@claude-flow/cli 3.11.0 → 3.12.0

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 (154) hide show
  1. package/dist/src/commands/doctor.d.ts.map +1 -1
  2. package/dist/src/commands/doctor.js +175 -1
  3. package/dist/src/commands/doctor.js.map +1 -1
  4. package/dist/src/commands/eject.d.ts +33 -0
  5. package/dist/src/commands/eject.d.ts.map +1 -0
  6. package/dist/src/commands/eject.js +195 -0
  7. package/dist/src/commands/eject.js.map +1 -0
  8. package/dist/src/commands/hooks.d.ts.map +1 -1
  9. package/dist/src/commands/hooks.js +2 -1
  10. package/dist/src/commands/hooks.js.map +1 -1
  11. package/dist/src/commands/index.d.ts.map +1 -1
  12. package/dist/src/commands/index.js +4 -0
  13. package/dist/src/commands/index.js.map +1 -1
  14. package/dist/src/commands/init.d.ts.map +1 -1
  15. package/dist/src/commands/init.js +3 -0
  16. package/dist/src/commands/init.js.map +1 -1
  17. package/dist/src/commands/metaharness.d.ts +39 -0
  18. package/dist/src/commands/metaharness.d.ts.map +1 -0
  19. package/dist/src/commands/metaharness.js +194 -0
  20. package/dist/src/commands/metaharness.js.map +1 -0
  21. package/dist/src/mcp-client.d.ts.map +1 -1
  22. package/dist/src/mcp-client.js +4 -0
  23. package/dist/src/mcp-client.js.map +1 -1
  24. package/dist/src/mcp-tools/index.d.ts +1 -0
  25. package/dist/src/mcp-tools/index.d.ts.map +1 -1
  26. package/dist/src/mcp-tools/index.js +2 -0
  27. package/dist/src/mcp-tools/index.js.map +1 -1
  28. package/dist/src/mcp-tools/metaharness-tools.d.ts +31 -0
  29. package/dist/src/mcp-tools/metaharness-tools.d.ts.map +1 -0
  30. package/dist/src/mcp-tools/metaharness-tools.js +372 -0
  31. package/dist/src/mcp-tools/metaharness-tools.js.map +1 -0
  32. package/dist/src/plugins/store/discovery.d.ts.map +1 -1
  33. package/dist/src/plugins/store/discovery.js +3 -0
  34. package/dist/src/plugins/store/discovery.js.map +1 -1
  35. package/dist/src/plugins/store/types.d.ts +6 -1
  36. package/dist/src/plugins/store/types.d.ts.map +1 -1
  37. package/dist/src/ruvector/model-router.d.ts.map +1 -1
  38. package/dist/src/ruvector/model-router.js +37 -0
  39. package/dist/src/ruvector/model-router.js.map +1 -1
  40. package/dist/src/ruvector/router-parallel-recorder.d.ts +127 -0
  41. package/dist/src/ruvector/router-parallel-recorder.d.ts.map +1 -0
  42. package/dist/src/ruvector/router-parallel-recorder.js +183 -0
  43. package/dist/src/ruvector/router-parallel-recorder.js.map +1 -0
  44. package/dist/tsconfig.tsbuildinfo +1 -1
  45. package/package.json +11 -9
  46. package/dist/src/benchmarks/capability-tasks.d.ts +0 -126
  47. package/dist/src/benchmarks/capability-tasks.d.ts.map +0 -1
  48. package/dist/src/benchmarks/capability-tasks.js +0 -147
  49. package/dist/src/benchmarks/capability-tasks.js.map +0 -1
  50. package/dist/src/benchmarks/gaia-agent-contrastive.d.ts +0 -53
  51. package/dist/src/benchmarks/gaia-agent-contrastive.d.ts.map +0 -1
  52. package/dist/src/benchmarks/gaia-agent-contrastive.js +0 -177
  53. package/dist/src/benchmarks/gaia-agent-contrastive.js.map +0 -1
  54. package/dist/src/benchmarks/gaia-agent-gemini.d.ts +0 -51
  55. package/dist/src/benchmarks/gaia-agent-gemini.d.ts.map +0 -1
  56. package/dist/src/benchmarks/gaia-agent-gemini.js +0 -352
  57. package/dist/src/benchmarks/gaia-agent-gemini.js.map +0 -1
  58. package/dist/src/benchmarks/gaia-claude-p.d.ts +0 -105
  59. package/dist/src/benchmarks/gaia-claude-p.d.ts.map +0 -1
  60. package/dist/src/benchmarks/gaia-claude-p.js +0 -269
  61. package/dist/src/benchmarks/gaia-claude-p.js.map +0 -1
  62. package/dist/src/benchmarks/gaia-claude-p.smoke.d.ts +0 -20
  63. package/dist/src/benchmarks/gaia-claude-p.smoke.d.ts.map +0 -1
  64. package/dist/src/benchmarks/gaia-claude-p.smoke.js +0 -168
  65. package/dist/src/benchmarks/gaia-claude-p.smoke.js.map +0 -1
  66. package/dist/src/benchmarks/gaia-codeagent-runner.py +0 -556
  67. package/dist/src/benchmarks/gaia-codeagent.d.ts +0 -129
  68. package/dist/src/benchmarks/gaia-codeagent.d.ts.map +0 -1
  69. package/dist/src/benchmarks/gaia-codeagent.js +0 -590
  70. package/dist/src/benchmarks/gaia-codeagent.js.map +0 -1
  71. package/dist/src/benchmarks/gaia-codeagent.smoke.d.ts +0 -33
  72. package/dist/src/benchmarks/gaia-codeagent.smoke.d.ts.map +0 -1
  73. package/dist/src/benchmarks/gaia-codeagent.smoke.js +0 -241
  74. package/dist/src/benchmarks/gaia-codeagent.smoke.js.map +0 -1
  75. package/dist/src/benchmarks/gaia-dag-pilot.d.ts +0 -22
  76. package/dist/src/benchmarks/gaia-dag-pilot.d.ts.map +0 -1
  77. package/dist/src/benchmarks/gaia-dag-pilot.js +0 -158
  78. package/dist/src/benchmarks/gaia-dag-pilot.js.map +0 -1
  79. package/dist/src/benchmarks/gaia-dag.d.ts +0 -110
  80. package/dist/src/benchmarks/gaia-dag.d.ts.map +0 -1
  81. package/dist/src/benchmarks/gaia-dag.js +0 -658
  82. package/dist/src/benchmarks/gaia-dag.js.map +0 -1
  83. package/dist/src/benchmarks/gaia-dag.smoke.d.ts +0 -21
  84. package/dist/src/benchmarks/gaia-dag.smoke.d.ts.map +0 -1
  85. package/dist/src/benchmarks/gaia-dag.smoke.js +0 -290
  86. package/dist/src/benchmarks/gaia-dag.smoke.js.map +0 -1
  87. package/dist/src/benchmarks/gaia-ensemble-pilot.d.ts +0 -17
  88. package/dist/src/benchmarks/gaia-ensemble-pilot.d.ts.map +0 -1
  89. package/dist/src/benchmarks/gaia-ensemble-pilot.js +0 -106
  90. package/dist/src/benchmarks/gaia-ensemble-pilot.js.map +0 -1
  91. package/dist/src/benchmarks/gaia-ensemble.d.ts +0 -103
  92. package/dist/src/benchmarks/gaia-ensemble.d.ts.map +0 -1
  93. package/dist/src/benchmarks/gaia-ensemble.js +0 -462
  94. package/dist/src/benchmarks/gaia-ensemble.js.map +0 -1
  95. package/dist/src/benchmarks/gaia-gemini-pilot.d.ts +0 -17
  96. package/dist/src/benchmarks/gaia-gemini-pilot.d.ts.map +0 -1
  97. package/dist/src/benchmarks/gaia-gemini-pilot.js +0 -168
  98. package/dist/src/benchmarks/gaia-gemini-pilot.js.map +0 -1
  99. package/dist/src/benchmarks/gaia-mode-router.d.ts +0 -67
  100. package/dist/src/benchmarks/gaia-mode-router.d.ts.map +0 -1
  101. package/dist/src/benchmarks/gaia-mode-router.js +0 -126
  102. package/dist/src/benchmarks/gaia-mode-router.js.map +0 -1
  103. package/dist/src/benchmarks/gaia-mode-router.smoke.d.ts +0 -15
  104. package/dist/src/benchmarks/gaia-mode-router.smoke.d.ts.map +0 -1
  105. package/dist/src/benchmarks/gaia-mode-router.smoke.js +0 -131
  106. package/dist/src/benchmarks/gaia-mode-router.smoke.js.map +0 -1
  107. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.d.ts +0 -24
  108. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.d.ts.map +0 -1
  109. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.js +0 -276
  110. package/dist/src/benchmarks/gaia-tools/grounded_query.smoke.js.map +0 -1
  111. package/dist/src/benchmarks/gaia-tools/image_describe.d.ts +0 -45
  112. package/dist/src/benchmarks/gaia-tools/image_describe.d.ts.map +0 -1
  113. package/dist/src/benchmarks/gaia-tools/image_describe.js +0 -222
  114. package/dist/src/benchmarks/gaia-tools/image_describe.js.map +0 -1
  115. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.d.ts +0 -24
  116. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.d.ts.map +0 -1
  117. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.js +0 -169
  118. package/dist/src/benchmarks/gaia-tools/image_describe.smoke.js.map +0 -1
  119. package/dist/src/benchmarks/gaia-tools/pdf_read.d.ts +0 -27
  120. package/dist/src/benchmarks/gaia-tools/pdf_read.d.ts.map +0 -1
  121. package/dist/src/benchmarks/gaia-tools/pdf_read.js +0 -226
  122. package/dist/src/benchmarks/gaia-tools/pdf_read.js.map +0 -1
  123. package/dist/src/benchmarks/gaia-tools/python_exec.d.ts +0 -31
  124. package/dist/src/benchmarks/gaia-tools/python_exec.d.ts.map +0 -1
  125. package/dist/src/benchmarks/gaia-tools/python_exec.js +0 -94
  126. package/dist/src/benchmarks/gaia-tools/python_exec.js.map +0 -1
  127. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.d.ts +0 -13
  128. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.d.ts.map +0 -1
  129. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.js +0 -98
  130. package/dist/src/benchmarks/gaia-tools/python_exec.smoke.js.map +0 -1
  131. package/dist/src/benchmarks/gaia-tools/visit_webpage.d.ts +0 -52
  132. package/dist/src/benchmarks/gaia-tools/visit_webpage.d.ts.map +0 -1
  133. package/dist/src/benchmarks/gaia-tools/visit_webpage.js +0 -304
  134. package/dist/src/benchmarks/gaia-tools/visit_webpage.js.map +0 -1
  135. package/dist/src/benchmarks/gaia-tools/web_browse.d.ts +0 -61
  136. package/dist/src/benchmarks/gaia-tools/web_browse.d.ts.map +0 -1
  137. package/dist/src/benchmarks/gaia-tools/web_browse.js +0 -217
  138. package/dist/src/benchmarks/gaia-tools/web_browse.js.map +0 -1
  139. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.d.ts +0 -26
  140. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.d.ts.map +0 -1
  141. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.js +0 -152
  142. package/dist/src/benchmarks/gaia-tools/web_browse.smoke.js.map +0 -1
  143. package/dist/src/benchmarks/gaia-tools/web_search.smoke.d.ts +0 -24
  144. package/dist/src/benchmarks/gaia-tools/web_search.smoke.d.ts.map +0 -1
  145. package/dist/src/benchmarks/gaia-tools/web_search.smoke.js +0 -311
  146. package/dist/src/benchmarks/gaia-tools/web_search.smoke.js.map +0 -1
  147. package/dist/src/benchmarks/gaia-visit-webpage.smoke.d.ts +0 -25
  148. package/dist/src/benchmarks/gaia-visit-webpage.smoke.d.ts.map +0 -1
  149. package/dist/src/benchmarks/gaia-visit-webpage.smoke.js +0 -304
  150. package/dist/src/benchmarks/gaia-visit-webpage.smoke.js.map +0 -1
  151. package/dist/src/commands/performance-capability.d.ts +0 -30
  152. package/dist/src/commands/performance-capability.d.ts.map +0 -1
  153. package/dist/src/commands/performance-capability.js +0 -349
  154. package/dist/src/commands/performance-capability.js.map +0 -1
@@ -1,276 +0,0 @@
1
- /**
2
- * Smoke tests for grounded_query.ts — ADR-135 / iter-33
3
- *
4
- * All tests are mock-based using groundedQueryTestHooks for ESM-safe injection.
5
- * NO live API calls are made.
6
- *
7
- * Coverage:
8
- * 1. resolveGoogleAIApiKey — returns key from env var
9
- * 2. resolveGoogleAIApiKey — throws when env var absent and gcloud unavailable
10
- * 3. parseGeminiResponse — well-formed response → correct GroundedQueryResult shape
11
- * 4. parseGeminiResponse — empty grounding metadata → grounded=false, empty sources
12
- * 5. parseGeminiResponse — deduplicates sources by URI
13
- * 6. parseGeminiResponse — cost calculation uses token counts
14
- * 7. GroundedQueryTool.execute — happy path: answer + sources in output
15
- * 8. GroundedQueryTool.execute — HTTP 4xx surfaces useful error message
16
- * 9. GroundedQueryTool.execute — ungrounded response includes note in output
17
- * 10. GroundedQueryTool.execute — rejects empty query
18
- * 11. GroundedQueryTool.execute — max_tokens clamped to 8192
19
- * 12. GroundedQueryTool.execute — uses injected apiKey from constructor
20
- *
21
- * Run with: npx tsx grounded_query.smoke.ts (from gaia-tools directory)
22
- */
23
- import * as assert from 'node:assert/strict';
24
- import { GroundedQueryTool, groundedQueryTestHooks, resolveGoogleAIApiKey, parseGeminiResponse, createGroundedQueryTool, } from './grounded_query.js';
25
- // ---------------------------------------------------------------------------
26
- // Minimal test harness (identical to web_search.smoke.ts — no external deps)
27
- // ---------------------------------------------------------------------------
28
- let passed = 0;
29
- let failed = 0;
30
- async function test(name, fn) {
31
- try {
32
- await fn();
33
- console.log(` PASS ${name}`);
34
- passed++;
35
- }
36
- catch (err) {
37
- console.error(` FAIL ${name}`);
38
- console.error(` ${err.message}`);
39
- failed++;
40
- }
41
- }
42
- function clearHooks() {
43
- delete groundedQueryTestHooks.resolveKey;
44
- delete groundedQueryTestHooks.fetchResponse;
45
- }
46
- // ---------------------------------------------------------------------------
47
- // Shared stub data
48
- // ---------------------------------------------------------------------------
49
- const STUB_RESPONSE_GROUNDED = {
50
- candidates: [
51
- {
52
- content: { parts: [{ text: 'Mercedes Sosa released 5 studio albums in that period.' }] },
53
- groundingMetadata: {
54
- groundingChunks: [
55
- { web: { title: 'Mercedes Sosa discography', uri: 'https://example.com/mercedes-sosa' } },
56
- { web: { title: 'AllMusic Mercedes Sosa', uri: 'https://allmusic.com/sosa' } },
57
- { web: { title: 'Wikipedia Mercedes Sosa', uri: 'https://en.wikipedia.org/wiki/Mercedes_Sosa' } },
58
- { web: { title: 'Duplicate entry', uri: 'https://example.com/mercedes-sosa' } }, // duplicate URI
59
- ],
60
- webSearchQueries: ['Mercedes Sosa studio albums count'],
61
- },
62
- },
63
- ],
64
- usageMetadata: { promptTokenCount: 200, candidatesTokenCount: 50 },
65
- };
66
- const STUB_RESPONSE_UNGROUNDED = {
67
- candidates: [
68
- {
69
- content: { parts: [{ text: 'I cannot verify this without current data.' }] },
70
- // No groundingMetadata
71
- },
72
- ],
73
- usageMetadata: { promptTokenCount: 100, candidatesTokenCount: 30 },
74
- };
75
- const STUB_RESPONSE_EMPTY_GROUNDING = {
76
- candidates: [
77
- {
78
- content: { parts: [{ text: 'Some answer.' }] },
79
- groundingMetadata: {
80
- groundingChunks: [],
81
- webSearchQueries: [],
82
- },
83
- },
84
- ],
85
- usageMetadata: { promptTokenCount: 50, candidatesTokenCount: 20 },
86
- };
87
- // ---------------------------------------------------------------------------
88
- // Test suite
89
- // ---------------------------------------------------------------------------
90
- async function runSuite() {
91
- console.log('\ngrounded_query.smoke.ts — ADR-135 / iter-33 Gemini Grounding\n');
92
- // -------------------------------------------------------------------------
93
- // 1. resolveGoogleAIApiKey — env var happy path
94
- // -------------------------------------------------------------------------
95
- await test('resolveGoogleAIApiKey: returns key from env var', async () => {
96
- const savedKey = process.env['GOOGLE_AI_API_KEY'];
97
- process.env['GOOGLE_AI_API_KEY'] = 'test-gemini-key-123';
98
- try {
99
- const key = await resolveGoogleAIApiKey();
100
- assert.equal(key, 'test-gemini-key-123');
101
- }
102
- finally {
103
- if (savedKey === undefined)
104
- delete process.env['GOOGLE_AI_API_KEY'];
105
- else
106
- process.env['GOOGLE_AI_API_KEY'] = savedKey;
107
- }
108
- });
109
- // -------------------------------------------------------------------------
110
- // 2. resolveGoogleAIApiKey — throws when missing (tested via tool hook path
111
- // to avoid depending on whether GOOGLE_AI_API_KEY is set in CI or dev env)
112
- // -------------------------------------------------------------------------
113
- await test('resolveGoogleAIApiKey: throws with guidance when no key available', async () => {
114
- // Use the resolveKey hook to simulate the case where no key is configured.
115
- // This avoids a brittle env-var delete that breaks when the key IS present.
116
- groundedQueryTestHooks.resolveKey = async () => {
117
- throw new Error('grounded_query: No Google AI API key found.\n' +
118
- 'Set GOOGLE_AI_API_KEY env var, or ensure `gcloud` is authenticated.\n' +
119
- 'Get a free key at: https://aistudio.google.com/apikey');
120
- };
121
- const tool = new GroundedQueryTool();
122
- try {
123
- await assert.rejects(() => tool.execute({ query: 'test missing key' }), /GOOGLE_AI_API_KEY/);
124
- }
125
- finally {
126
- clearHooks();
127
- }
128
- });
129
- // -------------------------------------------------------------------------
130
- // 3. parseGeminiResponse — well-formed grounded response
131
- // -------------------------------------------------------------------------
132
- await test('parseGeminiResponse: maps answer + sources + queries correctly', async () => {
133
- const result = parseGeminiResponse(STUB_RESPONSE_GROUNDED);
134
- assert.equal(result.model, 'gemini-2.5-flash');
135
- assert.ok(result.answer.includes('5 studio albums'));
136
- assert.equal(result.grounded, true);
137
- // 4 chunks but one duplicate URI → 3 unique sources
138
- assert.equal(result.sources.length, 3, `expected 3 unique sources, got ${result.sources.length}`);
139
- assert.equal(result.search_queries_used.length, 1);
140
- assert.equal(result.search_queries_used[0], 'Mercedes Sosa studio albums count');
141
- });
142
- // -------------------------------------------------------------------------
143
- // 4. parseGeminiResponse — empty grounding → grounded=false
144
- // -------------------------------------------------------------------------
145
- await test('parseGeminiResponse: empty grounding metadata → grounded=false', async () => {
146
- const result = parseGeminiResponse(STUB_RESPONSE_EMPTY_GROUNDING);
147
- assert.equal(result.grounded, false);
148
- assert.equal(result.sources.length, 0);
149
- assert.equal(result.search_queries_used.length, 0);
150
- });
151
- // -------------------------------------------------------------------------
152
- // 5. parseGeminiResponse — deduplicates sources by URI (already tested in #3
153
- // but let's verify explicitly)
154
- // -------------------------------------------------------------------------
155
- await test('parseGeminiResponse: deduplicates sources by URI', async () => {
156
- const result = parseGeminiResponse(STUB_RESPONSE_GROUNDED);
157
- const uris = result.sources.map((s) => s.uri);
158
- const uniqueUris = new Set(uris);
159
- assert.equal(uris.length, uniqueUris.size, 'duplicate URIs found in sources');
160
- });
161
- // -------------------------------------------------------------------------
162
- // 6. parseGeminiResponse — cost calculation
163
- // -------------------------------------------------------------------------
164
- await test('parseGeminiResponse: cost_usd calculated from token counts', async () => {
165
- const result = parseGeminiResponse(STUB_RESPONSE_GROUNDED);
166
- // 200 input @ $0.075/M + 50 output @ $0.30/M
167
- const expectedMin = (200 / 1_000_000) * 0.075 + (50 / 1_000_000) * 0.30;
168
- const expectedMax = expectedMin * 1.01; // 1% tolerance for float arithmetic
169
- assert.ok(result.cost_usd >= expectedMin && result.cost_usd <= expectedMax, `cost_usd=${result.cost_usd} not in [${expectedMin}, ${expectedMax}]`);
170
- });
171
- // -------------------------------------------------------------------------
172
- // 7. GroundedQueryTool.execute — happy path
173
- // -------------------------------------------------------------------------
174
- await test('GroundedQueryTool.execute: happy path produces answer + sources in output', async () => {
175
- groundedQueryTestHooks.resolveKey = async () => 'injected-key';
176
- groundedQueryTestHooks.fetchResponse = async () => STUB_RESPONSE_GROUNDED;
177
- const tool = new GroundedQueryTool();
178
- try {
179
- const output = await tool.execute({ query: 'How many albums did Mercedes Sosa release?' });
180
- assert.ok(output.includes('gemini-2.5-flash'), 'model missing from output');
181
- assert.ok(output.includes('5 studio albums'), 'answer text missing from output');
182
- assert.ok(output.includes('Sources:'), 'Sources section missing from output');
183
- assert.ok(output.includes('example.com/mercedes-sosa'), 'source URI missing from output');
184
- }
185
- finally {
186
- clearHooks();
187
- }
188
- });
189
- // -------------------------------------------------------------------------
190
- // 8. GroundedQueryTool.execute — HTTP 4xx surfaces useful error
191
- // -------------------------------------------------------------------------
192
- await test('GroundedQueryTool.execute: HTTP error surfaces in thrown Error', async () => {
193
- groundedQueryTestHooks.resolveKey = async () => 'bad-key';
194
- groundedQueryTestHooks.fetchResponse = async () => {
195
- throw new Error('grounded_query: Gemini HTTP 400: API key not valid');
196
- };
197
- const tool = new GroundedQueryTool();
198
- try {
199
- await assert.rejects(() => tool.execute({ query: 'test error' }), /Gemini HTTP 400/);
200
- }
201
- finally {
202
- clearHooks();
203
- }
204
- });
205
- // -------------------------------------------------------------------------
206
- // 9. GroundedQueryTool.execute — ungrounded response includes note
207
- // -------------------------------------------------------------------------
208
- await test('GroundedQueryTool.execute: ungrounded response includes note in output', async () => {
209
- groundedQueryTestHooks.resolveKey = async () => 'injected-key';
210
- groundedQueryTestHooks.fetchResponse = async () => STUB_RESPONSE_UNGROUNDED;
211
- const tool = new GroundedQueryTool();
212
- try {
213
- const output = await tool.execute({ query: 'some query' });
214
- assert.ok(output.includes('unverified'), `expected unverified note in: ${output}`);
215
- }
216
- finally {
217
- clearHooks();
218
- }
219
- });
220
- // -------------------------------------------------------------------------
221
- // 10. GroundedQueryTool.execute — rejects empty query
222
- // -------------------------------------------------------------------------
223
- await test('GroundedQueryTool.execute: rejects empty query', async () => {
224
- const tool = new GroundedQueryTool();
225
- await assert.rejects(() => tool.execute({ query: '' }), /query.*required/);
226
- });
227
- // -------------------------------------------------------------------------
228
- // 11. GroundedQueryTool.execute — max_tokens clamped
229
- // -------------------------------------------------------------------------
230
- await test('GroundedQueryTool.execute: max_tokens clamped to 8192', async () => {
231
- let capturedMaxTokens = 0;
232
- groundedQueryTestHooks.resolveKey = async () => 'injected-key';
233
- groundedQueryTestHooks.fetchResponse = async (_q, maxTokens) => {
234
- capturedMaxTokens = maxTokens;
235
- return STUB_RESPONSE_GROUNDED;
236
- };
237
- const tool = new GroundedQueryTool();
238
- try {
239
- await tool.execute({ query: 'test max tokens', max_tokens: 99_999 });
240
- assert.equal(capturedMaxTokens, 8192, `expected 8192, got ${capturedMaxTokens}`);
241
- }
242
- finally {
243
- clearHooks();
244
- }
245
- });
246
- // -------------------------------------------------------------------------
247
- // 12. GroundedQueryTool — uses injected apiKey from constructor
248
- // -------------------------------------------------------------------------
249
- await test('createGroundedQueryTool: uses apiKey passed to constructor', async () => {
250
- let capturedKey = '';
251
- groundedQueryTestHooks.fetchResponse = async (_q, _max, key) => {
252
- capturedKey = key;
253
- return STUB_RESPONSE_GROUNDED;
254
- };
255
- const tool = createGroundedQueryTool({ apiKey: 'constructor-key' });
256
- try {
257
- await tool.execute({ query: 'test constructor key' });
258
- assert.equal(capturedKey, 'constructor-key', `expected 'constructor-key', got '${capturedKey}'`);
259
- }
260
- finally {
261
- clearHooks();
262
- }
263
- });
264
- // ---------------------------------------------------------------------------
265
- // Summary
266
- // ---------------------------------------------------------------------------
267
- console.log(`\n${passed} passed, ${failed} failed\n`);
268
- if (failed > 0) {
269
- process.exit(1);
270
- }
271
- }
272
- runSuite().catch((err) => {
273
- console.error('Smoke test runner error:', err);
274
- process.exit(1);
275
- });
276
- //# sourceMappingURL=grounded_query.smoke.js.map
@@ -1 +0,0 @@
1
- {"version":3,"file":"grounded_query.smoke.js","sourceRoot":"","sources":["../../../../src/benchmarks/gaia-tools/grounded_query.smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;AAEH,OAAO,KAAK,MAAM,MAAM,oBAAoB,CAAC;AAC7C,OAAO,EACL,iBAAiB,EACjB,sBAAsB,EACtB,qBAAqB,EACrB,mBAAmB,EACnB,uBAAuB,GACxB,MAAM,qBAAqB,CAAC;AAE7B,8EAA8E;AAC9E,6EAA6E;AAC7E,8EAA8E;AAE9E,IAAI,MAAM,GAAG,CAAC,CAAC;AACf,IAAI,MAAM,GAAG,CAAC,CAAC;AAEf,KAAK,UAAU,IAAI,CAAC,IAAY,EAAE,EAAuB;IACvD,IAAI,CAAC;QACH,MAAM,EAAE,EAAE,CAAC;QACX,OAAO,CAAC,GAAG,CAAC,WAAW,IAAI,EAAE,CAAC,CAAC;QAC/B,MAAM,EAAE,CAAC;IACX,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,OAAO,CAAC,KAAK,CAAC,WAAW,IAAI,EAAE,CAAC,CAAC;QACjC,OAAO,CAAC,KAAK,CAAC,WAAY,GAAa,CAAC,OAAO,EAAE,CAAC,CAAC;QACnD,MAAM,EAAE,CAAC;IACX,CAAC;AACH,CAAC;AAED,SAAS,UAAU;IACjB,OAAO,sBAAsB,CAAC,UAAU,CAAC;IACzC,OAAO,sBAAsB,CAAC,aAAa,CAAC;AAC9C,CAAC;AAED,8EAA8E;AAC9E,mBAAmB;AACnB,8EAA8E;AAE9E,MAAM,sBAAsB,GAAG;IAC7B,UAAU,EAAE;QACV;YACE,OAAO,EAAE,EAAE,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,wDAAwD,EAAE,CAAC,EAAE;YACxF,iBAAiB,EAAE;gBACjB,eAAe,EAAE;oBACf,EAAE,GAAG,EAAE,EAAE,KAAK,EAAE,2BAA2B,EAAE,GAAG,EAAE,mCAAmC,EAAE,EAAE;oBACzF,EAAE,GAAG,EAAE,EAAE,KAAK,EAAE,wBAAwB,EAAE,GAAG,EAAE,2BAA2B,EAAE,EAAE;oBAC9E,EAAE,GAAG,EAAE,EAAE,KAAK,EAAE,yBAAyB,EAAE,GAAG,EAAE,6CAA6C,EAAE,EAAE;oBACjG,EAAE,GAAG,EAAE,EAAE,KAAK,EAAE,iBAAiB,EAAE,GAAG,EAAE,mCAAmC,EAAE,EAAE,EAAE,gBAAgB;iBAClG;gBACD,gBAAgB,EAAE,CAAC,mCAAmC,CAAC;aACxD;SACF;KACF;IACD,aAAa,EAAE,EAAE,gBAAgB,EAAE,GAAG,EAAE,oBAAoB,EAAE,EAAE,EAAE;CACnE,CAAC;AAEF,MAAM,wBAAwB,GAAG;IAC/B,UAAU,EAAE;QACV;YACE,OAAO,EAAE,EAAE,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,4CAA4C,EAAE,CAAC,EAAE;YAC5E,uBAAuB;SACxB;KACF;IACD,aAAa,EAAE,EAAE,gBAAgB,EAAE,GAAG,EAAE,oBAAoB,EAAE,EAAE,EAAE;CACnE,CAAC;AAEF,MAAM,6BAA6B,GAAG;IACpC,UAAU,EAAE;QACV;YACE,OAAO,EAAE,EAAE,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,CAAC,EAAE;YAC9C,iBAAiB,EAAE;gBACjB,eAAe,EAAE,EAAE;gBACnB,gBAAgB,EAAE,EAAE;aACrB;SACF;KACF;IACD,aAAa,EAAE,EAAE,gBAAgB,EAAE,EAAE,EAAE,oBAAoB,EAAE,EAAE,EAAE;CAClE,CAAC;AAEF,8EAA8E;AAC9E,aAAa;AACb,8EAA8E;AAE9E,KAAK,UAAU,QAAQ;IACrB,OAAO,CAAC,GAAG,CAAC,kEAAkE,CAAC,CAAC;IAEhF,4EAA4E;IAC5E,gDAAgD;IAChD,4EAA4E;IAC5E,MAAM,IAAI,CAAC,iDAAiD,EAAE,KAAK,IAAI,EAAE;QACvE,MAAM,QAAQ,GAAG,OAAO,CAAC,GAAG,CAAC,mBAAmB,CAAC,CAAC;QAClD,OAAO,CAAC,GAAG,CAAC,mBAAmB,CAAC,GAAG,qBAAqB,CAAC;QACzD,IAAI,CAAC;YACH,MAAM,GAAG,GAAG,MAAM,qBAAqB,EAAE,CAAC;YAC1C,MAAM,CAAC,KAAK,CAAC,GAAG,EAAE,qBAAqB,CAAC,CAAC;QAC3C,CAAC;gBAAS,CAAC;YACT,IAAI,QAAQ,KAAK,SAAS;gBAAE,OAAO,OAAO,CAAC,GAAG,CAAC,mBAAmB,CAAC,CAAC;;gBAC/D,OAAO,CAAC,GAAG,CAAC,mBAAmB,CAAC,GAAG,QAAQ,CAAC;QACnD,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,4EAA4E;IAC5E,8EAA8E;IAC9E,4EAA4E;IAC5E,MAAM,IAAI,CAAC,mEAAmE,EAAE,KAAK,IAAI,EAAE;QACzF,2EAA2E;QAC3E,4EAA4E;QAC5E,sBAAsB,CAAC,UAAU,GAAG,KAAK,IAAI,EAAE;YAC7C,MAAM,IAAI,KAAK,CACb,+CAA+C;gBAC7C,uEAAuE;gBACvE,uDAAuD,CAC1D,CAAC;QACJ,CAAC,CAAC;QACF,MAAM,IAAI,GAAG,IAAI,iBAAiB,EAAE,CAAC;QACrC,IAAI,CAAC;YACH,MAAM,MAAM,CAAC,OAAO,CAClB,GAAG,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,kBAAkB,EAAE,CAAC,EACjD,mBAAmB,CACpB,CAAC;QACJ,CAAC;gBAAS,CAAC;YACT,UAAU,EAAE,CAAC;QACf,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,yDAAyD;IACzD,4EAA4E;IAC5E,MAAM,IAAI,CAAC,gEAAgE,EAAE,KAAK,IAAI,EAAE;QACtF,MAAM,MAAM,GAAG,mBAAmB,CAAC,sBAAsB,CAAC,CAAC;QAC3D,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,KAAK,EAAE,kBAAkB,CAAC,CAAC;QAC/C,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC,QAAQ,CAAC,iBAAiB,CAAC,CAAC,CAAC;QACrD,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,QAAQ,EAAE,IAAI,CAAC,CAAC;QACpC,oDAAoD;QACpD,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC,EAAE,kCAAkC,MAAM,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;QAClG,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,mBAAmB,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;QACnD,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,mBAAmB,CAAC,CAAC,CAAC,EAAE,mCAAmC,CAAC,CAAC;IACnF,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,4DAA4D;IAC5D,4EAA4E;IAC5E,MAAM,IAAI,CAAC,gEAAgE,EAAE,KAAK,IAAI,EAAE;QACtF,MAAM,MAAM,GAAG,mBAAmB,CAAC,6BAA6B,CAAC,CAAC;QAClE,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,QAAQ,EAAE,KAAK,CAAC,CAAC;QACrC,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;QACvC,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,mBAAmB,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACrD,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,6EAA6E;IAC7E,kCAAkC;IAClC,4EAA4E;IAC5E,MAAM,IAAI,CAAC,kDAAkD,EAAE,KAAK,IAAI,EAAE;QACxE,MAAM,MAAM,GAAG,mBAAmB,CAAC,sBAAsB,CAAC,CAAC;QAC3D,MAAM,IAAI,GAAG,MAAM,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;QAC9C,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,IAAI,CAAC,CAAC;QACjC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,EAAE,UAAU,CAAC,IAAI,EAAE,iCAAiC,CAAC,CAAC;IAChF,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,4CAA4C;IAC5C,4EAA4E;IAC5E,MAAM,IAAI,CAAC,4DAA4D,EAAE,KAAK,IAAI,EAAE;QAClF,MAAM,MAAM,GAAG,mBAAmB,CAAC,sBAAsB,CAAC,CAAC;QAC3D,6CAA6C;QAC7C,MAAM,WAAW,GAAG,CAAC,GAAG,GAAG,SAAS,CAAC,GAAG,KAAK,GAAG,CAAC,EAAE,GAAG,SAAS,CAAC,GAAG,IAAI,CAAC;QACxE,MAAM,WAAW,GAAG,WAAW,GAAG,IAAI,CAAC,CAAC,oCAAoC;QAC5E,MAAM,CAAC,EAAE,CACP,MAAM,CAAC,QAAQ,IAAI,WAAW,IAAI,MAAM,CAAC,QAAQ,IAAI,WAAW,EAChE,YAAY,MAAM,CAAC,QAAQ,YAAY,WAAW,KAAK,WAAW,GAAG,CACtE,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,4CAA4C;IAC5C,4EAA4E;IAC5E,MAAM,IAAI,CAAC,2EAA2E,EAAE,KAAK,IAAI,EAAE;QACjG,sBAAsB,CAAC,UAAU,GAAG,KAAK,IAAI,EAAE,CAAC,cAAc,CAAC;QAC/D,sBAAsB,CAAC,aAAa,GAAG,KAAK,IAAI,EAAE,CAAC,sBAAsB,CAAC;QAE1E,MAAM,IAAI,GAAG,IAAI,iBAAiB,EAAE,CAAC;QACrC,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,4CAA4C,EAAE,CAAC,CAAC;YAC3F,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,kBAAkB,CAAC,EAAE,2BAA2B,CAAC,CAAC;YAC5E,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,iBAAiB,CAAC,EAAE,iCAAiC,CAAC,CAAC;YACjF,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,UAAU,CAAC,EAAE,qCAAqC,CAAC,CAAC;YAC9E,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,2BAA2B,CAAC,EAAE,gCAAgC,CAAC,CAAC;QAC5F,CAAC;gBAAS,CAAC;YACT,UAAU,EAAE,CAAC;QACf,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,gEAAgE;IAChE,4EAA4E;IAC5E,MAAM,IAAI,CAAC,gEAAgE,EAAE,KAAK,IAAI,EAAE;QACtF,sBAAsB,CAAC,UAAU,GAAG,KAAK,IAAI,EAAE,CAAC,SAAS,CAAC;QAC1D,sBAAsB,CAAC,aAAa,GAAG,KAAK,IAAI,EAAE;YAChD,MAAM,IAAI,KAAK,CAAC,oDAAoD,CAAC,CAAC;QACxE,CAAC,CAAC;QAEF,MAAM,IAAI,GAAG,IAAI,iBAAiB,EAAE,CAAC;QACrC,IAAI,CAAC;YACH,MAAM,MAAM,CAAC,OAAO,CAClB,GAAG,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,YAAY,EAAE,CAAC,EAC3C,iBAAiB,CAClB,CAAC;QACJ,CAAC;gBAAS,CAAC;YACT,UAAU,EAAE,CAAC;QACf,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,mEAAmE;IACnE,4EAA4E;IAC5E,MAAM,IAAI,CAAC,wEAAwE,EAAE,KAAK,IAAI,EAAE;QAC9F,sBAAsB,CAAC,UAAU,GAAG,KAAK,IAAI,EAAE,CAAC,cAAc,CAAC;QAC/D,sBAAsB,CAAC,aAAa,GAAG,KAAK,IAAI,EAAE,CAAC,wBAAwB,CAAC;QAE5E,MAAM,IAAI,GAAG,IAAI,iBAAiB,EAAE,CAAC;QACrC,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,YAAY,EAAE,CAAC,CAAC;YAC3D,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,YAAY,CAAC,EAAE,gCAAgC,MAAM,EAAE,CAAC,CAAC;QACrF,CAAC;gBAAS,CAAC;YACT,UAAU,EAAE,CAAC;QACf,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,sDAAsD;IACtD,4EAA4E;IAC5E,MAAM,IAAI,CAAC,gDAAgD,EAAE,KAAK,IAAI,EAAE;QACtE,MAAM,IAAI,GAAG,IAAI,iBAAiB,EAAE,CAAC;QACrC,MAAM,MAAM,CAAC,OAAO,CAClB,GAAG,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,EAAE,EAAE,CAAC,EACjC,iBAAiB,CAClB,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,qDAAqD;IACrD,4EAA4E;IAC5E,MAAM,IAAI,CAAC,uDAAuD,EAAE,KAAK,IAAI,EAAE;QAC7E,IAAI,iBAAiB,GAAG,CAAC,CAAC;QAC1B,sBAAsB,CAAC,UAAU,GAAG,KAAK,IAAI,EAAE,CAAC,cAAc,CAAC;QAC/D,sBAAsB,CAAC,aAAa,GAAG,KAAK,EAAE,EAAE,EAAE,SAAS,EAAE,EAAE;YAC7D,iBAAiB,GAAG,SAAS,CAAC;YAC9B,OAAO,sBAAsB,CAAC;QAChC,CAAC,CAAC;QAEF,MAAM,IAAI,GAAG,IAAI,iBAAiB,EAAE,CAAC;QACrC,IAAI,CAAC;YACH,MAAM,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,iBAAiB,EAAE,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;YACrE,MAAM,CAAC,KAAK,CAAC,iBAAiB,EAAE,IAAI,EAAE,sBAAsB,iBAAiB,EAAE,CAAC,CAAC;QACnF,CAAC;gBAAS,CAAC;YACT,UAAU,EAAE,CAAC;QACf,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,4EAA4E;IAC5E,gEAAgE;IAChE,4EAA4E;IAC5E,MAAM,IAAI,CAAC,4DAA4D,EAAE,KAAK,IAAI,EAAE;QAClF,IAAI,WAAW,GAAG,EAAE,CAAC;QACrB,sBAAsB,CAAC,aAAa,GAAG,KAAK,EAAE,EAAE,EAAE,IAAI,EAAE,GAAG,EAAE,EAAE;YAC7D,WAAW,GAAG,GAAG,CAAC;YAClB,OAAO,sBAAsB,CAAC;QAChC,CAAC,CAAC;QAEF,MAAM,IAAI,GAAG,uBAAuB,CAAC,EAAE,MAAM,EAAE,iBAAiB,EAAE,CAAC,CAAC;QACpE,IAAI,CAAC;YACH,MAAM,IAAI,CAAC,OAAO,CAAC,EAAE,KAAK,EAAE,sBAAsB,EAAE,CAAC,CAAC;YACtD,MAAM,CAAC,KAAK,CAAC,WAAW,EAAE,iBAAiB,EAAE,oCAAoC,WAAW,GAAG,CAAC,CAAC;QACnG,CAAC;gBAAS,CAAC;YACT,UAAU,EAAE,CAAC;QACf,CAAC;IACH,CAAC,CAAC,CAAC;IAEH,8EAA8E;IAC9E,UAAU;IACV,8EAA8E;IAC9E,OAAO,CAAC,GAAG,CAAC,KAAK,MAAM,YAAY,MAAM,WAAW,CAAC,CAAC;IACtD,IAAI,MAAM,GAAG,CAAC,EAAE,CAAC;QACf,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;AACH,CAAC;AAED,QAAQ,EAAE,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,EAAE;IACvB,OAAO,CAAC,KAAK,CAAC,0BAA0B,EAAE,GAAG,CAAC,CAAC;IAC/C,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC,CAAC,CAAC"}
@@ -1,45 +0,0 @@
1
- /**
2
- * GAIA Tool: image_describe — ADR-133-PR5
3
- *
4
- * Describes an image at a given URL or local file path using Anthropic's
5
- * claude-haiku-4-5 model with vision (image content blocks). Covers the
6
- * subset of GAIA Level-1 questions that provide image attachments (graphs,
7
- * screenshots, photos, diagrams).
8
- *
9
- * ============================================================
10
- * DESIGN NOTES
11
- * ============================================================
12
- * - Uses the Anthropic Messages API directly via `fetch` (same approach as
13
- * gaia-agent.ts / gaia-judge.ts) — no SDK dependency.
14
- * - API key resolution order mirrors gaia-agent.ts:
15
- * 1. `options.apiKey` (caller-supplied)
16
- * 2. ANTHROPIC_API_KEY env var
17
- * 3. gcloud secrets versions access latest --secret=ANTHROPIC_API_KEY
18
- * - Model: claude-haiku-4-5 (cheapest vision-capable model, ~$0.001/call).
19
- * - URL images: sent as { type: 'url', url } — Anthropic fetches the image.
20
- * - Local files: read as Buffer, base64-encoded, MIME detected from extension.
21
- * - execute() NEVER throws — returns a structured error string so the agent
22
- * loop can forward it to Claude rather than crashing.
23
- *
24
- * ============================================================
25
- * SUPPORTED IMAGE FORMATS
26
- * ============================================================
27
- * Anthropic vision accepts: JPEG, PNG, GIF, WebP.
28
- * Unsupported formats return a descriptive error that Claude can relay.
29
- *
30
- * Refs: ADR-133, #2156
31
- */
32
- import { GaiaTool, ToolDefinition } from './types.js';
33
- export declare class ImageDescribeTool implements GaiaTool {
34
- readonly name = "image_describe";
35
- private readonly apiKey?;
36
- constructor(apiKey?: string);
37
- readonly definition: ToolDefinition;
38
- execute(input: Record<string, unknown>): Promise<string>;
39
- }
40
- export interface ImageDescribeToolOptions {
41
- /** Anthropic API key (falls back to env / gcloud if omitted). */
42
- apiKey?: string;
43
- }
44
- export declare function createImageDescribeTool(opts?: ImageDescribeToolOptions): ImageDescribeTool;
45
- //# sourceMappingURL=image_describe.d.ts.map
@@ -1 +0,0 @@
1
- {"version":3,"file":"image_describe.d.ts","sourceRoot":"","sources":["../../../../src/benchmarks/gaia-tools/image_describe.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AAKH,OAAO,EAAE,QAAQ,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAwJtD,qBAAa,iBAAkB,YAAW,QAAQ;IAChD,QAAQ,CAAC,IAAI,oBAAoB;IACjC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAS;gBAErB,MAAM,CAAC,EAAE,MAAM;IAI3B,QAAQ,CAAC,UAAU,EAAE,cAAc,CA0BjC;IAEI,OAAO,CAAC,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,OAAO,CAAC,MAAM,CAAC;CA6D/D;AAMD,MAAM,WAAW,wBAAwB;IACvC,iEAAiE;IACjE,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED,wBAAgB,uBAAuB,CAAC,IAAI,CAAC,EAAE,wBAAwB,GAAG,iBAAiB,CAE1F"}
@@ -1,222 +0,0 @@
1
- /**
2
- * GAIA Tool: image_describe — ADR-133-PR5
3
- *
4
- * Describes an image at a given URL or local file path using Anthropic's
5
- * claude-haiku-4-5 model with vision (image content blocks). Covers the
6
- * subset of GAIA Level-1 questions that provide image attachments (graphs,
7
- * screenshots, photos, diagrams).
8
- *
9
- * ============================================================
10
- * DESIGN NOTES
11
- * ============================================================
12
- * - Uses the Anthropic Messages API directly via `fetch` (same approach as
13
- * gaia-agent.ts / gaia-judge.ts) — no SDK dependency.
14
- * - API key resolution order mirrors gaia-agent.ts:
15
- * 1. `options.apiKey` (caller-supplied)
16
- * 2. ANTHROPIC_API_KEY env var
17
- * 3. gcloud secrets versions access latest --secret=ANTHROPIC_API_KEY
18
- * - Model: claude-haiku-4-5 (cheapest vision-capable model, ~$0.001/call).
19
- * - URL images: sent as { type: 'url', url } — Anthropic fetches the image.
20
- * - Local files: read as Buffer, base64-encoded, MIME detected from extension.
21
- * - execute() NEVER throws — returns a structured error string so the agent
22
- * loop can forward it to Claude rather than crashing.
23
- *
24
- * ============================================================
25
- * SUPPORTED IMAGE FORMATS
26
- * ============================================================
27
- * Anthropic vision accepts: JPEG, PNG, GIF, WebP.
28
- * Unsupported formats return a descriptive error that Claude can relay.
29
- *
30
- * Refs: ADR-133, #2156
31
- */
32
- import * as fs from 'node:fs';
33
- import * as path from 'node:path';
34
- import { execSync } from 'node:child_process';
35
- // ---------------------------------------------------------------------------
36
- // Constants
37
- // ---------------------------------------------------------------------------
38
- const VISION_MODEL = 'claude-haiku-4-5';
39
- const ANTHROPIC_API_VERSION = '2023-06-01';
40
- const ANTHROPIC_API_URL = 'https://api.anthropic.com/v1/messages';
41
- const MAX_TOKENS = 512;
42
- const DEFAULT_PROMPT = 'Describe this image in detail. Note any text, charts, diagrams, tables, ' +
43
- 'numerical data, labels, axes, legends, or other content that would be ' +
44
- 'useful to answer a factual question about what the image shows.';
45
- const EXT_TO_MIME = {
46
- '.jpg': 'image/jpeg',
47
- '.jpeg': 'image/jpeg',
48
- '.png': 'image/png',
49
- '.gif': 'image/gif',
50
- '.webp': 'image/webp',
51
- };
52
- function detectMime(filePath, buf) {
53
- // Magic bytes first
54
- if (buf.length >= 2 && buf[0] === 0xff && buf[1] === 0xd8)
55
- return 'image/jpeg';
56
- if (buf.length >= 4 &&
57
- buf[0] === 0x89 &&
58
- buf[1] === 0x50 &&
59
- buf[2] === 0x4e &&
60
- buf[3] === 0x47)
61
- return 'image/png';
62
- if (buf.length >= 4 &&
63
- buf[0] === 0x47 &&
64
- buf[1] === 0x49 &&
65
- buf[2] === 0x46 &&
66
- buf[3] === 0x38)
67
- return 'image/gif';
68
- if (buf.length >= 4 &&
69
- buf[0] === 0x52 &&
70
- buf[1] === 0x49 &&
71
- buf[2] === 0x46 &&
72
- buf[3] === 0x46)
73
- return 'image/webp';
74
- // Fall back to extension
75
- const ext = path.extname(filePath).toLowerCase();
76
- return EXT_TO_MIME[ext] ?? 'image/png';
77
- }
78
- // ---------------------------------------------------------------------------
79
- // API key resolution (mirrors gaia-agent.ts)
80
- // ---------------------------------------------------------------------------
81
- function resolveAnthropicApiKey(suppliedKey) {
82
- if (suppliedKey && suppliedKey.trim())
83
- return suppliedKey.trim();
84
- const envKey = process.env.ANTHROPIC_API_KEY;
85
- if (envKey && envKey.trim())
86
- return envKey.trim();
87
- try {
88
- const out = execSync('gcloud secrets versions access latest --secret=ANTHROPIC_API_KEY 2>/dev/null', { encoding: 'utf-8', timeout: 10_000 }).trim();
89
- if (out)
90
- return out;
91
- }
92
- catch {
93
- /* fall through */
94
- }
95
- throw new Error('ANTHROPIC_API_KEY not found. Set the env var or store it in GCP Secret Manager ' +
96
- 'under "ANTHROPIC_API_KEY".');
97
- }
98
- async function callVisionApi(imageBlock, prompt, apiKey) {
99
- const body = JSON.stringify({
100
- model: VISION_MODEL,
101
- max_tokens: MAX_TOKENS,
102
- messages: [
103
- {
104
- role: 'user',
105
- content: [imageBlock, { type: 'text', text: prompt }],
106
- },
107
- ],
108
- });
109
- const resp = await fetch(ANTHROPIC_API_URL, {
110
- method: 'POST',
111
- headers: {
112
- 'x-api-key': apiKey,
113
- 'anthropic-version': ANTHROPIC_API_VERSION,
114
- 'content-type': 'application/json',
115
- },
116
- body,
117
- });
118
- if (!resp.ok) {
119
- const errText = await resp.text().catch(() => '(no body)');
120
- throw new Error(`Anthropic API error ${resp.status}: ${errText}`);
121
- }
122
- const json = (await resp.json());
123
- if (json.error)
124
- throw new Error(`Anthropic API error: ${json.error.message}`);
125
- const textBlock = json.content?.find((b) => b.type === 'text');
126
- return textBlock?.text ?? '(no description returned)';
127
- }
128
- // ---------------------------------------------------------------------------
129
- // GaiaTool implementation
130
- // ---------------------------------------------------------------------------
131
- export class ImageDescribeTool {
132
- name = 'image_describe';
133
- apiKey;
134
- constructor(apiKey) {
135
- this.apiKey = apiKey;
136
- }
137
- definition = {
138
- name: 'image_describe',
139
- description: 'Describe an image at a given URL or local absolute file path. ' +
140
- 'Returns a detailed text description suitable as input to answer ' +
141
- 'factual questions about the image content (charts, graphs, photos, ' +
142
- 'screenshots, diagrams, text in images). ' +
143
- `Uses ${VISION_MODEL} for cost-efficient vision (~$0.001/call). ` +
144
- 'Supported formats: JPEG, PNG, GIF, WebP.',
145
- input_schema: {
146
- type: 'object',
147
- properties: {
148
- source: {
149
- type: 'string',
150
- description: 'URL (http/https) or absolute local file path of the image to describe.',
151
- },
152
- prompt: {
153
- type: 'string',
154
- description: 'Optional instruction to guide the description ' +
155
- '(default: describe all visible content with focus on factual details).',
156
- },
157
- },
158
- required: ['source'],
159
- },
160
- };
161
- async execute(input) {
162
- const source = String(input['source'] ?? '').trim();
163
- if (!source)
164
- throw new Error('image_describe: `source` input is required and must be non-empty.');
165
- const prompt = input['prompt'] != null ? String(input['prompt']).trim() : DEFAULT_PROMPT;
166
- // Resolve API key — errors here are caught by the agent loop.
167
- let apiKey;
168
- try {
169
- apiKey = resolveAnthropicApiKey(this.apiKey);
170
- }
171
- catch (e) {
172
- return `[image_describe error] ${String(e)}`;
173
- }
174
- let imageBlock;
175
- if (source.startsWith('http://') || source.startsWith('https://')) {
176
- // URL image — Anthropic fetches it directly.
177
- imageBlock = {
178
- type: 'image',
179
- source: { type: 'url', url: source },
180
- };
181
- }
182
- else {
183
- // Local file — validate, read, base64-encode.
184
- if (!path.isAbsolute(source)) {
185
- return (`[image_describe error] Local file paths must be absolute. ` +
186
- `Got: "${source}".`);
187
- }
188
- let buf;
189
- try {
190
- buf = fs.readFileSync(source);
191
- }
192
- catch (e) {
193
- const err = e;
194
- if (err.code === 'ENOENT') {
195
- return `[image_describe error] File not found: ${source}`;
196
- }
197
- return `[image_describe error] Cannot read file "${source}": ${String(e)}`;
198
- }
199
- const mime = detectMime(source, buf);
200
- imageBlock = {
201
- type: 'image',
202
- source: {
203
- type: 'base64',
204
- media_type: mime,
205
- data: buf.toString('base64'),
206
- },
207
- };
208
- }
209
- try {
210
- const description = await callVisionApi(imageBlock, prompt, apiKey);
211
- return `[image_describe: ${VISION_MODEL}]\n${description}`;
212
- }
213
- catch (e) {
214
- // Return error as string — never throw so the agent loop stays alive.
215
- return `[image_describe error] ${String(e)}`;
216
- }
217
- }
218
- }
219
- export function createImageDescribeTool(opts) {
220
- return new ImageDescribeTool(opts?.apiKey);
221
- }
222
- //# sourceMappingURL=image_describe.js.map
@@ -1 +0,0 @@
1
- {"version":3,"file":"image_describe.js","sourceRoot":"","sources":["../../../../src/benchmarks/gaia-tools/image_describe.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AAEH,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AAG9C,8EAA8E;AAC9E,YAAY;AACZ,8EAA8E;AAE9E,MAAM,YAAY,GAAG,kBAAkB,CAAC;AACxC,MAAM,qBAAqB,GAAG,YAAY,CAAC;AAC3C,MAAM,iBAAiB,GAAG,uCAAuC,CAAC;AAClE,MAAM,UAAU,GAAG,GAAG,CAAC;AACvB,MAAM,cAAc,GAClB,0EAA0E;IAC1E,wEAAwE;IACxE,iEAAiE,CAAC;AAQpE,MAAM,WAAW,GAAkC;IACjD,MAAM,EAAE,YAAY;IACpB,OAAO,EAAE,YAAY;IACrB,MAAM,EAAE,WAAW;IACnB,MAAM,EAAE,WAAW;IACnB,OAAO,EAAE,YAAY;CACtB,CAAC;AAEF,SAAS,UAAU,CAAC,QAAgB,EAAE,GAAW;IAC/C,oBAAoB;IACpB,IAAI,GAAG,CAAC,MAAM,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI,IAAI,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QAAE,OAAO,YAAY,CAAC;IAC/E,IACE,GAAG,CAAC,MAAM,IAAI,CAAC;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QAEf,OAAO,WAAW,CAAC;IACrB,IACE,GAAG,CAAC,MAAM,IAAI,CAAC;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QAEf,OAAO,WAAW,CAAC;IACrB,IACE,GAAG,CAAC,MAAM,IAAI,CAAC;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QACf,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI;QAEf,OAAO,YAAY,CAAC;IACtB,yBAAyB;IACzB,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,WAAW,EAAE,CAAC;IACjD,OAAO,WAAW,CAAC,GAAG,CAAC,IAAI,WAAW,CAAC;AACzC,CAAC;AAED,8EAA8E;AAC9E,6CAA6C;AAC7C,8EAA8E;AAE9E,SAAS,sBAAsB,CAAC,WAAoB;IAClD,IAAI,WAAW,IAAI,WAAW,CAAC,IAAI,EAAE;QAAE,OAAO,WAAW,CAAC,IAAI,EAAE,CAAC;IACjE,MAAM,MAAM,GAAG,OAAO,CAAC,GAAG,CAAC,iBAAiB,CAAC;IAC7C,IAAI,MAAM,IAAI,MAAM,CAAC,IAAI,EAAE;QAAE,OAAO,MAAM,CAAC,IAAI,EAAE,CAAC;IAClD,IAAI,CAAC;QACH,MAAM,GAAG,GAAG,QAAQ,CAClB,8EAA8E,EAC9E,EAAE,QAAQ,EAAE,OAAO,EAAE,OAAO,EAAE,MAAM,EAAE,CACvC,CAAC,IAAI,EAAE,CAAC;QACT,IAAI,GAAG;YAAE,OAAO,GAAG,CAAC;IACtB,CAAC;IAAC,MAAM,CAAC;QACP,kBAAkB;IACpB,CAAC;IACD,MAAM,IAAI,KAAK,CACb,kFAAkF;QAChF,4BAA4B,CAC/B,CAAC;AACJ,CAAC;AAuBD,KAAK,UAAU,aAAa,CAC1B,UAA+B,EAC/B,MAAc,EACd,MAAc;IAEd,MAAM,IAAI,GAAG,IAAI,CAAC,SAAS,CAAC;QAC1B,KAAK,EAAE,YAAY;QACnB,UAAU,EAAE,UAAU;QACtB,QAAQ,EAAE;YACR;gBACE,IAAI,EAAE,MAAM;gBACZ,OAAO,EAAE,CAAC,UAAU,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAwB,CAAC;aAC5E;SACF;KACF,CAAC,CAAC;IAEH,MAAM,IAAI,GAAG,MAAM,KAAK,CAAC,iBAAiB,EAAE;QAC1C,MAAM,EAAE,MAAM;QACd,OAAO,EAAE;YACP,WAAW,EAAE,MAAM;YACnB,mBAAmB,EAAE,qBAAqB;YAC1C,cAAc,EAAE,kBAAkB;SACnC;QACD,IAAI;KACL,CAAC,CAAC;IAEH,IAAI,CAAC,IAAI,CAAC,EAAE,EAAE,CAAC;QACb,MAAM,OAAO,GAAG,MAAM,IAAI,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,WAAW,CAAC,CAAC;QAC3D,MAAM,IAAI,KAAK,CAAC,uBAAuB,IAAI,CAAC,MAAM,KAAK,OAAO,EAAE,CAAC,CAAC;IACpE,CAAC;IAED,MAAM,IAAI,GAAG,CAAC,MAAM,IAAI,CAAC,IAAI,EAAE,CAI9B,CAAC;IAEF,IAAI,IAAI,CAAC,KAAK;QAAE,MAAM,IAAI,KAAK,CAAC,wBAAwB,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;IAE9E,MAAM,SAAS,GAAG,IAAI,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,MAAM,CAAC,CAAC;IAC/D,OAAO,SAAS,EAAE,IAAI,IAAI,2BAA2B,CAAC;AACxD,CAAC;AAED,8EAA8E;AAC9E,0BAA0B;AAC1B,8EAA8E;AAE9E,MAAM,OAAO,iBAAiB;IACnB,IAAI,GAAG,gBAAgB,CAAC;IAChB,MAAM,CAAU;IAEjC,YAAY,MAAe;QACzB,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;IACvB,CAAC;IAEQ,UAAU,GAAmB;QACpC,IAAI,EAAE,gBAAgB;QACtB,WAAW,EACT,gEAAgE;YAChE,kEAAkE;YAClE,qEAAqE;YACrE,0CAA0C;YAC1C,QAAQ,YAAY,6CAA6C;YACjE,0CAA0C;QAC5C,YAAY,EAAE;YACZ,IAAI,EAAE,QAAQ;YACd,UAAU,EAAE;gBACV,MAAM,EAAE;oBACN,IAAI,EAAE,QAAQ;oBACd,WAAW,EACT,wEAAwE;iBAC3E;gBACD,MAAM,EAAE;oBACN,IAAI,EAAE,QAAQ;oBACd,WAAW,EACT,gDAAgD;wBAChD,wEAAwE;iBAC3E;aACF;YACD,QAAQ,EAAE,CAAC,QAAQ,CAAC;SACrB;KACF,CAAC;IAEF,KAAK,CAAC,OAAO,CAAC,KAA8B;QAC1C,MAAM,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;QACpD,IAAI,CAAC,MAAM;YAAE,MAAM,IAAI,KAAK,CAAC,mEAAmE,CAAC,CAAC;QAElG,MAAM,MAAM,GAAG,KAAK,CAAC,QAAQ,CAAC,IAAI,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,cAAc,CAAC;QAEzF,8DAA8D;QAC9D,IAAI,MAAc,CAAC;QACnB,IAAI,CAAC;YACH,MAAM,GAAG,sBAAsB,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAC/C,CAAC;QAAC,OAAO,CAAU,EAAE,CAAC;YACpB,OAAO,0BAA0B,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC;QAC/C,CAAC;QAED,IAAI,UAA+B,CAAC;QAEpC,IAAI,MAAM,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,MAAM,CAAC,UAAU,CAAC,UAAU,CAAC,EAAE,CAAC;YAClE,6CAA6C;YAC7C,UAAU,GAAG;gBACX,IAAI,EAAE,OAAO;gBACb,MAAM,EAAE,EAAE,IAAI,EAAE,KAAK,EAAE,GAAG,EAAE,MAAM,EAAE;aACrC,CAAC;QACJ,CAAC;aAAM,CAAC;YACN,8CAA8C;YAC9C,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC7B,OAAO,CACL,4DAA4D;oBAC5D,SAAS,MAAM,IAAI,CACpB,CAAC;YACJ,CAAC;YAED,IAAI,GAAW,CAAC;YAChB,IAAI,CAAC;gBACH,GAAG,GAAG,EAAE,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;YAChC,CAAC;YAAC,OAAO,CAAU,EAAE,CAAC;gBACpB,MAAM,GAAG,GAAG,CAA0B,CAAC;gBACvC,IAAI,GAAG,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;oBAC1B,OAAO,0CAA0C,MAAM,EAAE,CAAC;gBAC5D,CAAC;gBACD,OAAO,4CAA4C,MAAM,MAAM,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC;YAC7E,CAAC;YAED,MAAM,IAAI,GAAG,UAAU,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;YACrC,UAAU,GAAG;gBACX,IAAI,EAAE,OAAO;gBACb,MAAM,EAAE;oBACN,IAAI,EAAE,QAAQ;oBACd,UAAU,EAAE,IAAI;oBAChB,IAAI,EAAE,GAAG,CAAC,QAAQ,CAAC,QAAQ,CAAC;iBAC7B;aACF,CAAC;QACJ,CAAC;QAED,IAAI,CAAC;YACH,MAAM,WAAW,GAAG,MAAM,aAAa,CAAC,UAAU,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;YACpE,OAAO,oBAAoB,YAAY,MAAM,WAAW,EAAE,CAAC;QAC7D,CAAC;QAAC,OAAO,CAAU,EAAE,CAAC;YACpB,sEAAsE;YACtE,OAAO,0BAA0B,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC;QAC/C,CAAC;IACH,CAAC;CACF;AAWD,MAAM,UAAU,uBAAuB,CAAC,IAA+B;IACrE,OAAO,IAAI,iBAAiB,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;AAC7C,CAAC"}
@@ -1,24 +0,0 @@
1
- /**
2
- * Smoke tests for image_describe — ADR-133-PR5
3
- *
4
- * Run with:
5
- * cd v3/@claude-flow/cli
6
- * npx tsx src/benchmarks/gaia-tools/image_describe.smoke.ts
7
- *
8
- * Tests:
9
- * 1. URL image (Wikipedia logo PNG) — description contains 'wiki' or 'logo' or 'W'
10
- * 2. Missing API key — returns error string, does not crash
11
- * 3. Non-existent local file — returns error string, does not crash
12
- * 4. Invalid source (empty) — throws (caught by test harness)
13
- *
14
- * If ANTHROPIC_API_KEY is not set and gcloud is not available, Test 1 is
15
- * marked SKIP (not FAIL) — the tool's graceful error path is validated in
16
- * Test 2 instead.
17
- *
18
- * Cost estimate: ~$0.001 per live API call (Haiku vision, single small image).
19
- * Total smoke cost: ≤$0.001 (only Test 1 makes a live call).
20
- *
21
- * Refs: ADR-133, #2156
22
- */
23
- export {};
24
- //# sourceMappingURL=image_describe.smoke.d.ts.map
@@ -1 +0,0 @@
1
- {"version":3,"file":"image_describe.smoke.d.ts","sourceRoot":"","sources":["../../../../src/benchmarks/gaia-tools/image_describe.smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG"}