Files
gbrain/test/subagent-transcript.test.ts
T
Garry TanandClaude Opus 4.7 e9be322b8e feat(minions): subagent-audit JSONL + transcript renderer
Two small plumbing pieces the v0.15 subagent handler + `gbrain agent logs`
depend on:

subagent-audit.ts — JSONL-rotated audit log mirroring the shell-audit
pattern. Two event flavors: submission (one line per job submit) and
heartbeat (one line per turn boundary — llm_call_started / completed /
tool_called / tool_result / tool_failed). Heartbeats fix the "--follow on
a long Anthropic call shows nothing for 30 seconds" problem codex flagged.
Never logs prompts or tool inputs (PII risk — subagent input_vars may
carry user-supplied free text); DOES log tokens, ms_elapsed, tool_name,
first 200 chars of error text. Rotates weekly via ISO week. `readSubagent
AuditForJob` is the readback path for `gbrain agent logs` — scans the
current + prior week file so job boundaries across weeks still resolve.
`GBRAIN_AUDIT_DIR` overrides the default ~/.gbrain/audit/ for container
deploys.

transcript.ts — renders subagent_messages + subagent_tool_executions to
markdown. Message order is authoritative; tool rows splice under their
owning assistant tool_use by tool_use_id. Handles text, tool_use (with
pending / complete / failed execution rows), tool_result (skipped if
we already rendered the owning tool_use — avoids double-printing), and
unknown block types (fenced JSON dump for diagnostics). Output is
UTF-8-safe truncated at maxOutputBytes.

21 unit tests: ISO week filename rotation (incl. 2027-01-01 → W53-2026
boundary), submission + heartbeat write shapes, 200-char error cap, best-
effort write failure doesn't throw, readback filters by job_id and
sinceIso. Transcript: empty input, ordering, token line, tool_use +
complete/failed/pending execution rendering, truncation, unknown-block
diagnostic dump.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-20 21:24:06 +08:00

163 lines
6.5 KiB
TypeScript

/**
* transcript renderer tests. Uses PGLite in-memory to round-trip messages +
* tool executions through the actual schema so the loader path is exercised.
*/
import { describe, test, expect, beforeAll, afterAll, beforeEach } from 'bun:test';
import { PGLiteEngine } from '../src/core/pglite-engine.ts';
import { MinionQueue } from '../src/core/minions/queue.ts';
import {
loadTranscriptRows,
renderTranscript,
} from '../src/core/minions/transcript.ts';
import type { ContentBlock } from '../src/core/minions/types.ts';
let engine: PGLiteEngine;
let queue: MinionQueue;
let jobId: number;
beforeAll(async () => {
engine = new PGLiteEngine();
await engine.connect({ databaseUrl: '' });
await engine.initSchema();
queue = new MinionQueue(engine);
});
afterAll(async () => {
await engine.disconnect();
});
beforeEach(async () => {
await engine.executeRaw('DELETE FROM subagent_messages');
await engine.executeRaw('DELETE FROM subagent_tool_executions');
await engine.executeRaw('DELETE FROM minion_jobs');
const j = await queue.add('subagent', { prompt: 'hi' });
jobId = j.id;
});
async function insertMessage(
idx: number,
role: 'user' | 'assistant',
blocks: ContentBlock[],
tokens: { in?: number; out?: number; cache_read?: number; cache_create?: number } = {},
model = 'claude-sonnet-4-6',
) {
await engine.executeRaw(
`INSERT INTO subagent_messages (job_id, message_idx, role, content_blocks, tokens_in, tokens_out, tokens_cache_read, tokens_cache_create, model)
VALUES ($1, $2, $3, $4::jsonb, $5, $6, $7, $8, $9)`,
[jobId, idx, role, JSON.stringify(blocks), tokens.in ?? null, tokens.out ?? null, tokens.cache_read ?? null, tokens.cache_create ?? null, model],
);
}
async function insertTool(
idx: number,
toolUseId: string,
toolName: string,
input: unknown,
status: 'pending' | 'complete' | 'failed',
output: unknown = null,
error: string | null = null,
) {
await engine.executeRaw(
`INSERT INTO subagent_tool_executions (job_id, message_idx, tool_use_id, tool_name, input, status, output, error)
VALUES ($1, $2, $3, $4, $5::jsonb, $6, $7::jsonb, $8)`,
[jobId, idx, toolUseId, toolName, JSON.stringify(input), status, output == null ? null : JSON.stringify(output), error],
);
}
describe('loadTranscriptRows', () => {
test('empty job returns empty arrays', async () => {
const { messages, tools } = await loadTranscriptRows(engine, jobId);
expect(messages).toEqual([]);
expect(tools).toEqual([]);
});
test('returns messages in message_idx order', async () => {
await insertMessage(1, 'assistant', [{ type: 'text', text: 'second' }]);
await insertMessage(0, 'user', [{ type: 'text', text: 'first' }]);
const { messages } = await loadTranscriptRows(engine, jobId);
expect(messages.map(m => m.message_idx)).toEqual([0, 1]);
});
test('parses content_blocks from JSONB', async () => {
const block: ContentBlock = { type: 'tool_use', id: 'tu_1', name: 'brain_search', input: { q: 'x' } };
await insertMessage(0, 'assistant', [block]);
const { messages } = await loadTranscriptRows(engine, jobId);
expect(messages[0]!.content_blocks[0]!.type).toBe('tool_use');
});
});
describe('renderTranscript', () => {
test('empty messages produce a "no messages" placeholder', () => {
const md = renderTranscript([], []);
expect(md).toContain('# Subagent transcript');
expect(md).toContain('_(no messages)_');
});
test('renders text content under role headers', async () => {
await insertMessage(0, 'user', [{ type: 'text', text: 'hello' }]);
await insertMessage(1, 'assistant', [{ type: 'text', text: 'hi back' }], { in: 5, out: 3 });
const { messages, tools } = await loadTranscriptRows(engine, jobId);
const md = renderTranscript(messages, tools);
expect(md).toContain('## Message 0 — user');
expect(md).toContain('hello');
expect(md).toContain('## Message 1 — assistant');
expect(md).toContain('hi back');
expect(md).toContain('tokens:');
expect(md).toContain('in=5');
});
test('renders tool_use with matching execution row', async () => {
await insertMessage(0, 'assistant', [
{ type: 'tool_use', id: 'tu_42', name: 'brain_get_page', input: { slug: 'foo' } },
]);
await insertTool(0, 'tu_42', 'brain_get_page', { slug: 'foo' }, 'complete', { title: 'Foo' });
const { messages, tools } = await loadTranscriptRows(engine, jobId);
const md = renderTranscript(messages, tools);
expect(md).toContain('**tool_use** `brain_get_page`');
expect(md).toContain('status: **complete**');
expect(md).toContain('"title": "Foo"');
});
test('renders tool_use with failed execution row shows error', async () => {
await insertMessage(0, 'assistant', [
{ type: 'tool_use', id: 'tu_43', name: 'brain_put_page', input: { slug: 'bad' } },
]);
await insertTool(0, 'tu_43', 'brain_put_page', { slug: 'bad' }, 'failed', null, 'permission_denied');
const { messages, tools } = await loadTranscriptRows(engine, jobId);
const md = renderTranscript(messages, tools);
expect(md).toContain('status: **failed**');
expect(md).toContain('permission_denied');
});
test('pending tool execution is shown as pending', async () => {
await insertMessage(0, 'assistant', [
{ type: 'tool_use', id: 'tu_44', name: 'brain_search', input: { q: 'x' } },
]);
await insertTool(0, 'tu_44', 'brain_search', { q: 'x' }, 'pending');
const { messages, tools } = await loadTranscriptRows(engine, jobId);
const md = renderTranscript(messages, tools);
expect(md).toContain('pending (no resolution recorded yet)');
});
test('truncates huge tool outputs per maxOutputBytes', async () => {
await insertMessage(0, 'assistant', [
{ type: 'tool_use', id: 'tu_big', name: 'brain_search', input: {} },
]);
const huge = 'x'.repeat(8000);
await insertTool(0, 'tu_big', 'brain_search', {}, 'complete', { body: huge });
const { messages, tools } = await loadTranscriptRows(engine, jobId);
const md = renderTranscript(messages, tools, { maxOutputBytes: 1024 });
expect(md).toContain('[truncated at 1024 bytes]');
expect(md.length).toBeLessThan(huge.length);
});
test('unknown block types fall through to a JSON dump', async () => {
await insertMessage(0, 'assistant', [{ type: 'some_future_block_type', extra: 42 } as any]);
const { messages, tools } = await loadTranscriptRows(engine, jobId);
const md = renderTranscript(messages, tools);
expect(md).toContain('**some_future_block_type**');
expect(md).toContain('"extra": 42');
});
});