mirror of
https://github.com/NanmiCoder/cc-haha
synced 2026-07-16 13:03:31 +08:00
0c43b270b9
ChatGPT Official runs through Codex OAuth, so context accounting should follow the Codex model catalog and its effective context window instead of public API limits. This keeps GPT-5.5 at the same usable window shown by Codex while preserving GPT-5.4's larger Codex window. Constraint: Codex OAuth model metadata applies effective_context_window_percent to raw catalog windows Rejected: Use public API GPT-5.5 1.05M context | Codex app and OAuth runtime clamp this model to the smaller effective window Confidence: high Scope-risk: narrow Directive: Keep ChatGPT Official context values aligned to Codex model catalog behavior, not generic OpenAI API docs Tested: bun test src/services/openaiAuth/models.test.ts src/services/compact/autoCompact.test.ts src/server/__tests__/providers.test.ts Tested: bun run check:server Tested: git diff --check Not-tested: Manual desktop screenshot after rebuilding the app
62 lines
2.4 KiB
TypeScript
62 lines
2.4 KiB
TypeScript
import { describe, expect, test, beforeEach, afterEach } from 'bun:test'
|
|
|
|
import { getEffectiveContextWindowSize } from './autoCompact.js'
|
|
import { getContextWindowForModel } from '../../utils/context.js'
|
|
import { MODEL_CONTEXT_WINDOWS_ENV_KEY } from '../../utils/model/modelContextWindows.js'
|
|
|
|
let originalAutoCompactWindow: string | undefined
|
|
let originalContextWindows: string | undefined
|
|
|
|
beforeEach(() => {
|
|
originalAutoCompactWindow = process.env.CLAUDE_CODE_AUTO_COMPACT_WINDOW
|
|
originalContextWindows = process.env[MODEL_CONTEXT_WINDOWS_ENV_KEY]
|
|
delete process.env.CLAUDE_CODE_AUTO_COMPACT_WINDOW
|
|
delete process.env[MODEL_CONTEXT_WINDOWS_ENV_KEY]
|
|
})
|
|
|
|
afterEach(() => {
|
|
if (originalAutoCompactWindow === undefined) {
|
|
delete process.env.CLAUDE_CODE_AUTO_COMPACT_WINDOW
|
|
} else {
|
|
process.env.CLAUDE_CODE_AUTO_COMPACT_WINDOW = originalAutoCompactWindow
|
|
}
|
|
if (originalContextWindows === undefined) {
|
|
delete process.env[MODEL_CONTEXT_WINDOWS_ENV_KEY]
|
|
} else {
|
|
process.env[MODEL_CONTEXT_WINDOWS_ENV_KEY] = originalContextWindows
|
|
}
|
|
})
|
|
|
|
describe('model context window resolution', () => {
|
|
test('uses built-in windows for current third-party coding models', () => {
|
|
expect(getContextWindowForModel('deepseek-v4-pro')).toBe(1_000_000)
|
|
expect(getContextWindowForModel('MiniMax-M2.7')).toBe(204_800)
|
|
expect(getContextWindowForModel('kimi-k2.6')).toBe(262_144)
|
|
expect(getContextWindowForModel('glm-5.1')).toBe(200_000)
|
|
expect(getContextWindowForModel('glm-4.5-air')).toBe(128_000)
|
|
})
|
|
|
|
test('uses Codex OAuth effective context windows for OpenAI GPT models', () => {
|
|
expect(getContextWindowForModel('gpt-5.5')).toBe(258_400)
|
|
expect(getContextWindowForModel('gpt-5.4')).toBe(950_000)
|
|
expect(getContextWindowForModel('gpt-5.4-mini')).toBe(258_400)
|
|
expect(getContextWindowForModel('gpt-5.3-codex-spark')).toBe(121_600)
|
|
})
|
|
|
|
test('uses per-model provider overrides before built-in defaults', () => {
|
|
process.env[MODEL_CONTEXT_WINDOWS_ENV_KEY] = JSON.stringify({
|
|
'deepseek-v4-pro': 500_000,
|
|
'custom-model': 300_000,
|
|
})
|
|
|
|
expect(getContextWindowForModel('deepseek-v4-pro')).toBe(500_000)
|
|
expect(getContextWindowForModel('provider/custom-model')).toBe(300_000)
|
|
})
|
|
|
|
test('global auto compact window can raise unknown models above the default', () => {
|
|
process.env.CLAUDE_CODE_AUTO_COMPACT_WINDOW = '1000000'
|
|
|
|
expect(getEffectiveContextWindowSize('unknown-future-model')).toBe(980_000)
|
|
})
|
|
})
|