRoundItem,
|
|
removeRoundItem,
|
|
runTurn,
|
|
validateTurnResult,
|
|
} from './engine';
|
|
|
|
/** A fake LlmClient that returns a fixed list of raw responses in order. */
|
|
function fakeLlm(responses: string[]): LlmClient {
|
|
let i = 0;
|
|
return {
|
|
chat: async () => responses[Math.min(i++, responses.length - 1)],
|
|
listModels: async () => [],
|
|
};
|
|
}
|
|
|
|
const preset: Preset = {
|
|
id: 'test',
|
|
title: 'Test',
|
|
startDate: '1939-09-01',
|
|
summary: '',
|
|
worldBackground: 'A test world.',
|
|
simulationRules: ['Prefer gradual change'],
|
|
nations: [
|
|
{
|
|
id: 'uk',
|
|
name: 'UK',
|
|
type: 'major',
|
|
stats: { military: 80, economy: 70, stability: 60, population: 50 },
|
|
territory: ['GBR'],
|
|
relations: { us: 'neutral', de: 'hostile' },
|
|
},
|
|
{
|
|
id: 'us',
|
|
name: 'US',
|
|
type: 'major',
|
|
stats: { military: 60, economy: 90, stability: 80, population: 80 },
|
|
territory: ['USA'],
|
|
relations: { uk: 'neutral' },
|
|
},
|
|
{
|
|
id: 'de',
|
|
name: 'DE',
|
|
type: 'major',
|
|
stats: { military: 95, economy: 70, stability: 75, population: 90 },
|
|
territory: ['DEU'],
|
|
relations: {},
|
|
},
|
|
],
|
|
};
|
|
|
|
function baseState(): GameState {
|
|
return {
|
|
id: 'g1',
|
|
presetId: 'test',
|
|
presetTitle: 'Test',
|
|
playerNationId: 'uk',
|
|
turn: 1,
|
|
date: '1939-09-01',
|
|
nations: {
|
|
uk: { ...preset.nations[0], stats: { ...preset.nations[0].stats }, territory: [...preset.nations[0].territory], relations: { ...preset.nations[0].relations } },
|
|
us: { ...preset.nations[1], stats: { ...preset.nations[1].stats }, territory: [...preset.nations[1].territory], relations: { ...preset.nations[1].relations } },
|
|
de: { ...preset.nations[2], stats: { ...preset.nations[2].stats }, territory: [...preset.nations[2].territory], relations: { ...preset.nations[2].relations } },
|
|
},
|
|
events: [],
|
|
pending: [],
|
|
version: 1,
|
|
};
|
|
}
|
|
|
|
const validResult = JSON.stringify({
|
|
date: '1939-09-03',
|
|
events: [
|
|
{ nation: 'uk', type: 'military', title: 'T-26s cross the border', text: 'The border is breached.', outcome: 'success' },
|
|
],
|
|
stateChanges: [
|
|
{ nation: 'uk', stat: 'military', op: 'set', value: 85 },
|
|
{ nation: 'de', stat: 'economy', op: 'add', value: 10 },
|
|
],
|
|
relations: [{ a: 'uk', b: 'us', set: 'ally' }],
|
|
// v0.21 (#16 P2): DE gains the UK's GBR — the key has exactly one previous
|
|
// owner, so the transfer applies (an unowned key like 'POL' would be ignored
|
|
// by the new sole-ownership rules in applyTurnResult).
|
|
territoryChanges: [{ nation: 'de', gain: ['GBR'], lose: [] }],
|
|
});
|
|
|
|
describe('runTurn', () => {
|
|
it('applies a canned TurnResult: stats, symmetric relations, territory, event, turn, date, pending cleared', async () => {
|
|
const original = baseState();
|
|
const { state, result } = await runTurn({
|
|
llm: fakeLlm([validResult]),
|
|
preset,
|
|
state: original,
|
|
items: [{ kind: 'action', text: 'We rearm the coast' }],
|
|
});
|
|
|
|
expect(state.turn).toBe(2); // the turn counter advanced
|
|
expect(state.date).toBe('1939-09-03');
|
|
expect(state.pending).toEqual([]); // committing a round always clears it
|
|
expect(state.nations.uk.stats.military).toBe(85);
|
|
expect(state.nations.de.stats.economy).toBe(80); // 70 + 10 (add)
|
|
// relations are symmetric
|
|
expect(state.nations.uk.relations.us).toBe('ally');
|
|
expect(state.nations.us.relations.uk).toBe('ally');
|
|
// territory moved to DE: it gains GBR (the UK's sole previous owner) and
|
|
// the old owner correspondingly loses it (v0.21 #16 P2: one owner per key)
|
|
expect(state.nations.de.territory).toContain('GBR');
|
|
expect(state.nations.uk.territory).not.toContain('GBR');
|
|
// one event appended, tagged with the turn that was resolved (1)
|
|
expect(state.events).toHaveLength(1);
|
|
expect(state.events[0].turn).toBe(1);
|
|
expect(state.events[0].title).toBe('T-26s cross the border');
|
|
expect(state.events[0].nation).toBe('uk');
|
|
// the return value carries the validated result too
|
|
expect(result.events[0].title).toBe('T-26s cross the border');
|
|
// and the original state object is NOT mutated
|
|
expect(original.turn).toBe(1);
|
|
expect(original.nations.uk.stats.military).toBe(80);
|
|
expect(original.nations.de.territory).not.toContain('GBR');
|
|
expect(original.nations.uk.territory).toContain('GBR');
|
|
});
|
|
|
|
it('clamps out-of-range stat values into 0-100', async () => {
|
|
const clampResult = JSON.stringify({
|
|
date: '1939-09-03',
|
|
events: [{ nation: 'uk', type: 'economic', title: 'cap', text: 'x', outcome: 'neutral' }],
|
|
stateChanges: [
|
|
{ nation: 'uk', stat: 'military', op: 'set', value: 250 },
|
|
{ nation: 'uk', stat: 'stability', op: 'set', value: -20 },
|
|
],
|
|
relations: [],
|
|
territoryChanges: [],
|
|
});
|
|
const { state } = await runTurn({
|
|
llm: fakeLlm([clampResult]),
|
|
preset,
|
|
state: baseState(),
|
|
items: [],
|
|
});
|
|
expect(state.nations.uk.stats.military).toBe(100);
|
|
expect(state.nations.uk.stats.stability).toBe(0);
|
|
});
|
|
|
|
it('survives an invalid-JSON response: no throw, advances turn, emits a generic event', async () => {
|
|
const { state, result } = await runTurn({
|
|
llm: fakeLlm(['total garbage', 'still not json']),
|
|
preset,
|
|
state: baseState(),
|
|
items: [],
|
|
});
|
|
expect(state.turn).toBe(2);
|
|
expect(state.events).toHaveLength(1);
|
|
expect(state.events[0].title).toBe('The turn was unclear');
|
|
expect(result.events[0].title).toBe('The turn was unclear');
|
|
});
|
|
|
|
it('treats a valid-but-empty result as unclear (not meaningful)', async () => {
|
|
const emptyResult = JSON.stringify({
|
|
date: '1939-09-03',
|
|
events: [],
|
|
stateChanges: [],
|
|
relations: [],
|
|
territoryChanges: [],
|
|
});
|
|
const { state } = await runTurn({
|
|
llm: fakeLlm([emptyResult, emptyResult]),
|
|
preset,
|
|
state: baseState(),
|
|
items: [],
|
|
});
|
|
expect(state.turn).toBe(2);
|
|
expect(state.events[0].title).toBe('The turn was unclear');
|
|
});
|
|
|
|
it('"jump forward" (no action) still advances the turn and emits an event', async () => {
|
|
const state = baseState();
|
|
state.turn = 3; // start mid-game
|
|
const { state: next } = await runTurn({
|
|
llm: fakeLlm([validResult]),
|
|
preset,
|
|
state, // no items = an empty round / jump forward
|
|
items: [],
|
|
});
|
|
expect(next.turn).toBe(4);
|
|
expect(next.events).toHaveLength(1);
|
|
});
|
|
});
|
|
|
|
describe('runTurn (batched round, deterministic stub)', () => {
|
|
it('a single action item: turn+1, date+2, pending cleared, one military event, military set to clamp(cur+5)', async () => {
|
|
const original = baseState();
|
|
original.pending = [{ kind: 'action', text: 'Fortify the coast and mobilize the home guard' }];
|
|
const { state, result } = await runTurn({
|
|
llm: makeStubLlmClient(),
|
|
preset,
|
|
state: original,
|
|
items: original.pending,
|
|
});
|
|
expect(state.turn).toBe(2);
|
|
expect(state.date).toBe('1939-09-03');
|
|
expect(state.pending).toEqual([]);
|
|
expect(state.nations.uk.stats.military).toBe(85); // clamp(80 + 5)
|
|
expect(state.events).toHaveLength(1);
|
|
expect(state.events[0].type).toBe('military');
|
|
expect(state.events[0].nation).toBe('uk');
|
|
expect(state.events[0].outcome).toBe('success');
|
|
expect(result.stateChanges).toEqual([{ nation: 'uk', stat: 'military', op: 'set', value: 85 }]);
|
|
// committing a turn is a new revision too
|
|
expect(state.version).toBe(2);
|
|
// the original object is not mutated (a fresh queue was consumed)
|
|
expect(original.pending).toHaveLength(1);
|
|
expect(original.version).toBe(1);
|
|
});
|
|
|
|
it('an empty round: turn advances, one political "world moves on" event, pending stays empty', async () => {
|
|
const { state } = await runTurn({
|
|
llm: makeStubLlmClient(),
|
|
preset,
|
|
state: baseState(), // pending: []
|
|
items: [],
|
|
});
|
|
expect(state.turn).toBe(2);
|
|
expect(state.date).toBe('1939-09-03');
|
|
expect(state.pending).toEqual([]);
|
|
expect(state.events).toHaveLength(1);
|
|
expect(state.events[0].type).toBe('political');
|
|
expect(state.events[0].title).toBe('The world moves on');
|
|
expect(state.events[0].outcome).toBe('neutral');
|
|
});
|
|
|
|
it('a round with two actions and one diplomacy: 3 events (2 military + 1 diplomatic), exactly one turn advance', async () => {
|
|
const state = baseState();
|
|
state.pending = [
|
|
{ kind: 'action', text: 'Reinforce the Atlantic convoys' },
|
|
{ kind: 'action', text: 'Expand the shipyards' },
|
|
{ kind: 'diplomacy', to: 'us', text: 'Propose a mutual defense pact' },
|
|
];
|
|
const { state: next } = await runTurn({
|
|
llm: makeStubLlmClient(),
|
|
preset,
|
|
state,
|
|
items: state.pending,
|
|
});
|
|
expect(next.turn).toBe(2); // a whole batch = one turn
|
|
expect(next.pending).toEqual([]);
|
|
expect(next.events).toHaveLength(3);
|
|
expect(next.events.filter((e) => e.type === 'military')).toHaveLength(2);
|
|
expect(next.events.filter((e) => e.type === 'diplomatic')).toHaveLength(1);
|
|
expect(next.nations.uk.stats.military).toBe(85); // one +5 per turn, not per item
|
|
});
|
|
});
|
|
|
|
describe('buildUserPrompt (batched round)', () => {
|
|
it('includes every queued item (action + diplomacy) in the emitted payload', () => {
|
|
const items: RoundItem[] = [
|
|
{ kind: 'action', text: 'Fortify Dover' },
|
|
{ kind: 'diplomacy', to: 'us', text: 'Propose a pact' },
|
|
];
|
|
const parsed = JSON.parse(buildUserPrompt(baseState(), items)) as {
|
|
pending: RoundItem[];
|
|
playerActions: unknown[];
|
|
roundInstruction: string;
|
|
};
|
|
expect(parsed.pending).toEqual(items);
|
|
expect(parsed.playerActions).toEqual(['Fortify Dover', { to: 'us', message: 'Propose a pact' }]);
|
|
expect(parsed.roundInstruction).toContain('Fortify Dover');
|
|
expect(parsed.roundInstruction).toContain('Propose a pact');
|
|
});
|
|
|
|
it('frames an empty round as "the world advances on its own momentum"', () => {
|
|
const parsed = JSON.parse(buildUserPrompt(baseState(), [])) as {
|
|
pending: RoundItem[];
|
|
roundInstruction: string;
|
|
};
|
|
expect(parsed.pending).toEqual([]);
|
|
expect(parsed.roundInstruction).toMatch(/no action/i);
|
|
expect(parsed.roundInstruction).toMatch(/own momentum/i);
|
|
});
|
|
|
|
it('caps recentEvents at the last 5 (in order) when the state has more than 5 events', () => {
|
|
const state = baseState();
|
|
const events = Array.from({ length: 8 }, (_, i) =>
|
|
makeEvent(
|
|
{
|
|
nation: 'uk',
|
|
type: 'political',
|
|
title: `Event ${i + 1}`,
|
|
text: `Narrative ${i + 1}`,
|
|
outcome: 'neutral',
|
|
},
|
|
i + 1,
|
|
`1939-09-0${i + 1}`,
|
|
),
|
|
);
|
|
state.events = events;
|
|
|
|
const parsed = JSON.parse(buildUserPrompt(state, [])) as {
|
|
state: { recentEvents: Array<{ title: string; turn: number; date: string }> };
|
|
};
|
|
|
|
expect(parsed.state.recentEvents).toHaveLength(5);
|
|
// the last 5 of the 8 input events, in order
|
|
expect(parsed.state.recentEvents[0].title).toBe(events[3].title); // 'Event 4'
|
|
expect(parsed.state.recentEvents[4].title).toBe(events[7].title); // 'Event 8'
|
|
expect(parsed.state.recentEvents.map((e) => e.title)).toEqual([
|
|
events[3].title,
|
|
events[4].title,
|
|
events[5].title,
|
|
events[6].title,
|
|
events[7].title,
|
|
]);
|
|
});
|
|
|
|
it('asks the model to keep its internal reasoning brief (concise-CoT constraint)', () => {
|
|
expect(buildSystemPrompt(preset)).toMatch(/reasoning brief|a few sentences/i);
|
|
});
|
|
|
|
it('includes a regions reference table only when one is provided', () => {
|
|
// with a reference table: the payload carries it (region key -> "name (parent)")
|
|
const withRef = JSON.parse(
|
|
buildUserPrompt(baseState(), [], { 'ain-lqtt': 'Ain (FRA)', 'wales-bbbb': 'Wales (GBR)' }),
|
|
) as { regions?: Record<string, string> };
|
|
expect(withRef.regions).toEqual({ 'ain-lqtt': 'Ain (FRA)', 'wales-bbbb': 'Wales (GBR)' });
|
|
// without one: the field is absent entirely (the legacy prompt shape is unchanged)
|
|
const withoutRef = JSON.parse(buildUserPrompt(baseState(), [])) as { regions?: Record<string, string> };
|
|
expect(withoutRef.regions).toBeUndefined();
|
|
});
|
|
});
|
|
|
|
describe('queueRoundItem', () => {
|
|
it('appends an item to pending without advancing the turn or the date', () => {
|
|
const before = baseState();
|
|
const after = queueRoundItem(before, { kind: 'diplomacy', to: 'us', text: 'hello' });
|
|
expect(after.turn).toBe(before.turn);
|
|
expect(after.date).toBe(before.date);
|
|
expect(after.pending).toHaveLength(1);
|
|
expect(after.pending[0]).toEqual({ kind: 'diplomacy', to: 'us', text: 'hello' });
|
|
// the original state is not mutated
|
|
expect(before.pending).toHaveLength(0);
|
|
expect(before.version).toBe(1);
|
|
});
|
|
|
|
it('bumps the monotonic version on each queue mutation', () => {
|
|
const a = queueRoundItem(baseState(), { kind: 'action', text: 'one' });
|
|
expect(a.version).toBe(2);
|
|
const b = queueRoundItem(a, { kind: 'action', text: 'two' });
|
|
expect(b.version).toBe(3);
|
|
// and never backwards: two consecutive adds from the same base both
|
|
// advance past it, so a late snapshot of the older state is always older.
|
|
const b2 = queueRoundItem(baseState(), { kind: 'action', text: 'two' });
|
|
expect(b2.version).toBe(2);
|
|
expect(b.version).toBeGreaterThan(b2.version);
|
|
});
|
|
|
|
it('treats a legacy state without a version field as version 0', () => {
|
|
const legacy = { ...baseState() } as Omit<GameState, 'version'> & { version?: number };
|
|
delete legacy.version; // a row saved before the version field existed
|
|
// simulate a legacy row coming back from SQLite: the JSON has no field.
|
|
const after = queueRoundItem(legacy as unknown as GameState, {
|
|
kind: 'action',
|
|
text: 'first after upgrade',
|
|
});
|
|
expect(after.version).toBe(1);
|
|
});
|
|
});
|
|
|
|
describe('removeRoundItem', () => {
|
|
it('removes the item at the index, bumps the version, and leaves the input untouched', () => {
|
|
const state = baseState();
|
|
state.pending = [
|
|
{ kind: 'action', text: 'one' },
|
|
{ kind: 'diplomacy', to: 'us', text: 'two' },
|
|
{ kind: 'action', text: 'three' },
|
|
];
|
|
const next = removeRoundItem(state, 1);
|
|
expect(next).not.toBeNull();
|
|
expect(next!.version).toBe(2);
|
|
expect(next!.pending).toEqual([
|
|
{ kind: 'action', text: 'one' },
|
|
{ kind: 'action', text: 'three' },
|
|
]);
|
|
expect(state.pending).toHaveLength(3); // original untouched
|
|
expect(state.version).toBe(1);
|
|
});
|
|
|
|
it('returns null for out-of-bounds or non-integer indices (route answers 400, nothing mutates)', () => {
|
|
const state = baseState();
|
|
state.pending = [{ kind: 'action', text: 'one' }];
|
|
expect(removeRoundItem(state, -1)).toBeNull();
|
|
expect(removeRoundItem(state, 1)).toBeNull(); // == length
|
|
expect(removeRoundItem(state, 5)).toBeNull();
|
|
expect(removeRoundItem(state, 0.5)).toBeNull();
|
|
expect(removeRoundItem(state, Number.NaN)).toBeNull();
|
|
expect(state.pending).toHaveLength(1); // still untouched
|
|
});
|
|
});
|
|
|
|
describe('parseTurnResult', () => {
|
|
it('parses a fenced JSON block', () => {
|
|
const r = parseTurnResult('```json\n{"date":"1939-09-03","events":[{"nation":"uk","type":"military","title":"t","text":"b","outcome":"success"}]}\n```');
|
|
expect(r?.events[0].title).toBe('t');
|
|
});
|
|
|
|
it('parses JSON prefixed by leading whitespace/newlines (reasoning model emits "\\n\\n{...}")', () => {
|
|
const r = parseTurnResult('\n\n{"date":"1939-09-03","events":[{"nation":"uk","type":"military","title":"t","text":"b","outcome":"success"}]}\n');
|
|
expect(r?.events[0].title).toBe('t');
|
|
expect(r?.date).toBe('1939-09-03');
|
|
});
|
|
|
|
it('extracts the object from surrounding prose', () => {
|
|
const r = parseTurnResult('Sure! Here is the result: {"date":"1939-09-03","events":[]} hope that helps');
|
|
expect(r).not.toBeNull();
|
|
});
|
|
|
|
it('returns null for non-JSON output', () => {
|
|
expect(parseTurnResult('no json here at all')).toBeNull();
|
|
});
|
|
|
|
it('validateTurnResult ignores unknown/invalid entries and clamps', () => {
|
|
const r = validateTurnResult({
|
|
date: '1939-09-03',
|
|
events: [
|
|
{ nation: 'uk', type: 'bogus', title: 't', text: 'b', outcome: 'weird' },
|
|
{ nation: 'uk', title: 'missing text' },
|
|
],
|
|
stateChanges: [
|
|
{ nation: 'uk', stat: 'notastat', op: 'set', value: 5 },
|
|
{ nation: 'uk', stat: 'military', op: 'set', value: 150 },
|
|
],
|
|
});
|
|
expect(r).not.toBeNull();
|
|
expect(r!.events).toHaveLength(1); // the invalid one was dropped
|
|
expect(r!.events[0].type).toBe('political'); // bogus type -> default
|
|
expect(r!.events[0].outcome).toBe('neutral'); // bogus outcome -> default
|
|
expect(r!.stateChanges).toHaveLength(1);
|
|
expect(r!.stateChanges[0].value).toBe(100); // clamped
|
|
});
|
|
});
|
|
|
|
describe('applyTurnResult', () => {
|
|
it('does not mutate its input', () => {
|
|
const state = baseState();
|
|
const snapshot = JSON.stringify(state);
|
|
const result: TurnResult = {
|
|
date: '1939-09-03',
|
|
events: [{ nation: 'uk', type: 'military', title: 't', text: 'b', outcome: 'success' }],
|
|
stateChanges: [{ nation: 'uk', stat: 'military', op: 'set', value: 99 }],
|
|
relations: [],
|
|
territoryChanges: [],
|
|
};
|
|
applyTurnResult(state, result);
|
|
expect(JSON.stringify(state)).toBe(snapshot);
|
|
});
|
|
|
|
it('clears the pending round on commit and bumps the version', () => {
|
|
const state = baseState();
|
|
state.pending = [{ kind: 'action', text: 'x' }];
|
|
const result: TurnResult = {
|
|
date: '1939-09-03',
|
|
events: [],
|
|
stateChanges: [],
|
|
relations: [],
|
|
territoryChanges: [],
|
|
};
|
|
const next = applyTurnResult(state, result);
|
|
expect(next.pending).toEqual([]);
|
|
expect(next.version).toBe(2);
|
|
expect(state.pending).toHaveLength(1); // original untouched
|
|
expect(state.version).toBe(1);
|
|
});
|
|
|
|
it('with a validKeys set: applies real region keys, ignores hallucinated ones', () => {
|
|
const state = baseState();
|
|
// give the fixture's uk a real key ('wales-bbbb') for DE to be able to take
|
|
state.nations.uk.territory = ['GBR', 'wales-bbbb'];
|
|
const result: TurnResult = {
|
|
date: '1939-09-03',
|
|
events: [],
|
|
stateChanges: [],
|
|
relations: [],
|
|
// 'wales-bbbb' is a real key (owned by uk in the fixture); 'not-a-region-xyz'
|
|
// is not a real key at all. Both are proposed as a gain by DE.
|
|
territoryChanges: [{ nation: 'de', gain: ['wales-bbbb', 'not-a-region-xyz'], lose: [] }],
|
|
};
|
|
const next = applyTurnResult(state, result, new Set(['wales-bbbb']));
|
|
// the valid key was gained (taken from its sole previous owner); the bogus one was ignored
|
|
expect(next.nations.de.territory).toContain('wales-bbbb');
|
|
expect(next.nations.uk.territory).not.toContain('wales-bbbb');
|
|
// the bogus key never entered any nation's territory
|
|
for (const n of Object.values(next.nations)) {
|
|
expect(n.territory).not.toContain('not-a-region-xyz');
|
|
}
|
|
});
|
|
|
|
it('without a validKeys set: applies every listed key (legacy/unit path)', () => {
|
|
const state = baseState();
|
|
state.nations.uk.territory = ['GBR', 'wales-bbbb'];
|
|
const result: TurnResult = {
|
|
date: '1939-09-03',
|
|
events: [],
|
|
stateChanges: [],
|
|
relations: [],
|
|
// 'wales-bbbb' has a sole previous owner (uk) -> it transfers; 'POL' is
|
|
// owned by no one -> it is IGNORED (a key may not end a turn unowned).
|
|
territoryChanges: [{ nation: 'de', gain: ['POL', 'wales-bbbb'], lose: [] }],
|
|
};
|
|
const next = applyTurnResult(state, result); // no validKeys -> nothing filtered
|
|
expect(next.nations.de.territory).toContain('wales-bbbb');
|
|
expect(next.nations.uk.territory).not.toContain('wales-bbbb');
|
|
expect(next.nations.de.territory).not.toContain('POL');
|
|
});
|
|
});
|
|
|
|
describe('applyTurnResult (sole-ownership invariants, v0.21 #16 P2)', () => {
|
|
const noEvents = {
|
|
date: '1939-09-03',
|
|
events: [],
|
|
stateChanges: [],
|
|
relations: [],
|
|
};
|
|
|
|
it('a double claim (two different gainers, one key) is rejected: the key stays with its original owner', () => {
|
|
const state = baseState();
|
|
state.nations.uk.territory = ['GBR', 'wales-bbbb'];
|
|
const next = applyTurnResult(state, {
|
|
...noEvents,
|
|
territoryChanges: [
|
|
{ nation: 'de', gain: ['wales-bbbb'], lose: [] },
|
|
{ nation: 'us', gain: ['wales-bbbb'], lose: [] },
|
|
],
|
|
});
|
|
expect(next.nations.uk.territory).toContain('wales-bbbb');
|
|
expect(next.nations.de.territory).not.toContain('wales-bbbb');
|
|
expect(next.nations.us.territory).not.toContain('wales-bbbb');
|
|
});
|
|
|
|
it('duplicate gain entries for the SAME nation are deduped (one claim, one copy)', () => {
|
|
const state = baseState();
|
|
state.nations.uk.territory = ['GBR', 'wales-bbbb'];
|
|
const next = applyTurnResult(state, {
|
|
...noEvents,
|
|
territoryChanges: [
|
|
{ nation: 'de', gain: ['wales-bbbb'], lose: [] },
|
|
{ nation: 'de', gain: ['wales-bbbb'], lose: [] },
|
|
],
|
|
});
|
|
expect(next.nations.de.territory.filter((k) => k === 'wales-bbbb')).toHaveLength(1);
|
|
expect(next.nations.uk.territory).not.toContain('wales-bbbb');
|
|
});
|
|
|
|
it('a lone "lose" is a no-op: the key cannot be orphaned (zero owners)', () => {
|
|
const state = baseState();
|
|
const next = applyTurnResult(state, {
|
|
...noEvents,
|
|
territoryChanges: [{ nation: 'uk', gain: [], lose: ['GBR'] }],
|
|
});
|
|
expect(next.nations.uk.territory).toContain('GBR');
|
|
});
|
|
|
|
it('a self-gain is a no-op (a nation keeps exactly one copy of its own key)', () => {
|
|
const state = baseState();
|
|
const next = applyTurnResult(state, {
|
|
...noEvents,
|
|
territoryChanges: [{ nation: 'uk', gain: ['GBR'], lose: [] }],
|
|
});
|
|
expect(next.nations.uk.territory.filter((k) => k === 'GBR')).toHaveLength(1);
|
|
});
|
|
|
|
it('a pre-existing double ownership is never touched (a key with 2 owners keeps 2)', () => {
|
|
const state = baseState();
|
|
state.nations.us.territory = ['USA', 'GBR']; // uk AND us already share GBR
|
|
const next = applyTurnResult(state, {
|
|
...noEvents,
|
|
territoryChanges: [{ nation: 'de', gain: ['GBR'], lose: [] }],
|
|
});
|
|
expect(next.nations.uk.territory).toContain('GBR');
|
|
expect(next.nations.us.territory).toContain('GBR');
|
|
expect(next.nations.de.territory).not.toContain('GBR');
|
|
});
|
|
|
|
it('ownership is conserved across a transfer: exactly one owner before and after', () => {
|
|
const state = baseState();
|
|
state.nations.uk.territory = ['GBR', 'wales-bbbb'];
|
|
const count = (s: GameState) =>
|
|
Object.values(s.nations).reduce(
|
|
(acc, n) => acc + n.territory.filter((k) => k === 'wales-bbbb').length,
|
|
0,
|
|
);
|
|
const next = applyTurnResult(state, {
|
|
...noEvents,
|
|
territoryChanges: [{ nation: 'de', gain: ['wales-bbbb'], lose: [] }],
|
|
});
|
|
expect(count(state)).toBe(1);
|
|
expect(count(next)).toBe(1);
|
|
expect(next.nations.de.territory).toContain('wales-bbbb');
|
|
expect(next.nations.uk.territory).not.toContain('wales-bbbb');
|
|
});
|
|
});
|
|
|
|
describe('makeNeighboursExecutor (v0.21 #16 P2)', () => {
|
|
const getNeighbours = (keys: string[]) => keys.map((k) => `nb-${k}`);
|
|
|
|
it('answers getNeighbours with the resolver output', () => {
|
|
const exec = makeNeighboursExecutor(getNeighbours);
|
|
expect(exec('getNeighbours', { regionKeys: ['a', 'b'] })).toBe(
|
|
JSON.stringify({ neighbours: ['nb-a', 'nb-b'] }),
|
|
);
|
|
});
|
|
|
|
it('degrades unknown tools, malformed args, and unknown keys without throwing', () => {
|
|
const exec = makeNeighboursExecutor(getNeighbours, new Set(['a']));
|
|
expect(exec('otherTool', {})).toBe(JSON.stringify({ error: 'unknown tool: otherTool' }));
|
|
expect(exec('getNeighbours', { regionKeys: 'not-an-array' })).toBe(
|
|
JSON.stringify({ neighbours: [] }),
|
|
);
|
|
expect(exec('getNeighbours', { regionKeys: ['a', 'zzz-not-valid'] })).toBe(
|
|
JSON.stringify({ neighbours: ['nb-a'] }), // the unknown key is filtered out
|
|
);
|
|
});
|
|
});
|
|
|
|
describe('playerFrontier (v0.21 #16 P2)', () => {
|
|
const adjacency: Record<string, string[]> = {
|
|
GBR: ['DEU', 'USA'],
|
|
GBR2: ['DEU'],
|
|
DEU: ['GBR', 'GBR2', 'USA'],
|
|
USA: ['GBR', 'DEU'],
|
|
};
|
|
const getNeighbours = (keys: string[]) =>
|
|
[...new Set(keys.flatMap((k) => adjacency[k] ?? []))]
|
|
.filter((k) => !keys.includes(k))
|
|
.sort();
|
|
|
|
it('is the sorted, deduped set of other-nation keys bordering the player', () => {
|
|
const state = baseState();
|
|
state.nations.uk.territory = ['GBR', 'GBR2']; // the player owns both
|
|
// DEU (de) and USA (us) border the player; deduped and sorted
|
|
expect(playerFrontier(state, getNeighbours)).toEqual(['DEU', 'USA']);
|
|
});
|
|
|
|
it('drops candidates the player itself owns (a sloppy resolver may report them)', () => {
|
|
const state = baseState();
|
|
state.nations.uk.territory = ['GBR', 'GBR2'];
|
|
const sloppy = () => ['GBR2', 'DEU'];
|
|
expect(playerFrontier(state, sloppy)).toEqual(['DEU']);
|
|
});
|
|
|
|
it('is empty when the player has no territory or no controllable border', () => {
|
|
const empty = baseState();
|
|
empty.nations.uk.territory = [];
|
|
expect(playerFrontier(empty, getNeighbours)).toEqual([]);
|
|
const noBorders = () => [];
|
|
expect(playerFrontier(baseState(), noBorders)).toEqual([]);
|
|
});
|
|
});
|
|
|
|
describe('runTurn (getNeighbours tool path, v0.21 #16 P2)', () => {
|
|
const adjacency: Record<string, string[]> = {
|
|
GBR: ['DEU'],
|
|
DEU: ['GBR', 'USA'],
|
|
USA: ['DEU'],
|
|
};
|
|
const getNeighbours = (keys: string[]) =>
|
|
[...new Set(keys.flatMap((k) => adjacency[k] ?? []))]
|
|
.filter((k) => !keys.includes(k))
|
|
.sort();
|
|
|
|
const finalTurn = JSON.stringify({
|
|
date: '1939-09-03',
|
|
events: [
|
|
{
|
|
nation: 'uk',
|
|
type: 'military',
|
|
title: 'The border region falls',
|
|
text: 'UK seizes the contiguous region.',
|
|
outcome: 'success',
|
|
},
|
|
],
|
|
stateChanges: [],
|
|
relations: [],
|
|
// the player gains its only frontier region (DEU, owned by de)
|
|
territoryChanges: [{ nation: 'uk', gain: ['DEU'], lose: [] }],
|
|
});
|
|
|
|
/**
|
|
* A fake LlmClient implementing BOTH chat() and chatWithTools(). The
|
|
* chatWithTools fake plays the model's side of the protocol: it records
|
|
* the initial messages + tools, executes the scripted tool calls through
|
|
* the executor the engine handed it, and returns the final content.
|
|
*/
|
|
function toolLlm(final: string, scriptedCalls: Array<{ name: string; args: Record<string, unknown> }>) {
|
|
const seen: {
|
|
firstMessages: WireMessage[];
|
|
tools: ToolDefinition[];
|
|
executorCalls: Array<{ name: string; args: Record<string, unknown>; result: string }>;
|
|
chatCalls: number;
|
|
} = { firstMessages: [], tools: [], executorCalls: [], chatCalls: 0 };
|
|
const client: LlmClient = {
|
|
chat: async (messages) => {
|
|
seen.chatCalls += 1;
|
|
seen.firstMessages = messages as WireMessage[];
|
|
return final;
|
|
},
|
|
listModels: async () => [],
|
|
chatWithTools: async (messages, tools, executeTool) => {
|
|
seen.firstMessages = messages;
|
|
seen.tools = tools;
|
|
for (const call of scriptedCalls) {
|
|
seen.executorCalls.push({ ...call, result: executeTool(call.name, call.args) });
|
|
}
|
|
return final;
|
|
},
|
|
};
|
|
return { client, seen };
|
|
}
|
|
|
|
it('uses chatWithTools with the tool, a frontier in the prompt, the wired executor, and applies the transfer', async () => {
|
|
const original = baseState();
|
|
original.pending = [{ kind: 'action', text: 'Seize the border region' }];
|
|
const { client, seen } = toolLlm(finalTurn, [
|
|
{ name: 'getNeighbours', args: { regionKeys: ['GBR'] } },
|
|
]);
|
|
const { state } = await runTurn({
|
|
llm: client,
|
|
preset,
|
|
state: original,
|
|
items: original.pending,
|
|
getNeighbours,
|
|
});
|
|
|
|
// the tool was offered and the system prompt advertises it
|
|
expect(seen.tools.map((t) => t.function.name)).toEqual([GET_NEIGHBOURS_TOOL.function.name]);
|
|
const system = seen.firstMessages[0] as { role: 'system'; content: string };
|
|
expect(system.content).toContain('getNeighbours');
|
|
|
|
// the user payload carries the player's frontier (DEU: the player's only
|
|
// controllable border, owned by de)
|
|
const user = seen.firstMessages[1] as { role: 'user'; content: string };
|
|
const payload = JSON.parse(user.content) as { frontier?: string[] };
|
|
expect(payload.frontier).toEqual(['DEU']);
|
|
|
|
// the executor the engine built answers from the resolver
|
|
expect(seen.executorCalls).toHaveLength(1);
|
|
expect(seen.executorCalls[0].args).toEqual({ regionKeys: ['GBR'] });
|
|
expect(seen.executorCalls[0].result).toBe(JSON.stringify({ neighbours: ['DEU'] }));
|
|
|
|
// chat() was never used on the tool path
|
|
expect(seen.chatCalls).toBe(0);
|
|
|
|
// the contiguous transfer was applied: uk gains DEU, de loses it
|
|
expect(state.nations.uk.territory).toContain('DEU');
|
|
expect(state.nations.de.territory).not.toContain('DEU');
|
|
});
|
|
|
|
it('falls back to plain chat() when the client lacks chatWithTools (no tool; the frontier hint stays in the prompt)', async () => {
|
|
const original = baseState();
|
|
const { client, seen } = toolLlm(validResult, []);
|
|
delete client.chatWithTools; // simulate a minimal client (chat only)
|
|
const { state } = await runTurn({
|
|
llm: client,
|
|
preset,
|
|
state: original,
|
|
items: original.pending,
|
|
getNeighbours,
|
|
});
|
|
expect(seen.chatCalls).toBe(1);
|
|
expect(seen.tools).toEqual([]); // the tool path was never entered
|
|
const system = seen.firstMessages[0] as { content: string };
|
|
expect(system.content).not.toContain('getNeighbours');
|
|
const user = seen.firstMessages[1] as { content: string };
|
|
const payload = JSON.parse(user.content) as { frontier?: string[] };
|
|
// the frontier hint stays (a resolver was given) even without the tool —
|
|
// it also tells the model which borders are realistically contestable
|
|
expect(payload.frontier).toEqual(['DEU']);
|
|
// the canned result still applied (GBR moved uk -> de)
|
|
expect(state.nations.de.territory).toContain('GBR');
|
|
expect(state.nations.uk.territory).not.toContain('GBR');
|
|
});
|
|
});
|
|
|
|
describe('deterministic stub (one contiguous conquest per turn, v0.21 #16 P2)', () => {
|
|
const adjacency: Record<string, string[]> = {
|
|
GBR: ['DEU'],
|
|
DEU: ['GBR', 'USA'],
|
|
USA: ['DEU'],
|
|
};
|
|
const getNeighbours = (keys: string[]) =>
|
|
[...new Set(keys.flatMap((k) => adjacency[k] ?? []))]
|
|
.filter((k) => !keys.includes(k))
|
|
.sort();
|
|
|
|
it('transfers the first frontier region to the player (and its owner loses it)', async () => {
|
|
const original = baseState();
|
|
original.pending = [{ kind: 'action', text: 'March on the border region' }];
|
|
const { state } = await runTurn({
|
|
llm: makeStubLlmClient(),
|
|
preset,
|
|
state: original,
|
|
items: original.pending,
|
|
getNeighbours,
|
|
});
|
|
expect(state.nations.uk.territory).toContain('DEU'); // frontier[0]
|
|
expect(state.nations.de.territory).not.toContain('DEU');
|
|
// the rest of the stub behavior is unchanged
|
|
expect(state.turn).toBe(2);
|
|
expect(state.date).toBe('1939-09-03');
|
|
expect(state.nations.uk.stats.military).toBe(85); // clamp(80 + 5)
|
|
expect(state.events).toHaveLength(1);
|
|
});
|
|
|
|
it('with an empty frontier (no controllable border, like the UK in wwii-1939): no transfer, the turn still advances', async () => {
|
|
const original = baseState();
|
|
original.pending = [{ kind: 'action', text: 'Hold the line' }];
|
|
const { state } = await runTurn({
|
|
llm: makeStubLlmClient(),
|
|
preset,
|
|
state: original,
|
|
items: original.pending,
|
|
getNeighbours: () => [], // no borders at all
|
|
});
|
|
expect(state.nations.uk.territory).toEqual(['GBR']); // untouched
|
|
expect(state.nations.de.territory).toEqual(['DEU']);
|
|
expect(state.nations.us.territory).toEqual(['USA']);
|
|
expect(state.turn).toBe(2);
|
|
expect(state.events).toHaveLength(1);
|
|
});
|
|
});
|
|
|
|
describe('buildSystemPrompt (tool mode, v0.21 #16 P2)', () => {
|
|
it('mentions the getNeighbours tool and the contiguity rules only when the tool is sent', () => {
|
|
const plain = buildSystemPrompt(preset);
|
|
expect(plain).not.toContain('getNeighbours');
|
|
expect(plain).toMatch(/Only change borders when war/i); // the legacy rule stays
|
|
const withTool = buildSystemPrompt(preset, true);
|
|
expect(withTool).toContain('getNeighbours');
|
|
expect(withTool).toMatch(/contiguous/i);
|
|
expect(withTool).toMatch(/frontier/i);
|
|
expect(withTool).not.toMatch(/Only change borders when war/i);
|
|
});
|
|
});
|
|
|
|
describe('isMeaningful', () => {
|
|
it('is true with an event, false when everything is empty', () => {
|
|
expect(isMeaningful({ date: 'x', events: [{ nation: 'uk', type: 'military', title: 't', text: 'b', outcome: 'neutral' }], stateChanges: [], relations: [], territoryChanges: [] })).toBe(true);
|
|
expect(isMeaningful({ date: 'x', events: [], stateChanges: [{ nation: 'uk', stat: 'military', op: 'set', value: 5 }], relations: [], territoryChanges: [] })).toBe(true);
|
|
expect(isMeaningful({ date: 'x', events: [], stateChanges: [], relations: [], territoryChanges: [] })).toBe(false);
|
|
});
|
|
});
|