fix(codex): read full first line for session validation

`readFirstLine` allocated a fixed 16 KB buffer, but Codex CLI 0.128+
embeds the entire base_instructions / system prompt in the
`session_meta` line, pushing it past 20 KB. When the buffer doesn't
catch a newline, `isValidCodexSession` rejects the session, so every
recent Codex session is silently excluded from totals.

Switch to a streaming readline read so the first line is captured
regardless of length, and add a regression test that creates a
40 KB session_meta payload.

Locally, this changes my 30-day Codex total from €267 (only ~half
of sessions parsed) to €878 (all sessions parsed).
This commit is contained in:
ozymandiashh 2026-05-02 02:17:53 +03:00
parent 033da415c5
commit 945da9f0ba
2 changed files with 48 additions and 12 deletions

View file

@ -1,4 +1,6 @@
import { readdir, stat, open } from 'fs/promises'
import { readdir, stat } from 'fs/promises'
import { createReadStream } from 'fs'
import { createInterface } from 'readline'
import { basename, join } from 'path'
import { homedir } from 'os'
@ -70,21 +72,29 @@ function sanitizeProject(cwd: string): string {
}
async function readFirstLine(filePath: string): Promise<CodexEntry | null> {
let fh
// Codex CLI 0.128+ writes a session_meta line that can exceed 20 KB because
// it embeds the full base_instructions / system prompt. A fixed-size buffer
// would miss the trailing newline and reject the session as invalid.
// Stream the file via readline to read the first line regardless of length.
const stream = createReadStream(filePath, { encoding: 'utf-8' })
const rl = createInterface({ input: stream, crlfDelay: Infinity })
let firstLine: string | undefined
try {
fh = await open(filePath, 'r')
const buf = Buffer.alloc(16384)
const { bytesRead } = await fh.read(buf, 0, 16384, 0)
if (bytesRead === 0) return null
const text = buf.toString('utf-8', 0, bytesRead)
const nl = text.indexOf('\n')
const line = nl >= 0 ? text.slice(0, nl) : text
if (!line.trim()) return null
return JSON.parse(line) as CodexEntry
for await (const line of rl) {
firstLine = line
break
}
} catch {
return null
} finally {
await fh?.close()
rl.close()
stream.destroy()
}
if (!firstLine || !firstLine.trim()) return null
try {
return JSON.parse(firstLine) as CodexEntry
} catch {
return null
}
}

View file

@ -123,6 +123,32 @@ describe('codex provider - session discovery', () => {
expect(sessions).toHaveLength(1)
})
it('accepts session_meta lines larger than 16 KB (Codex CLI 0.128+)', async () => {
// Codex CLI 0.128+ embeds the full base_instructions / system prompt in the
// first session_meta line, often pushing it past 20 KB. Regression guard
// against a fixed-size buffer in readFirstLine.
const bigPayload = JSON.stringify({
type: 'session_meta',
timestamp: '2026-05-02T00:00:00Z',
payload: {
cwd: '/Users/test/big',
originator: 'codex-tui',
session_id: 'sess-big',
model: 'gpt-5.5',
base_instructions: { text: 'x'.repeat(40_000) },
},
})
await writeSession(tmpDir, '2026-05-02', 'rollout-big.jsonl', [
bigPayload,
tokenCount({ last: { input: 100, output: 50 }, total: { total: 150 } }),
])
const provider = createCodexProvider(tmpDir)
const sessions = await provider.discoverSessions()
expect(sessions).toHaveLength(1)
expect(sessions[0]!.path).toContain('rollout-big.jsonl')
})
it('skips files without codex session_meta', async () => {
const [year, month, day] = '2026-04-14'.split('-')
const sessionDir = join(tmpDir, 'sessions', year!, month!, day!)