@pi-unipi/background-tasks 2.6.2 → 2.6.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (59) hide show
  1. package/package.json +3 -2
  2. package/src/tools.ts +3 -1
  3. package/src/__tests__/anthropic-attribution.test.ts +0 -195
  4. package/src/__tests__/config.test.ts +0 -137
  5. package/src/__tests__/core.test.ts +0 -493
  6. package/src/__tests__/delegate-artifacts.test.ts +0 -528
  7. package/src/__tests__/delegate-budget.test.ts +0 -456
  8. package/src/__tests__/delegate-launch.test.ts +0 -676
  9. package/src/__tests__/delegate-result-package.test.ts +0 -350
  10. package/src/__tests__/delegate-seed.test.ts +0 -392
  11. package/src/__tests__/durable-fs.test.ts +0 -559
  12. package/src/__tests__/extension-api.test.ts +0 -579
  13. package/src/__tests__/fusion-artifacts.test.ts +0 -1039
  14. package/src/__tests__/fusion-budget.test.ts +0 -1356
  15. package/src/__tests__/fusion-claude-cache.test.ts +0 -320
  16. package/src/__tests__/fusion-config.test.ts +0 -335
  17. package/src/__tests__/fusion-context-prompts.test.ts +0 -670
  18. package/src/__tests__/fusion-evaluation.test.ts +0 -315
  19. package/src/__tests__/fusion-extraction-equivalence.test.ts +0 -58
  20. package/src/__tests__/fusion-golden-bytes.test.ts +0 -35
  21. package/src/__tests__/fusion-high-cardinality.test.ts +0 -192
  22. package/src/__tests__/fusion-model-selector.test.ts +0 -205
  23. package/src/__tests__/fusion-orchestrator.test.ts +0 -1194
  24. package/src/__tests__/fusion-rpc.test.ts +0 -369
  25. package/src/__tests__/fusion-sdk.test.ts +0 -1226
  26. package/src/__tests__/fusion-v5-core.test.ts +0 -219
  27. package/src/__tests__/fusion-validate-orchestrator.test.ts +0 -240
  28. package/src/__tests__/fusion-web-fetch.test.ts +0 -485
  29. package/src/__tests__/fusion-workflows.test.ts +0 -59
  30. package/src/__tests__/helpers/delegate-deterministic-seed.ts +0 -109
  31. package/src/__tests__/helpers/delegate-seed-subprocess.ts +0 -10
  32. package/src/__tests__/helpers/fusion-canonical-subprocess.ts +0 -21
  33. package/src/__tests__/helpers/fusion-canonical.ts +0 -140
  34. package/src/__tests__/helpers/fusion-fake-pi.ts +0 -279
  35. package/src/__tests__/helpers/fusion-golden-corpus.ts +0 -500
  36. package/src/__tests__/helpers/fusion-high-cardinality.ts +0 -140
  37. package/src/__tests__/helpers/normalize.ts +0 -22
  38. package/src/__tests__/helpers/pi-hook-contract-evidence.json +0 -18
  39. package/src/__tests__/pi-launch.test.ts +0 -202
  40. package/src/__tests__/registry.test.ts +0 -1580
  41. package/src/__tests__/scripted-provider/delegate-ambient-provider.test.ts +0 -130
  42. package/src/__tests__/scripted-provider/delegate-child-guard.test.ts +0 -631
  43. package/src/__tests__/scripted-provider/delegate-guard-provider.ts +0 -403
  44. package/src/__tests__/scripted-provider/follow-up.test.ts +0 -448
  45. package/src/__tests__/scripted-provider/fusion-output-recovery.test.ts +0 -132
  46. package/src/__tests__/scripted-provider/fusion-reason.test.ts +0 -310
  47. package/src/__tests__/scripted-provider/fusion-runtime-guard.test.ts +0 -163
  48. package/src/__tests__/scripted-provider/hook-contract-provider.ts +0 -179
  49. package/src/__tests__/scripted-provider/hook-probe-a.ts +0 -3
  50. package/src/__tests__/scripted-provider/hook-probe-b.ts +0 -3
  51. package/src/__tests__/scripted-provider/hook-probe-extension.ts +0 -126
  52. package/src/__tests__/scripted-provider/output-recovery-provider.ts +0 -153
  53. package/src/__tests__/scripted-provider/pi-hook-contract-evidence.json +0 -18
  54. package/src/__tests__/scripted-provider/pi-hook-contract.test.ts +0 -477
  55. package/src/__tests__/scripted-provider/runtime-guard-probe.ts +0 -28
  56. package/src/__tests__/scripted-provider/runtime-guard-provider.ts +0 -49
  57. package/src/__tests__/scripted-provider/scripted-provider-extension.ts +0 -408
  58. package/src/__tests__/task-manager.test.ts +0 -479
  59. package/src/__tests__/windows-taskkill.test.ts +0 -161
@@ -1,350 +0,0 @@
1
- import { describe, it } from 'node:test';
2
- import assert from 'node:assert/strict';
3
- import { createHash } from 'node:crypto';
4
- import {
5
- assertWellFormedUtf8,
6
- buildDelegateResultPackage,
7
- serializeDelegateResultPackage,
8
- verifyDelegateResultPackage,
9
- } from '../delegate/result-package.js';
10
- import { DelegateError, type DelegateUsageReport } from '../delegate/types.js';
11
-
12
- const EXPECTED = {
13
- taskId: 'd0123456789abcdef0123456789abcdef',
14
- launchNonce: 'ffeeddccbbaa99887766554433221100',
15
- seedSha256: 'a'.repeat(64),
16
- route: { provider: 'anthropic', model: 'claude-test' },
17
- };
18
-
19
- const OBSERVED_USAGE: DelegateUsageReport = {
20
- status: 'observed',
21
- usage: {
22
- input: 100,
23
- output: 50,
24
- cacheRead: 10,
25
- cacheWrite: 5,
26
- totalTokens: 165,
27
- cost: { input: 0.1, output: 0.2, cacheRead: 0.01, cacheWrite: 0.02, total: 0.33 },
28
- },
29
- };
30
-
31
- function pkg(answerBlocks: readonly string[], overrides: Record<string, unknown> = {}) {
32
- const built = buildDelegateResultPackage({
33
- taskId: EXPECTED.taskId,
34
- launchNonce: EXPECTED.launchNonce,
35
- seedSha256: EXPECTED.seedSha256,
36
- directiveSha256: 'b'.repeat(64),
37
- route: EXPECTED.route,
38
- routeAttestations: [
39
- { provider: 'anthropic', model: 'claude-test', stop_reason: 'stop' },
40
- ],
41
- stopReason: 'stop',
42
- turns: 3,
43
- toolCalls: 7,
44
- usage: OBSERVED_USAGE,
45
- answerBlocks,
46
- spilledArtifacts: [],
47
- });
48
- return { ...built, ...overrides };
49
- }
50
-
51
- function roundTrip(answerBlocks: readonly string[]) {
52
- return verifyDelegateResultPackage(
53
- serializeDelegateResultPackage(pkg(answerBlocks)),
54
- EXPECTED,
55
- );
56
- }
57
-
58
- function tamper(mutate: (record: Record<string, unknown>) => void, blocks = ['answer text']) {
59
- const record = JSON.parse(serializeDelegateResultPackage(pkg(blocks))) as Record<string, unknown>;
60
- mutate(record);
61
- return JSON.stringify(record);
62
- }
63
-
64
- void describe('delegate result package', () => {
65
- void it('round-trips a single-block answer with verified hashes', () => {
66
- const verified = roundTrip(['the delegate answer']);
67
- assert.equal(verified.answer, 'the delegate answer');
68
- assert.equal(verified.package.turns, 3);
69
- assert.equal(verified.package.tool_calls, 7);
70
- assert.equal(verified.package.usage.status, 'observed');
71
- });
72
-
73
- void it('round-trips a multi-block answer preserving exact concatenation', () => {
74
- const verified = roundTrip(['first part ', 'second part ', 'third part']);
75
- assert.equal(verified.answer, 'first part second part third part');
76
- assert.equal(verified.package.answer.blocks.length, 3);
77
- });
78
-
79
- void it('preserves U+2028, U+2029, emoji, and unnormalized sequences exactly', () => {
80
- const tricky = 'line\u2028sep\u2029para 👩‍👩‍👧‍👦 e\u0301 combining \u200D zwj';
81
- const verified = roundTrip([tricky]);
82
- assert.equal(verified.answer, tricky);
83
- assert.equal(verified.answerBytes.length, Buffer.byteLength(tricky, 'utf8'));
84
- });
85
-
86
- void it('rejects a lone surrogate rather than substituting U+FFFD', () => {
87
- assert.throws(
88
- () => assertWellFormedUtf8('broken \uD800 surrogate', 'answer'),
89
- (error: unknown) =>
90
- error instanceof DelegateError && error.code === 'child_result_encoding_invalid',
91
- );
92
- assert.throws(
93
- () => buildDelegateResultPackage({
94
- taskId: EXPECTED.taskId,
95
- launchNonce: EXPECTED.launchNonce,
96
- seedSha256: EXPECTED.seedSha256,
97
- directiveSha256: 'b'.repeat(64),
98
- route: EXPECTED.route,
99
- routeAttestations: [{ provider: 'anthropic', model: 'claude-test', stop_reason: 'stop' }],
100
- stopReason: 'stop',
101
- turns: 1,
102
- toolCalls: 0,
103
- usage: OBSERVED_USAGE,
104
- answerBlocks: ['\uD800'],
105
- spilledArtifacts: [],
106
- }),
107
- (error: unknown) =>
108
- error instanceof DelegateError && error.code === 'child_result_encoding_invalid',
109
- );
110
- });
111
-
112
- void it('detects a corrupted answer block and never returns a prefix', () => {
113
- const corrupted = tamper((record) => {
114
- const answer = record['answer'];
115
- assert.ok(typeof answer === 'object' && answer !== null);
116
- const blocks = Reflect.get(answer, 'blocks');
117
- assert.ok(Array.isArray(blocks));
118
- const first = blocks[0];
119
- assert.ok(typeof first === 'object' && first !== null);
120
- Reflect.set(first, 'data_base64', Buffer.from('tampered text', 'utf8').toString('base64'));
121
- });
122
- assert.throws(
123
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
124
- (error: unknown) => error instanceof DelegateError && error.code === 'answer_hash_mismatch',
125
- );
126
- });
127
-
128
- void it('detects an aggregate hash mismatch even when every block hash is valid', () => {
129
- const corrupted = tamper((record) => {
130
- const answer = record['answer'];
131
- assert.ok(typeof answer === 'object' && answer !== null);
132
- Reflect.set(answer, 'sha256', 'c'.repeat(64));
133
- });
134
- assert.throws(
135
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
136
- (error: unknown) => error instanceof DelegateError && error.code === 'answer_hash_mismatch',
137
- );
138
- });
139
-
140
- void it('detects a declared length that disagrees with the carried bytes', () => {
141
- const corrupted = tamper((record) => {
142
- const answer = record['answer'];
143
- assert.ok(typeof answer === 'object' && answer !== null);
144
- Reflect.set(answer, 'byte_length', 999_999);
145
- });
146
- assert.throws(
147
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
148
- (error: unknown) => error instanceof DelegateError && error.code === 'answer_hash_mismatch',
149
- );
150
- });
151
-
152
- void it('rejects non-strict base64 rather than absorbing it', () => {
153
- const corrupted = tamper((record) => {
154
- const answer = record['answer'];
155
- assert.ok(typeof answer === 'object' && answer !== null);
156
- const blocks = Reflect.get(answer, 'blocks');
157
- assert.ok(Array.isArray(blocks));
158
- const first = blocks[0];
159
- assert.ok(typeof first === 'object' && first !== null);
160
- const encoded: unknown = Reflect.get(first, 'data_base64');
161
- assert.ok(typeof encoded === 'string');
162
- Reflect.set(first, 'data_base64', `${encoded}===extra`);
163
- });
164
- assert.throws(
165
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
166
- (error: unknown) => error instanceof DelegateError && error.code === 'child_result_invalid',
167
- );
168
- });
169
-
170
- void it('rejects a package produced from a different seed', () => {
171
- assert.throws(
172
- () =>
173
- verifyDelegateResultPackage(serializeDelegateResultPackage(pkg(['x'])), {
174
- ...EXPECTED,
175
- seedSha256: 'd'.repeat(64),
176
- }),
177
- (error: unknown) => error instanceof DelegateError && error.code === 'seed_hash_mismatch',
178
- );
179
- });
180
-
181
- void it('rejects a package carrying a foreign task identity or launch nonce', () => {
182
- const serialized = serializeDelegateResultPackage(pkg(['x']));
183
- assert.throws(
184
- () => verifyDelegateResultPackage(serialized, { ...EXPECTED, taskId: 'dfff' }),
185
- (error: unknown) => error instanceof DelegateError && error.code === 'child_result_invalid',
186
- );
187
- assert.throws(
188
- () => verifyDelegateResultPackage(serialized, { ...EXPECTED, launchNonce: 'other' }),
189
- (error: unknown) => error instanceof DelegateError && error.code === 'child_result_invalid',
190
- );
191
- });
192
-
193
- void it('rejects a route that does not match the pinned route', () => {
194
- const corrupted = tamper((record) => {
195
- Reflect.set(record, 'route', { provider: 'openai', model: 'gpt-test' });
196
- });
197
- assert.throws(
198
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
199
- (error: unknown) => error instanceof DelegateError && error.code === 'route_mismatch',
200
- );
201
- });
202
-
203
- void it('rejects an attestation from a substituted route', () => {
204
- const corrupted = tamper((record) => {
205
- Reflect.set(record, 'route_attestations', [
206
- { provider: 'anthropic', model: 'claude-test', stop_reason: 'stop' },
207
- { provider: 'openai', model: 'gpt-test', stop_reason: 'stop' },
208
- ]);
209
- });
210
- assert.throws(
211
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
212
- (error: unknown) => error instanceof DelegateError && error.code === 'route_mismatch',
213
- );
214
- });
215
-
216
- void it('refuses a package with no route attestation at all', () => {
217
- const corrupted = tamper((record) => {
218
- Reflect.set(record, 'route_attestations', []);
219
- });
220
- assert.throws(
221
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
222
- (error: unknown) =>
223
- error instanceof DelegateError && error.code === 'route_attestation_missing',
224
- );
225
- });
226
-
227
- void it('reports unavailable usage explicitly rather than as zero', () => {
228
- const built = buildDelegateResultPackage({
229
- taskId: EXPECTED.taskId,
230
- launchNonce: EXPECTED.launchNonce,
231
- seedSha256: EXPECTED.seedSha256,
232
- directiveSha256: 'b'.repeat(64),
233
- route: EXPECTED.route,
234
- routeAttestations: [{ provider: 'anthropic', model: 'claude-test', stop_reason: 'stop' }],
235
- stopReason: 'stop',
236
- turns: 1,
237
- toolCalls: 0,
238
- usage: { status: 'unavailable', reason: 'the provider reported no usage' },
239
- answerBlocks: ['answer'],
240
- spilledArtifacts: [],
241
- });
242
- const verified = verifyDelegateResultPackage(
243
- serializeDelegateResultPackage(built),
244
- EXPECTED,
245
- );
246
- assert.equal(verified.package.usage.status, 'unavailable');
247
- const serialized = serializeDelegateResultPackage(built);
248
- assert.ok(!/"totalTokens":0/.test(serialized), 'must not synthesize zero usage');
249
- });
250
-
251
- void it('rejects a usage record missing cost components', () => {
252
- const corrupted = tamper((record) => {
253
- Reflect.set(record, 'usage', {
254
- status: 'observed',
255
- usage: { input: 1, output: 1, cacheRead: 0, cacheWrite: 0, totalTokens: 2 },
256
- });
257
- });
258
- assert.throws(
259
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
260
- (error: unknown) => error instanceof DelegateError && error.code === 'child_result_invalid',
261
- );
262
- });
263
-
264
- void it('rejects malformed JSON and a wrong schema version', () => {
265
- assert.throws(
266
- () => verifyDelegateResultPackage('{not json', EXPECTED),
267
- (error: unknown) => error instanceof DelegateError && error.code === 'child_result_invalid',
268
- );
269
- const corrupted = tamper((record) => {
270
- Reflect.set(record, 'schema_version', 'pi-background-tasks.delegate-result.v99');
271
- });
272
- assert.throws(
273
- () => verifyDelegateResultPackage(corrupted, EXPECTED),
274
- (error: unknown) => error instanceof DelegateError && error.code === 'child_result_invalid',
275
- );
276
- });
277
-
278
- void it('accepts a historical v1 spill receipt without content_format', () => {
279
- const built = buildDelegateResultPackage({
280
- taskId: EXPECTED.taskId,
281
- launchNonce: EXPECTED.launchNonce,
282
- seedSha256: EXPECTED.seedSha256,
283
- directiveSha256: 'b'.repeat(64),
284
- route: EXPECTED.route,
285
- routeAttestations: [{ provider: 'anthropic', model: 'claude-test', stop_reason: 'stop' }],
286
- stopReason: 'stop',
287
- turns: 1,
288
- toolCalls: 1,
289
- usage: OBSERVED_USAGE,
290
- answerBlocks: ['answer'],
291
- spilledArtifacts: [
292
- {
293
- schema_version: 'pi-background-tasks.delegate-receipt.v1',
294
- artifact: 'spill/historical.bin',
295
- tool_name: 'read',
296
- tool_call_id: 'historical',
297
- turn_sequence: 1,
298
- source_call_index: 0,
299
- byte_length: 7,
300
- sha256: createHash('sha256').update('payload').digest('hex'),
301
- },
302
- ],
303
- });
304
- const verified = verifyDelegateResultPackage(
305
- serializeDelegateResultPackage(built),
306
- EXPECTED,
307
- );
308
- assert.equal(verified.package.spilled_artifacts[0]?.content_format, undefined);
309
- });
310
-
311
- void it('preserves spill receipts with their call coordinates', () => {
312
- const built = buildDelegateResultPackage({
313
- taskId: EXPECTED.taskId,
314
- launchNonce: EXPECTED.launchNonce,
315
- seedSha256: EXPECTED.seedSha256,
316
- directiveSha256: 'b'.repeat(64),
317
- route: EXPECTED.route,
318
- routeAttestations: [{ provider: 'anthropic', model: 'claude-test', stop_reason: 'stop' }],
319
- stopReason: 'stop',
320
- turns: 2,
321
- toolCalls: 2,
322
- usage: OBSERVED_USAGE,
323
- answerBlocks: ['answer'],
324
- spilledArtifacts: [
325
- {
326
- schema_version: 'pi-background-tasks.delegate-receipt.v1',
327
- artifact: 'spill/t0001-c0000-abc.bin',
328
- tool_name: 'read',
329
- tool_call_id: 'abc',
330
- turn_sequence: 1,
331
- source_call_index: 0,
332
- byte_length: 2_097_152,
333
- sha256: createHash('sha256').update('payload').digest('hex'),
334
- content_format: 'single_text_utf8',
335
- },
336
- ],
337
- });
338
- const verified = verifyDelegateResultPackage(
339
- serializeDelegateResultPackage(built),
340
- EXPECTED,
341
- );
342
- assert.equal(verified.package.spilled_artifacts.length, 1);
343
- const receipt = verified.package.spilled_artifacts[0];
344
- assert.ok(receipt);
345
- assert.equal(receipt.turn_sequence, 1);
346
- assert.equal(receipt.source_call_index, 0);
347
- assert.equal(receipt.byte_length, 2_097_152);
348
- assert.equal(receipt.content_format, 'single_text_utf8');
349
- });
350
- });