bowenliang123
87a2893190
Apply the AGENTS.md comments guideline to every .ts under tests/: remove what the code already expresses, keep only what it cannot. Removed: spec-header coverage inventories the describe/test tree already names; comments restating the next line's assertion or its message; helper JSDoc whose body already says it; within-file duplicates; five stale claims (trendChart's "six aggregated turns" against agg.length === 8, agentGraph's "six categories" against CATS' seven, fold-timing's "cheapest call LAST" against a loop that feeds it first, fold-surface's unreachable plain-JSON note, contextView's "two turns" for a one-turn fixture). Kept and condensed: the projection-freeze class, strict wire/state-schema hazards, retention and pricing windows, pairing rules, hostile-input rationale, fixture and expectation-array ground truth, jsdom mechanics, and issue-numbered regressions. Comment lines 2604 -> 2014 (-22%); 85 module headers kept, 30 coverage inventories dropped; every surviving comment line is within the 140-column limit, with 36 wrapped lines joined onto fuller ones. No logic, string, identifier, import or directive changed: a comment-strip gate proves each file's comment-free code is byte-identical to HEAD, and a line-level proof pairs all 845 changed lines with an unchanged code part. pnpm run lint && pnpm run test (106 files / 2061 tests, per-file 100% coverage) && pnpm run build all pass.
67 lines
2.8 KiB
TypeScript
67 lines
2.8 KiB
TypeScript
// The official DeepSeek vision image-token calculator port (src/shared/imageTokens.ts). Reference values verified against the docs
|
||
// calculator itself (see the module header).
|
||
|
||
import assert from 'node:assert/strict'
|
||
import { describe, test } from 'vitest'
|
||
import { estimateImageTokens } from '../../src/shared/imageTokens'
|
||
|
||
describe('estimateImageTokens', () => {
|
||
// Verified against https://api-docs.deepseek.com/zh-cn/quick_start/token_usage.
|
||
test.each([
|
||
[2048, 1365, 313],
|
||
[800, 600, 341],
|
||
[2048, 2048, 349],
|
||
[512, 512, 201],
|
||
[100, 100, 117],
|
||
[1920, 1080, 369],
|
||
[400, 900, 249],
|
||
])('%i×%i → %i tokens (docs calculator reference)', (w, h, expected) => {
|
||
assert.equal(estimateImageTokens(w, h), expected)
|
||
})
|
||
|
||
test('every valid size lands inside the documented cap band', () => {
|
||
// The documented band is 117–384; aspect ratios at the 8:1 clamp price
|
||
// slightly under it (128×16 → 113) — same patch grid as the docs
|
||
// calculator, so pin the observed floor rather than the marketing one.
|
||
for (let w = 16; w <= 4096; w *= 2) {
|
||
for (let h = 16; h <= 4096; h *= 2) {
|
||
const tokens = estimateImageTokens(w, h)
|
||
assert.ok(tokens !== null && tokens >= 113 && tokens <= 384, `${w}×${h} → ${tokens}`)
|
||
}
|
||
}
|
||
})
|
||
|
||
test('extreme aspect ratios clamp to the 8:1 band before patching', () => {
|
||
assert.ok(estimateImageTokens(8000, 1000) !== null)
|
||
assert.ok(estimateImageTokens(1000, 8000) !== null)
|
||
})
|
||
|
||
test('a tall strip beyond the ~190:1 grid threshold solves on the collapsed-width arm', () => {
|
||
// The padded grid overflows the budget (860 tokens) and ratio 214 >
|
||
// 189.5, so the solve collapses width to one patch unit (gridW < 1 arm).
|
||
assert.equal(estimateImageTokens(50, 12000), 381)
|
||
})
|
||
|
||
test('a first-solve overflow walks the budget down until the grid fits', () => {
|
||
// Parity terms can push the solved grid a token or two over budget; the
|
||
// loop decrements until it fits (found by fuzz: 360×3243 exercises it).
|
||
assert.equal(estimateImageTokens(360, 3243), 281)
|
||
})
|
||
|
||
test('non-positive or non-finite dimensions return null', () => {
|
||
assert.equal(estimateImageTokens(0, 100), null)
|
||
assert.equal(estimateImageTokens(100, 0), null)
|
||
assert.equal(estimateImageTokens(-5, 100), null)
|
||
assert.equal(estimateImageTokens(100, -5), null)
|
||
assert.equal(estimateImageTokens(NaN, 100), null)
|
||
assert.equal(estimateImageTokens(100, Infinity), null)
|
||
})
|
||
|
||
test('astronomical (finite) dimensions fail to converge and return null instead of throwing', () => {
|
||
// w×h overflows to Infinity: the fixed-point iteration never stabilizes,
|
||
// so the 9-iteration cap narrows to null (untrusted-log defense).
|
||
assert.equal(estimateImageTokens(Number.MAX_VALUE, Number.MAX_VALUE), null)
|
||
assert.equal(estimateImageTokens(1e308, 1e308), null)
|
||
})
|
||
})
|