@apmantza/greedysearch-pi 1.4.1 → 1.5.0

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 (26) hide show
  1. package/.pi-lens/cache/jscpd.json +112 -0
  2. package/.pi-lens/cache/jscpd.meta.json +3 -0
  3. package/.pi-lens/cache/knip.json +111 -0
  4. package/.pi-lens/cache/knip.meta.json +4 -0
  5. package/.pi-lens/fix-plan.md +13 -0
  6. package/.pi-lens/fix-session.json +11 -0
  7. package/.pi-lens/metrics-history.json +182 -0
  8. package/.pi-lens/reports/fix-plan.tsv +38 -0
  9. package/.pi-lens/turn-state.json +6 -0
  10. package/CHANGELOG.md +30 -0
  11. package/README.md +52 -27
  12. package/cdp.mjs +1004 -799
  13. package/coding-task.mjs +392 -369
  14. package/extractors/bing-copilot.mjs +167 -204
  15. package/extractors/common.mjs +237 -0
  16. package/extractors/consent.mjs +273 -248
  17. package/extractors/gemini.mjs +142 -221
  18. package/extractors/google-ai.mjs +156 -165
  19. package/extractors/perplexity.mjs +126 -184
  20. package/extractors/selectors.mjs +52 -52
  21. package/index.ts +760 -623
  22. package/launch.mjs +288 -166
  23. package/package.json +26 -26
  24. package/search.mjs +1219 -997
  25. package/skills/greedy-search/SKILL.md +74 -145
  26. package/test.mjs +308 -0
package/.pi-lens/cache/jscpd.json ADDED
@@ -0,0 +1,112 @@
1
+ {
2
+ "success": true,
3
+ "clones": [
4
+ {
5
+ "fileA": "extractors\\gemini.mjs",
6
+ "startA": 63,
7
+ "fileB": "extractors\\perplexity.mjs",
8
+ "startB": 34,
9
+ "lines": 20,
10
+ "tokens": 0
11
+ },
12
+ {
13
+ "fileA": "extractors\\gemini.mjs",
14
+ "startA": 82,
15
+ "fileB": "extractors\\google-ai.mjs",
16
+ "startB": 96,
17
+ "lines": 13,
18
+ "tokens": 0
19
+ },
20
+ {
21
+ "fileA": "extractors\\bing-copilot.mjs",
22
+ "startA": 44,
23
+ "fileB": "extractors\\gemini.mjs",
24
+ "startB": 60,
25
+ "lines": 24,
26
+ "tokens": 0
27
+ },
28
+ {
29
+ "fileA": "extractors\\bing-copilot.mjs",
30
+ "startA": 67,
31
+ "fileB": "extractors\\google-ai.mjs",
32
+ "startB": 96,
33
+ "lines": 13,
34
+ "tokens": 0
35
+ },
36
+ {
37
+ "fileA": "extractors\\bing-copilot.mjs",
38
+ "startA": 111,
39
+ "fileB": "extractors\\perplexity.mjs",
40
+ "startB": 74,
41
+ "lines": 8,
42
+ "tokens": 0
43
+ },
44
+ {
45
+ "fileA": "extractors\\bing-copilot.mjs",
46
+ "startA": 134,
47
+ "fileB": "extractors\\perplexity.mjs",
48
+ "startB": 85,
49
+ "lines": 7,
50
+ "tokens": 0
51
+ },
52
+ {
53
+ "fileA": "extractors\\bing-copilot.mjs",
54
+ "startA": 151,
55
+ "fileB": "extractors\\perplexity.mjs",
56
+ "startB": 110,
57
+ "lines": 17,
58
+ "tokens": 0
59
+ },
60
+ {
61
+ "fileA": "test.mjs",
62
+ "startA": 35,
63
+ "fileB": "extractors\\common.mjs",
64
+ "startB": 25,
65
+ "lines": 6,
66
+ "tokens": 0
67
+ },
68
+ {
69
+ "fileA": "index.ts",
70
+ "startA": 410,
71
+ "fileB": "index.ts",
72
+ "startB": 396,
73
+ "lines": 11,
74
+ "tokens": 0
75
+ },
76
+ {
77
+ "fileA": "index.ts",
78
+ "startA": 573,
79
+ "fileB": "index.ts",
80
+ "startB": 504,
81
+ "lines": 7,
82
+ "tokens": 0
83
+ },
84
+ {
85
+ "fileA": "index.ts",
86
+ "startA": 601,
87
+ "fileB": "index.ts",
88
+ "startB": 531,
89
+ "lines": 9,
90
+ "tokens": 0
91
+ },
92
+ {
93
+ "fileA": "index.ts",
94
+ "startA": 699,
95
+ "fileB": "index.ts",
96
+ "startB": 36,
97
+ "lines": 15,
98
+ "tokens": 0
99
+ },
100
+ {
101
+ "fileA": "coding-task.mjs",
102
+ "startA": 161,
103
+ "fileB": "coding-task.mjs",
104
+ "startB": 103,
105
+ "lines": 11,
106
+ "tokens": 0
107
+ }
108
+ ],
109
+ "duplicatedLines": 148,
110
+ "totalLines": 3187,
111
+ "percentage": 4.64
112
+ }
package/.pi-lens/cache/jscpd.meta.json ADDED
@@ -0,0 +1,3 @@
1
+ {
2
+ "timestamp": "2026-03-29T12:39:01.584Z"
3
+ }
package/.pi-lens/cache/knip.json ADDED
@@ -0,0 +1,111 @@
1
+ {
2
+ "success": true,
3
+ "issues": [
4
+ {
5
+ "type": "file",
6
+ "name": "cdp.mjs",
7
+ "file": "cdp.mjs"
8
+ },
9
+ {
10
+ "type": "file",
11
+ "name": "coding-task.mjs",
12
+ "file": "coding-task.mjs"
13
+ },
14
+ {
15
+ "type": "file",
16
+ "name": "launch.mjs",
17
+ "file": "launch.mjs"
18
+ },
19
+ {
20
+ "type": "file",
21
+ "name": "search.mjs",
22
+ "file": "search.mjs"
23
+ },
24
+ {
25
+ "type": "file",
26
+ "name": "extractors/bing-copilot.mjs",
27
+ "file": "extractors/bing-copilot.mjs"
28
+ },
29
+ {
30
+ "type": "file",
31
+ "name": "extractors/consent.mjs",
32
+ "file": "extractors/consent.mjs"
33
+ },
34
+ {
35
+ "type": "file",
36
+ "name": "extractors/gemini.mjs",
37
+ "file": "extractors/gemini.mjs"
38
+ },
39
+ {
40
+ "type": "file",
41
+ "name": "extractors/google-ai.mjs",
42
+ "file": "extractors/google-ai.mjs"
43
+ },
44
+ {
45
+ "type": "file",
46
+ "name": "extractors/perplexity.mjs",
47
+ "file": "extractors/perplexity.mjs"
48
+ },
49
+ {
50
+ "type": "file",
51
+ "name": "extractors/selectors.mjs",
52
+ "file": "extractors/selectors.mjs"
53
+ }
54
+ ],
55
+ "unusedExports": [],
56
+ "unusedFiles": [
57
+ {
58
+ "type": "file",
59
+ "name": "cdp.mjs",
60
+ "file": "cdp.mjs"
61
+ },
62
+ {
63
+ "type": "file",
64
+ "name": "coding-task.mjs",
65
+ "file": "coding-task.mjs"
66
+ },
67
+ {
68
+ "type": "file",
69
+ "name": "launch.mjs",
70
+ "file": "launch.mjs"
71
+ },
72
+ {
73
+ "type": "file",
74
+ "name": "search.mjs",
75
+ "file": "search.mjs"
76
+ },
77
+ {
78
+ "type": "file",
79
+ "name": "extractors/bing-copilot.mjs",
80
+ "file": "extractors/bing-copilot.mjs"
81
+ },
82
+ {
83
+ "type": "file",
84
+ "name": "extractors/consent.mjs",
85
+ "file": "extractors/consent.mjs"
86
+ },
87
+ {
88
+ "type": "file",
89
+ "name": "extractors/gemini.mjs",
90
+ "file": "extractors/gemini.mjs"
91
+ },
92
+ {
93
+ "type": "file",
94
+ "name": "extractors/google-ai.mjs",
95
+ "file": "extractors/google-ai.mjs"
96
+ },
97
+ {
98
+ "type": "file",
99
+ "name": "extractors/perplexity.mjs",
100
+ "file": "extractors/perplexity.mjs"
101
+ },
102
+ {
103
+ "type": "file",
104
+ "name": "extractors/selectors.mjs",
105
+ "file": "extractors/selectors.mjs"
106
+ }
107
+ ],
108
+ "unusedDeps": [],
109
+ "unlistedDeps": [],
110
+ "summary": "Found 10 issues"
111
+ }
package/.pi-lens/cache/knip.meta.json ADDED
@@ -0,0 +1,4 @@
1
+ {
2
+ "timestamp": "2026-03-29T11:47:12.910Z",
3
+ "scanDurationMs": 2543
4
+ }
package/.pi-lens/fix-plan.md ADDED
@@ -0,0 +1,13 @@
1
+ # Fix Plan — Iteration 1
2
+
3
+ 📋 FIX PLAN — Iteration 1/3 — 37 issues
4
+ 📄 Full plan: .pi-lens/reports/fix-plan.tsv
5
+
6
+ 🔁 Duplicates: 22 block(s) — extract to shared utils
7
+ 🗑️ Dead code: 10 item(s) — remove unused exports
8
+ 🟠 Biome: 5 item(s) — auto-fixable
9
+
10
+ ---
11
+ 📖 **Read plan**: `read .pi-lens/reports/fix-plan.tsv` for full details
12
+ 🚀 **Fix & loop**: Fix items, then run `/lens-booboo-fix --loop`
13
+ 🚫 **False positive**: `/lens-booboo-fix --false-positive "type:file:line"`
package/.pi-lens/fix-session.json ADDED
@@ -0,0 +1,11 @@
1
+ {
2
+ "iteration": 1,
3
+ "counts": {
4
+ "duplicates": 22,
5
+ "dead_code": 10,
6
+ "ast_issues": 0,
7
+ "biome_issues": 5,
8
+ "slop_files": 0
9
+ },
10
+ "falsePositives": []
11
+ }
package/.pi-lens/metrics-history.json ADDED
@@ -0,0 +1,182 @@
1
+ {
2
+ "version": 1,
3
+ "files": {
4
+ "extractors\\perplexity.mjs": {
5
+ "latest": {
6
+ "commit": "unknown",
7
+ "timestamp": "2026-03-29T11:49:19.267Z",
8
+ "mi": 30.3,
9
+ "cognitive": 50,
10
+ "nesting": 5,
11
+ "lines": 186
12
+ },
13
+ "history": [
14
+ {
15
+ "commit": "unknown",
16
+ "timestamp": "2026-03-29T11:49:19.267Z",
17
+ "mi": 30.3,
18
+ "cognitive": 50,
19
+ "nesting": 5,
20
+ "lines": 186
21
+ }
22
+ ],
23
+ "trend": "stable"
24
+ },
25
+ "extractors\\gemini.mjs": {
26
+ "latest": {
27
+ "commit": "unknown",
28
+ "timestamp": "2026-03-29T11:49:41.697Z",
29
+ "mi": 31.2,
30
+ "cognitive": 40,
31
+ "nesting": 4,
32
+ "lines": 185
33
+ },
34
+ "history": [
35
+ {
36
+ "commit": "unknown",
37
+ "timestamp": "2026-03-29T11:49:41.697Z",
38
+ "mi": 31.2,
39
+ "cognitive": 40,
40
+ "nesting": 4,
41
+ "lines": 185
42
+ }
43
+ ],
44
+ "trend": "stable"
45
+ },
46
+ "extractors\\bing-copilot.mjs": {
47
+ "latest": {
48
+ "commit": "unknown",
49
+ "timestamp": "2026-03-29T11:50:05.399Z",
50
+ "mi": 29.9,
51
+ "cognitive": 53,
52
+ "nesting": 5,
53
+ "lines": 203
54
+ },
55
+ "history": [
56
+ {
57
+ "commit": "unknown",
58
+ "timestamp": "2026-03-29T11:50:05.399Z",
59
+ "mi": 29.9,
60
+ "cognitive": 53,
61
+ "nesting": 5,
62
+ "lines": 203
63
+ }
64
+ ],
65
+ "trend": "stable"
66
+ },
67
+ "extractors\\google-ai.mjs": {
68
+ "latest": {
69
+ "commit": "unknown",
70
+ "timestamp": "2026-03-29T11:50:24.174Z",
71
+ "mi": 31.7,
72
+ "cognitive": 49,
73
+ "nesting": 4,
74
+ "lines": 164
75
+ },
76
+ "history": [
77
+ {
78
+ "commit": "unknown",
79
+ "timestamp": "2026-03-29T11:50:24.174Z",
80
+ "mi": 31.7,
81
+ "cognitive": 49,
82
+ "nesting": 4,
83
+ "lines": 164
84
+ }
85
+ ],
86
+ "trend": "stable"
87
+ },
88
+ "coding-task.mjs": {
89
+ "latest": {
90
+ "commit": "unknown",
91
+ "timestamp": "2026-03-29T12:01:31.394Z",
92
+ "mi": 22.2,
93
+ "cognitive": 133,
94
+ "nesting": 4,
95
+ "lines": 355
96
+ },
97
+ "history": [
98
+ {
99
+ "commit": "unknown",
100
+ "timestamp": "2026-03-29T11:51:12.673Z",
101
+ "mi": 20.1,
102
+ "cognitive": 138,
103
+ "nesting": 4,
104
+ "lines": 374
105
+ },
106
+ {
107
+ "commit": "unknown",
108
+ "timestamp": "2026-03-29T12:01:31.394Z",
109
+ "mi": 22.2,
110
+ "cognitive": 133,
111
+ "nesting": 4,
112
+ "lines": 355
113
+ }
114
+ ],
115
+ "trend": "improving"
116
+ },
117
+ "extractors\\common.mjs": {
118
+ "latest": {
119
+ "commit": "unknown",
120
+ "timestamp": "2026-03-29T11:53:06.232Z",
121
+ "mi": 36.1,
122
+ "cognitive": 35,
123
+ "nesting": 5,
124
+ "lines": 191
125
+ },
126
+ "history": [
127
+ {
128
+ "commit": "unknown",
129
+ "timestamp": "2026-03-29T11:53:06.232Z",
130
+ "mi": 36.1,
131
+ "cognitive": 35,
132
+ "nesting": 5,
133
+ "lines": 191
134
+ }
135
+ ],
136
+ "trend": "stable"
137
+ },
138
+ "search.mjs": {
139
+ "latest": {
140
+ "commit": "unknown",
141
+ "timestamp": "2026-03-29T12:16:28.243Z",
142
+ "mi": 5.1,
143
+ "cognitive": 463,
144
+ "nesting": 8,
145
+ "lines": 1066
146
+ },
147
+ "history": [
148
+ {
149
+ "commit": "unknown",
150
+ "timestamp": "2026-03-29T12:16:28.243Z",
151
+ "mi": 5.1,
152
+ "cognitive": 463,
153
+ "nesting": 8,
154
+ "lines": 1066
155
+ }
156
+ ],
157
+ "trend": "stable"
158
+ },
159
+ "test.mjs": {
160
+ "latest": {
161
+ "commit": "unknown",
162
+ "timestamp": "2026-03-29T12:20:38.595Z",
163
+ "mi": 23.4,
164
+ "cognitive": 113,
165
+ "nesting": 4,
166
+ "lines": 259
167
+ },
168
+ "history": [
169
+ {
170
+ "commit": "unknown",
171
+ "timestamp": "2026-03-29T12:20:38.595Z",
172
+ "mi": 23.4,
173
+ "cognitive": 113,
174
+ "nesting": 4,
175
+ "lines": 259
176
+ }
177
+ ],
178
+ "trend": "stable"
179
+ }
180
+ },
181
+ "capturedAt": "2026-03-29T12:20:45.753Z"
182
+ }
package/.pi-lens/reports/fix-plan.tsv ADDED
@@ -0,0 +1,38 @@
1
+ type file rule message
2
+ dup extractors\google-ai.mjs:16 duplicate-code 8 lines duplicated with extractors\perplexity.mjs:16
3
+ dup extractors\google-ai.mjs:28 duplicate-code 42 lines duplicated with extractors\perplexity.mjs:26
4
+ dup extractors\google-ai.mjs:135 duplicate-code 18 lines duplicated with extractors\perplexity.mjs:154
5
+ dup extractors\google-ai.mjs:186 duplicate-code 8 lines duplicated with extractors\perplexity.mjs:211
6
+ dup extractors\google-ai.mjs:193 duplicate-code 12 lines duplicated with extractors\perplexity.mjs:218
7
+ dup extractors\gemini.mjs:8 duplicate-code 17 lines duplicated with extractors\google-ai.mjs:8
8
+ dup extractors\gemini.mjs:26 duplicate-code 10 lines duplicated with extractors\perplexity.mjs:26
9
+ dup extractors\gemini.mjs:36 duplicate-code 36 lines duplicated with extractors\perplexity.mjs:36
10
+ dup extractors\gemini.mjs:140 duplicate-code 21 lines duplicated with extractors\perplexity.mjs:134
11
+ dup extractors\gemini.mjs:160 duplicate-code 21 lines duplicated with extractors\perplexity.mjs:154
12
+ dup extractors\gemini.mjs:212 duplicate-code 7 lines duplicated with extractors\perplexity.mjs:212
13
+ dup extractors\gemini.mjs:218 duplicate-code 12 lines duplicated with extractors\perplexity.mjs:218
14
+ dup extractors\bing-copilot.mjs:8 duplicate-code 16 lines duplicated with extractors\google-ai.mjs:8
15
+ dup extractors\bing-copilot.mjs:26 duplicate-code 46 lines duplicated with extractors\perplexity.mjs:26
16
+ dup extractors\bing-copilot.mjs:100 duplicate-code 12 lines duplicated with extractors\gemini.mjs:118
17
+ dup extractors\bing-copilot.mjs:111 duplicate-code 13 lines duplicated with extractors\perplexity.mjs:122
18
+ dup extractors\bing-copilot.mjs:123 duplicate-code 21 lines duplicated with extractors\perplexity.mjs:134
19
+ dup extractors\bing-copilot.mjs:143 duplicate-code 21 lines duplicated with extractors\perplexity.mjs:154
20
+ dup extractors\bing-copilot.mjs:195 duplicate-code 8 lines duplicated with extractors\perplexity.mjs:182
21
+ dup extractors\bing-copilot.mjs:218 duplicate-code 7 lines duplicated with extractors\perplexity.mjs:193
22
+ dup extractors\bing-copilot.mjs:235 duplicate-code 18 lines duplicated with extractors\perplexity.mjs:212
23
+ dup coding-task.mjs:35 duplicate-code 26 lines duplicated with extractors\perplexity.mjs:26
24
+ dead cdp.mjs file cdp.mjs is unused
25
+ dead coding-task.mjs file coding-task.mjs is unused
26
+ dead launch.mjs file launch.mjs is unused
27
+ dead search.mjs file search.mjs is unused
28
+ dead extractors/bing-copilot.mjs file extractors/bing-copilot.mjs is unused
29
+ dead extractors/consent.mjs file extractors/consent.mjs is unused
30
+ dead extractors/gemini.mjs file extractors/gemini.mjs is unused
31
+ dead extractors/google-ai.mjs file extractors/google-ai.mjs is unused
32
+ dead extractors/perplexity.mjs file extractors/perplexity.mjs is unused
33
+ dead extractors/selectors.mjs file extractors/selectors.mjs is unused
34
+ biome :1 lint/suspicious/noExplicitAny Unexpected any. Specify a different type.
35
+ biome :1 lint/suspicious/noExplicitAny Unexpected any. Specify a different type.
36
+ biome :1 lint/suspicious/noExplicitAny Unexpected any. Specify a different type.
37
+ biome :1 lint/suspicious/useIterableCallbackReturn This callback passed to forEach() iterable method should not return a value.
38
+ biome :1 lint/suspicious/noAssignInExpressions The assignment should not be in an expression.
package/.pi-lens/turn-state.json ADDED
@@ -0,0 +1,6 @@
1
+ {
2
+ "files": {},
3
+ "turnCycles": 0,
4
+ "maxCycles": 3,
5
+ "lastUpdated": "2026-03-29T12:39:01.588Z"
6
+ }
package/CHANGELOG.md ADDED
@@ -0,0 +1,30 @@
1
+ # Changelog
2
+
3
+ ## v1.5.0 (2026-03-29)
4
+
5
+ ### Features
6
+ - **Code extraction fixed** — `coding_task` now uses clipboard interception to preserve markdown code blocks (was losing them via DOM scraping)
7
+ - **Chrome targeting hardened** — all tools now consistently target the dedicated GreedySearch Chrome via `CDP_PROFILE_DIR`, preventing fallback to user's main Chrome session
8
+ - **Shared utilities** — extracted ~220 lines of duplicate code from extractors into `common.mjs` (cdp wrapper, tab management, clipboard interception)
9
+ - **Documentation leaner** — skill documentation reduced 61% (180 → 70 lines) while preserving all decision-making info
10
+
11
+ ### Notable
12
+ - **NO API KEYS** — updated messaging to emphasize this works via browser automation, no API keys needed
13
+
14
+ ## v1.4.2 (2026-03-25)
15
+
16
+ - **Fresh isolated tabs** — each search now always creates a new `about:blank` tab via `Target.createTarget` and refreshes the CDP page cache immediately after, preventing SPA navigation failures and stale DOM state from prior queries
17
+ - **Regex-based citation extraction** — all extractors (Perplexity, Bing, Gemini) now parse sources from clipboard Markdown links (`[title](url)`) instead of DOM selectors that break on UI updates
18
+ - **Relaxed verification detection** — `consent.mjs` now uses broad keyword matching (`includes('verify')`, `includes('human')`) instead of anchored regexes, correctly catching button text variants like "Verify you are human" across Cloudflare, Microsoft, and generic modals
19
+
20
+ ## v1.4.1
21
+
22
+ - **Fixed parallel synthesis** — multiple `greedy_search` calls with `synthesize: true` now run safely in parallel. Each search creates a fresh Gemini tab that gets cleaned up after synthesis, preventing tab conflicts and "Uncaught" errors.
23
+
24
+ ## v1.4.0
25
+
26
+ - **Grounded synthesis** — Gemini now receives a normalized source registry with stable source IDs, agreement summaries, caveats, and cited claims
27
+ - **Real deep research** — top sources are fetched before synthesis so deep research answers are grounded in fetched evidence, not just engine summaries
28
+ - **Richer source metadata** — source output now includes canonical URLs, domains, source types, per-engine attribution, and confidence metadata
29
+ - **Cleaner tab lifecycle** — temporary Perplexity, Bing, and Google tabs are closed after each fan-out search, and synthesis finishes on the Gemini tab
30
+ - **Isolated Chrome targeting** — GreedySearch now refuses to fall back to your normal Chrome session, preventing stray remote-debugging prompts
package/README.md CHANGED
@@ -1,20 +1,14 @@
1
1
  # GreedySearch for Pi
2
2
 
3
- Pi extension that adds a `greedy_search` tool — fans out queries to Perplexity, Bing Copilot, and Google AI simultaneously and returns AI-synthesized answers with deduped sources. Streams progress as each engine completes.
3
+ Pi extension that adds `greedy_search`, `deep_research`, and `coding_task` tools — multi-engine AI search via browser automation. **NO API KEYS needed.**
4
4
 
5
- Forked from [GreedySearch-claude](https://github.com/apmantza/GreedySearch-claude).
6
-
7
- ## What's New (v1.4.1)
5
+ Fans out queries to Perplexity, Bing Copilot, and Google AI simultaneously. Returns AI-synthesized answers with deduped sources. Streams progress as each engine completes.
8
6
 
9
- - **Fixed parallel synthesis** — multiple `greedy_search` calls with `synthesize: true` now run safely in parallel. Each search creates a fresh Gemini tab that gets cleaned up after synthesis, preventing tab conflicts and "Uncaught" errors.
7
+ Forked from [GreedySearch-claude](https://github.com/apmantza/GreedySearch-claude).
10
8
 
11
- ## What's New (v1.4.0)
9
+ ## Quick Note
12
10
 
13
- - **Grounded synthesis** — Gemini now receives a normalized source registry with stable source IDs, agreement summaries, caveats, and cited claims
14
- - **Real deep research** — top sources are fetched before synthesis so deep research answers are grounded in fetched evidence, not just engine summaries
15
- - **Richer source metadata** — source output now includes canonical URLs, domains, source types, per-engine attribution, and confidence metadata
16
- - **Cleaner tab lifecycle** — temporary Perplexity, Bing, and Google tabs are closed after each fan-out search, and synthesis finishes on the Gemini tab
17
- - **Isolated Chrome targeting** — GreedySearch now refuses to fall back to your normal Chrome session, preventing stray remote-debugging prompts
11
+ **No API keys required** — this tool uses Chrome DevTools Protocol (CDP) to interact with search engines directly through a browser. It launches its own isolated Chrome instance, so it won't interfere with your main browser session.
18
12
 
19
13
  ## Install
20
14
 
@@ -74,15 +68,7 @@ For complex research questions, use `synthesize: true` with `engine: "all"`:
74
68
  greedy_search({ query: "best auth patterns for SaaS in 2026", engine: "all", synthesize: true })
75
69
  ```
76
70
 
77
- This deduplicates sources across engines, builds a normalized source registry, and feeds that context to Gemini for one clean synthesized answer. Adds ~30s but now returns agreement summaries, caveats, key claims, and better-labeled top sources.
78
-
79
- For the most grounded mode, use deep research from the CLI:
80
-
81
- ```bash
82
- node search.mjs all "best auth patterns for SaaS in 2026" --deep-research
83
- ```
84
-
85
- Deep research fetches top source pages before synthesis and reports source confidence metadata such as agreement level, fetched-source success rate, and source mix.
71
+ This deduplicates sources across engines, builds a normalized source registry, and feeds that context to Gemini for one clean synthesized answer. Adds ~30s but returns agreement summaries, caveats, key claims, and better-labeled top sources.
86
72
 
87
73
  **Use synthesis when:**
88
74
  - You need one definitive answer, not multiple perspectives
@@ -104,21 +90,25 @@ greedy_search({ query: "explain the React compiler", engine: "perplexity", fullA
104
90
  ## Examples
105
91
 
106
92
  **Quick technical lookup:**
93
+
107
94
  ```
108
95
  greedy_search({ query: "How to use async await in Python", engine: "perplexity" })
109
96
  ```
110
97
 
111
98
  **Compare tools (see where engines agree/disagree):**
99
+
112
100
  ```
113
101
  greedy_search({ query: "Prisma vs Drizzle in 2026", engine: "all" })
114
102
  ```
115
103
 
116
104
  **Research with synthesis:**
105
+
117
106
  ```
118
107
  greedy_search({ query: "Best practices for monorepo structure", engine: "all", synthesize: true })
119
108
  ```
120
109
 
121
110
  **Debug an error:**
111
+
122
112
  ```
123
113
  greedy_search({ query: "Error: Cannot find module 'react-dom/client' Next.js 15", engine: "all" })
124
114
  ```
@@ -167,32 +157,37 @@ Tests verify:
167
157
  ## Troubleshooting
168
158
 
169
159
  ### "Chrome not found"
160
+
170
161
  Set the path explicitly:
162
+
171
163
  ```bash
172
164
  export CHROME_PATH="/path/to/chrome"
173
165
  ```
174
166
 
175
167
  ### "CDP timeout" or "Chrome may have crashed"
168
+
176
169
  Restart GreedySearch Chrome:
170
+
177
171
  ```bash
178
172
  node ~/.pi/agent/git/GreedySearch-pi/launch.mjs --kill
179
173
  node ~/.pi/agent/git/GreedySearch-pi/launch.mjs
180
174
  ```
181
175
 
182
176
  ### Google / Bing "verify you're human"
183
- The extension auto-clicks verification buttons and Cloudflare Turnstile challenges. For hard CAPTCHAs (image puzzles), solve manually in the Chrome window that opens.
177
+
178
+ The extension auto-clicks verification buttons and Cloudflare Turnstile challenges using broad keyword matching — resilient to variations like "Verify you are human" or localised button text. For hard CAPTCHAs (image puzzles), solve manually in the Chrome window that opens.
184
179
 
185
180
  ### Parallel searches failing
186
- Earlier versions shared Chrome tabs between concurrent searches, causing `ERR_ABORTED` errors. Version 1.2.0+ creates fresh tabs for each search, allowing safe parallel execution.
181
+
182
+ Each search creates a fresh isolated browser tab that is closed after completion, allowing safe parallel execution without tab state conflicts.
187
183
 
188
184
  ### Search hangs
185
+
189
186
  Chrome may be unresponsive. Restart it with `launch.mjs --kill` then `launch.mjs`.
190
187
 
191
- ### Sources are junk links
192
- This was a known issue with Gemini sources. If you're on an older version, update:
193
- ```bash
194
- pi install npm:@apmantza/greedysearch-pi
195
- ```
188
+ ### Sources are empty or junk links
189
+
190
+ Sources are now extracted by regex-parsing Markdown links (`[title](url)`) from the clipboard text captured after each engine responds — not from DOM selectors that break when the engine's UI updates. If sources are empty, the engine's clipboard copy didn't include formatted links (Bing Copilot currently falls into this category).
196
191
 
197
192
  ## How It Works
198
193
 
@@ -203,6 +198,36 @@ pi install npm:@apmantza/greedysearch-pi
203
198
  - `cdp.mjs` — Chrome DevTools Protocol CLI for browser automation
204
199
  - `skills/greedy-search/SKILL.md` — skill file that guides the model on when/how to use greedy_search
205
200
 
201
+ ## Changelog
202
+
203
+ ### v1.5.0 (2026-03-29)
204
+
205
+ - **Code extraction fixed** — `coding_task` now uses clipboard interception to preserve markdown code blocks (was losing them via DOM scraping)
206
+ - **Chrome targeting hardened** — all tools now consistently target the dedicated GreedySearch Chrome via `CDP_PROFILE_DIR`, preventing fallback to user's main Chrome session
207
+ - **Shared utilities** — extracted ~220 lines of duplicate code from extractors into `common.mjs` (cdp wrapper, tab management, clipboard interception)
208
+ - **Documentation leaner** — skill documentation reduced 61% (180 → 70 lines) while preserving all decision-making info
209
+ - **NO API KEYS** — updated messaging to emphasize this works via browser automation, no API keys needed
210
+
211
+ ### v1.4.2 (2026-03-25)
212
+
213
+ - **Fresh isolated tabs** — each search now always creates a new `about:blank` tab via `Target.createTarget` and refreshes the CDP page cache immediately after, preventing SPA navigation failures and stale DOM state from prior queries
214
+ - **Regex-based citation extraction** — all extractors (Perplexity, Bing, Gemini) now parse sources from clipboard Markdown links (`[title](url)`) instead of DOM selectors that break on UI updates
215
+ - **Relaxed verification detection** — `consent.mjs` now uses broad keyword matching (`includes('verify')`, `includes('human')`) instead of anchored regexes, correctly catching button text variants like "Verify you are human" across Cloudflare, Microsoft, and generic modals
216
+
217
+ ---
218
+
219
+ ### v1.4.1
220
+
221
+ - **Fixed parallel synthesis** — multiple `greedy_search` calls with `synthesize: true` now run safely in parallel. Each search creates a fresh Gemini tab that gets cleaned up after synthesis, preventing tab conflicts and "Uncaught" errors.
222
+
223
+ ### v1.4.0
224
+
225
+ - **Grounded synthesis** — Gemini now receives a normalized source registry with stable source IDs, agreement summaries, caveats, and cited claims
226
+ - **Real deep research** — top sources are fetched before synthesis so deep research answers are grounded in fetched evidence, not just engine summaries
227
+ - **Richer source metadata** — source output now includes canonical URLs, domains, source types, per-engine attribution, and confidence metadata
228
+ - **Cleaner tab lifecycle** — temporary Perplexity, Bing, and Google tabs are closed after each fan-out search, and synthesis finishes on the Gemini tab
229
+ - **Isolated Chrome targeting** — GreedySearch now refuses to fall back to your normal Chrome session, preventing stray remote-debugging prompts
230
+
206
231
  ## License
207
232
 
208
233
  MIT