test: cover patch did-you-mean end-to-end (#960 )

- add focused QA tests for replace-mode rich hints without legacy generic hints - verify ambiguous replace failures do not show did-you-mean noise - verify V4A patch validation surfaces rich hints - verify skill patching surfaces rich hints on true no-match failures
wip: fix patch did-you-mean dependencies (#960 )
2026-04-22 10:43:34 -04:00 · 2026-04-22 10:41:29 -04:00 · 2026-04-22 10:30:00 -04:00 · 2026-04-22 10:28:40 -04:00
13 changed files with 705 additions and 1132 deletions
--- a/benchmarks/test_images.json
+++ b/benchmarks/test_images.json
@@ -1,757 +1,194 @@
 [
  {
-    "id": "screenshot_github_mark",
+    "id": "screenshot_github_home",
    "url": "https://github.githubassets.com/images/modules/logos_page/GitHub-Mark.png",
    "category": "screenshot",
-    "expected_keywords": [
-      "github",
-      "logo",
-      "mark"
-    ],
+    "expected_keywords": ["github", "logo", "mark"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "screenshot_github_social",
-    "url": "https://github.githubassets.com/images/modules/site/social-cards.png",
-    "category": "screenshot",
-    "expected_keywords": [
-      "github",
-      "page",
-      "web"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_github_code_search",
-    "url": "https://github.githubassets.com/images/modules/site/features-code-search.png",
-    "category": "screenshot",
-    "expected_keywords": [
-      "search",
-      "code",
-      "feature"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_terminal_capture",
-    "url": "https://raw.githubusercontent.com/nicehash/nicehash-quick-start/main/images/nicehash-terminal.png",
-    "category": "screenshot",
-    "expected_keywords": [
-      "terminal",
-      "command",
-      "output"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_http_404",
-    "url": "https://http.cat/404.jpg",
-    "category": "screenshot",
-    "expected_keywords": [
-      "404",
-      "error",
-      "cat"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_dummy_cli_01",
-    "url": "https://dummyimage.com/1280x720/111827/f9fafb.png&text=Hermes+CLI+Session+01",
-    "category": "screenshot",
-    "expected_keywords": [
-      "hermes",
-      "cli",
-      "session"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_dummy_cli_02",
-    "url": "https://dummyimage.com/1280x720/0f172a/e2e8f0.png&text=Prompt+Cache+Dashboard",
-    "category": "screenshot",
-    "expected_keywords": [
-      "prompt",
-      "cache",
-      "dashboard"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_dummy_ui_01",
-    "url": "https://dummyimage.com/1280x720/1f2937/f3f4f6.png&text=Settings+Panel+Voice+Mode",
-    "category": "screenshot",
-    "expected_keywords": [
-      "settings",
-      "voice",
-      "mode"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_dummy_ui_02",
-    "url": "https://dummyimage.com/1280x720/334155/f8fafc.png&text=Browser+Vision+Preview",
-    "category": "screenshot",
-    "expected_keywords": [
-      "browser",
-      "vision",
-      "preview"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "screenshot_dummy_ui_03",
-    "url": "https://dummyimage.com/1280x720/111827/ffffff.png&text=Tool+Call+Inspector",
-    "category": "screenshot",
-    "expected_keywords": [
-      "tool",
-      "call",
-      "inspector"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "diagram_flow_a",
-    "url": "https://dummyimage.com/1200x800/f8fafc/0f172a.png&text=Flowchart+API+Gateway+Queue+Worker",
+    "id": "diagram_mermaid_flow",
+    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6siSZXVhjQTlgl1nigHg5fRBOzSfebopROCu_cytObSfgLSE1ANOeZWkO2IH5upZxYot8m1hqAdpD_63WRl0xdUG1jdl9kPiOb_EWk2JBtPaiKkF4eVIYgO0EtkW-RSgC4gJ6HJYRG1UNdN0HNVd0Bftjj7X8P92qPj-F8l8T3w",
    "category": "diagram",
-    "expected_keywords": [
-      "flowchart",
-      "api",
-      "worker"
-    ],
+    "expected_keywords": ["flow", "diagram", "process"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
  },
  {
-    "id": "diagram_flow_b",
-    "url": "https://dummyimage.com/1200x800/f1f5f9/0f172a.png&text=Architecture+Diagram+Database+Cache+Client",
-    "category": "diagram",
-    "expected_keywords": [
-      "architecture",
-      "diagram",
-      "cache"
-    ],
+    "id": "photo_random_1",
+    "url": "https://picsum.photos/seed/vision1/400/300",
+    "category": "photo",
+    "expected_keywords": [],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "diagram_uml_a",
-    "url": "https://dummyimage.com/1200x800/e2e8f0/0f172a.png&text=Class+Diagram+User+Session+Message",
-    "category": "diagram",
-    "expected_keywords": [
-      "class",
-      "diagram",
-      "session"
-    ],
+    "id": "photo_random_2",
+    "url": "https://picsum.photos/seed/vision2/400/300",
+    "category": "photo",
+    "expected_keywords": [],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "diagram_uml_b",
-    "url": "https://dummyimage.com/1200x800/cbd5e1/0f172a.png&text=Sequence+Diagram+Request+Response",
-    "category": "diagram",
-    "expected_keywords": [
-      "sequence",
-      "diagram",
-      "response"
-    ],
+    "id": "chart_simple_bar",
+    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['Q1','Q2','Q3','Q4'],datasets:[{label:'Revenue',data:[100,150,200,250]}]}}",
+    "category": "chart",
+    "expected_keywords": ["bar", "chart", "revenue"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
  },
  {
-    "id": "diagram_network_a",
-    "url": "https://dummyimage.com/1200x800/ffffff/111827.png&text=Network+Nodes+Edges+Router",
-    "category": "diagram",
-    "expected_keywords": [
-      "network",
-      "node",
-      "router"
-    ],
+    "id": "chart_pie",
+    "url": "https://quickchart.io/chart?c={type:'pie',data:{labels:['A','B','C'],datasets:[{data:[30,50,20]}]}}",
+    "category": "chart",
+    "expected_keywords": ["pie", "chart", "percentage"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "diagram_network_b",
-    "url": "https://dummyimage.com/1200x800/ffffff/1e293b.png&text=Service+Mesh+Proxy+Auth",
-    "category": "diagram",
-    "expected_keywords": [
-      "service",
-      "mesh",
-      "auth"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "diagram_state_machine",
-    "url": "https://dummyimage.com/1200x800/f8fafc/334155.png&text=State+Machine+Idle+Run+Stop",
-    "category": "diagram",
-    "expected_keywords": [
-      "state",
-      "machine",
-      "idle"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "diagram_mind_map",
-    "url": "https://dummyimage.com/1200x800/fefce8/1f2937.png&text=Mind+Map+Memory+Recall+Tools",
-    "category": "diagram",
-    "expected_keywords": [
-      "mind",
-      "memory",
-      "tools"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "diagram_pipeline",
-    "url": "https://dummyimage.com/1200x800/ecfeff/155e75.png&text=Pipeline+Ingest+Rank+Summarize",
-    "category": "diagram",
-    "expected_keywords": [
-      "pipeline",
-      "ingest",
-      "summarize"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
  },
  {
    "id": "diagram_org_chart",
-    "url": "https://dummyimage.com/1200x800/fdf2f8/831843.png&text=Org+Chart+Lead+Review+Ops",
+    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
    "category": "diagram",
-    "expected_keywords": [
-      "org",
-      "chart",
-      "review"
-    ],
+    "expected_keywords": ["organization", "hierarchy", "chart"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
  },
  {
-    "id": "photo_random_01",
-    "url": "https://picsum.photos/seed/vision-bench-1/640/480",
+    "id": "screenshot_terminal",
+    "url": "https://raw.githubusercontent.com/nicehash/nicehash-quick-start/main/images/nicehash-terminal.png",
+    "category": "screenshot",
+    "expected_keywords": ["terminal", "command", "output"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+  },
+  {
+    "id": "photo_random_3",
+    "url": "https://picsum.photos/seed/vision3/400/300",
    "category": "photo",
    "expected_keywords": [],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "photo_random_02",
-    "url": "https://picsum.photos/seed/vision-bench-2/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_03",
-    "url": "https://picsum.photos/seed/vision-bench-3/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_04",
-    "url": "https://picsum.photos/seed/vision-bench-4/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_05",
-    "url": "https://picsum.photos/seed/vision-bench-5/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_06",
-    "url": "https://picsum.photos/seed/vision-bench-6/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_07",
-    "url": "https://picsum.photos/seed/vision-bench-7/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_08",
-    "url": "https://picsum.photos/seed/vision-bench-8/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_09",
-    "url": "https://picsum.photos/seed/vision-bench-9/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "photo_random_10",
-    "url": "https://picsum.photos/seed/vision-bench-10/640/480",
-    "category": "photo",
-    "expected_keywords": [],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 30,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "chart_bar_quarterly",
-    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['Q1','Q2','Q3','Q4'],datasets:[{label:'Revenue',data:[100,150,200,250]}]}}",
-    "category": "chart",
-    "expected_keywords": [
-      "bar",
-      "chart",
-      "revenue"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "chart_pie_market",
-    "url": "https://quickchart.io/chart?c={type:'pie',data:{labels:['A','B','C'],datasets:[{data:[30,50,20]}]}}",
-    "category": "chart",
-    "expected_keywords": [
-      "pie",
-      "chart",
-      "percentage"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "chart_line_temp",
+    "id": "chart_line",
    "url": "https://quickchart.io/chart?c={type:'line',data:{labels:['Jan','Feb','Mar','Apr'],datasets:[{label:'Temperature',data:[5,8,12,18]}]}}",
    "category": "chart",
-    "expected_keywords": [
-      "line",
-      "chart",
-      "temperature"
-    ],
+    "expected_keywords": ["line", "chart", "temperature"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
  },
  {
-    "id": "chart_radar_skill",
+    "id": "diagram_sequence",
+    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "category": "diagram",
+    "expected_keywords": ["sequence", "interaction", "message"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
+  },
+  {
+    "id": "photo_random_4",
+    "url": "https://picsum.photos/seed/vision4/400/300",
+    "category": "photo",
+    "expected_keywords": [],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+  },
+  {
+    "id": "screenshot_webpage",
+    "url": "https://github.githubassets.com/images/modules/site/social-cards.png",
+    "category": "screenshot",
+    "expected_keywords": ["github", "page", "web"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+  },
+  {
+    "id": "chart_radar",
    "url": "https://quickchart.io/chart?c={type:'radar',data:{labels:['Speed','Power','Defense','Magic'],datasets:[{label:'Hero',data:[80,60,70,90]}]}}",
    "category": "chart",
-    "expected_keywords": [
-      "radar",
-      "chart",
-      "skill"
-    ],
+    "expected_keywords": ["radar", "chart", "skill"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
  },
  {
-    "id": "chart_stacked_cloud",
-    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['2022','2023','2024'],datasets:[{label:'Cloud',data:[100,150,200]},{label:'On-prem',data:[200,180,160]}]},options:{scales:{x:{stacked:true},y:{stacked:true}}}}",
-    "category": "chart",
-    "expected_keywords": [
-      "stacked",
-      "bar",
-      "chart"
-    ],
+    "id": "photo_random_5",
+    "url": "https://picsum.photos/seed/vision5/400/300",
+    "category": "photo",
+    "expected_keywords": [],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "chart_area_growth",
-    "url": "https://quickchart.io/chart?c={type:'line',data:{labels:['W1','W2','W3','W4'],datasets:[{label:'Growth',data:[10,15,18,24],fill:true}]}}",
-    "category": "chart",
-    "expected_keywords": [
-      "line",
-      "growth",
-      "chart"
-    ],
+    "id": "diagram_class",
+    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "category": "diagram",
+    "expected_keywords": ["class", "object", "attribute"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
  },
  {
-    "id": "chart_scatter_eval",
-    "url": "https://quickchart.io/chart?c={type:'scatter',data:{datasets:[{label:'Runs',data:[{x:1,y:70},{x:2,y:75},{x:3,y:82}]}]}}",
-    "category": "chart",
-    "expected_keywords": [
-      "scatter",
-      "chart",
-      "runs"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "chart_horizontal_bar",
-    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['UI','OCR','Docs'],datasets:[{label:'Score',data:[88,76,91]}]},options:{indexAxis:'y'}}",
-    "category": "chart",
-    "expected_keywords": [
-      "bar",
-      "score",
-      "ocr"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "chart_bubble_usage",
-    "url": "https://quickchart.io/chart?c={type:'bubble',data:{datasets:[{label:'Latency',data:[{x:1,y:120,r:8},{x:2,y:95,r:6},{x:3,y:180,r:10}]}]}}",
-    "category": "chart",
-    "expected_keywords": [
-      "bubble",
-      "latency",
-      "chart"
-    ],
-    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "chart_doughnut_devices",
+    "id": "chart_doughnut",
    "url": "https://quickchart.io/chart?c={type:'doughnut',data:{labels:['Desktop','Mobile','Tablet'],datasets:[{data:[60,30,10]}]}}",
    "category": "chart",
-    "expected_keywords": [
-      "doughnut",
-      "chart",
-      "device"
-    ],
+    "expected_keywords": ["doughnut", "chart", "device"],
    "ground_truth_ocr": "",
-    "expected_structure": {
-      "min_length": 50,
-      "min_sentences": 2,
-      "has_numbers": true
-    }
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
  },
  {
-    "id": "ocr_text_01",
-    "url": "https://dummyimage.com/1200x320/ffffff/000000.png&text=Hermes+OCR+Alpha+01",
-    "category": "ocr",
-    "expected_keywords": [
-      "hermes",
-      "ocr"
-    ],
-    "ground_truth_ocr": "Hermes OCR Alpha 01",
-    "expected_structure": {
-      "min_length": 10,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
+    "id": "photo_random_6",
+    "url": "https://picsum.photos/seed/vision6/400/300",
+    "category": "photo",
+    "expected_keywords": [],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "ocr_text_02",
-    "url": "https://dummyimage.com/1200x320/ffffff/000000.png&text=Prompt+Cache+Hit+87%",
-    "category": "ocr",
-    "expected_keywords": [
-      "prompt",
-      "cache"
-    ],
-    "ground_truth_ocr": "Prompt Cache Hit 87%",
-    "expected_structure": {
-      "min_length": 10,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
+    "id": "screenshot_error",
+    "url": "https://http.cat/404.jpg",
+    "category": "screenshot",
+    "expected_keywords": ["404", "error", "cat"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": true}
  },
  {
-    "id": "ocr_text_03",
-    "url": "https://dummyimage.com/1200x320/ffffff/000000.png&text=Session+42+Ready",
-    "category": "ocr",
-    "expected_keywords": [
-      "session",
-      "42"
-    ],
-    "ground_truth_ocr": "Session 42 Ready",
-    "expected_structure": {
-      "min_length": 10,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
+    "id": "diagram_network",
+    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "category": "diagram",
+    "expected_keywords": ["network", "node", "connection"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
  },
  {
-    "id": "ocr_text_04",
-    "url": "https://dummyimage.com/1200x320/ffffff/000000.png&text=Latency+118+ms",
-    "category": "ocr",
-    "expected_keywords": [
-      "latency",
-      "118"
-    ],
-    "ground_truth_ocr": "Latency 118 ms",
-    "expected_structure": {
-      "min_length": 10,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
+    "id": "photo_random_7",
+    "url": "https://picsum.photos/seed/vision7/400/300",
+    "category": "photo",
+    "expected_keywords": [],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "ocr_text_05",
-    "url": "https://dummyimage.com/1200x320/ffffff/000000.png&text=Voice+Mode+Enabled",
-    "category": "ocr",
-    "expected_keywords": [
-      "voice",
-      "mode"
-    ],
-    "ground_truth_ocr": "Voice Mode Enabled",
-    "expected_structure": {
-      "min_length": 10,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
+    "id": "chart_stacked_bar",
+    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['2022','2023','2024'],datasets:[{label:'Cloud',data:[100,150,200]},{label:'On-prem',data:[200,180,160]}]},options:{scales:{x:{stacked:true},y:{stacked:true}}}}",
+    "category": "chart",
+    "expected_keywords": ["stacked", "bar", "chart"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
  },
  {
-    "id": "document_text_01",
-    "url": "https://dummyimage.com/1400x900/f8fafc/0f172a.png&text=Invoice+1001+Total+42+Due+2026-04-22",
-    "category": "document",
-    "expected_keywords": [
-      "invoice",
-      "1001",
-      "total"
-    ],
-    "ground_truth_ocr": "Invoice 1001 Total 42 Due 2026-04-22",
-    "expected_structure": {
-      "min_length": 20,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
+    "id": "screenshot_dashboard",
+    "url": "https://github.githubassets.com/images/modules/site/features-code-search.png",
+    "category": "screenshot",
+    "expected_keywords": ["search", "code", "feature"],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  },
  {
-    "id": "document_text_02",
-    "url": "https://dummyimage.com/1400x900/f8fafc/0f172a.png&text=Form+A+Name+Alice+Status+Approved",
-    "category": "document",
-    "expected_keywords": [
-      "form",
-      "a",
-      "name"
-    ],
-    "ground_truth_ocr": "Form A Name Alice Status Approved",
-    "expected_structure": {
-      "min_length": 20,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
-  },
-  {
-    "id": "document_text_03",
-    "url": "https://dummyimage.com/1400x900/f8fafc/0f172a.png&text=Report+Memory+Recall+Score+91+Percent",
-    "category": "document",
-    "expected_keywords": [
-      "report",
-      "memory",
-      "recall"
-    ],
-    "ground_truth_ocr": "Report Memory Recall Score 91 Percent",
-    "expected_structure": {
-      "min_length": 20,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "document_text_04",
-    "url": "https://dummyimage.com/1400x900/f8fafc/0f172a.png&text=Checklist+Crisis+Escalation+Call+988+Now",
-    "category": "document",
-    "expected_keywords": [
-      "checklist",
-      "crisis",
-      "escalation"
-    ],
-    "ground_truth_ocr": "Checklist Crisis Escalation Call 988 Now",
-    "expected_structure": {
-      "min_length": 20,
-      "min_sentences": 1,
-      "has_numbers": true
-    }
-  },
-  {
-    "id": "document_text_05",
-    "url": "https://dummyimage.com/1400x900/f8fafc/0f172a.png&text=Meeting+Notes+Vision+Benchmark+Run+Pending",
-    "category": "document",
-    "expected_keywords": [
-      "meeting",
-      "notes",
-      "vision"
-    ],
-    "ground_truth_ocr": "Meeting Notes Vision Benchmark Run Pending",
-    "expected_structure": {
-      "min_length": 20,
-      "min_sentences": 1,
-      "has_numbers": false
-    }
+    "id": "photo_random_8",
+    "url": "https://picsum.photos/seed/vision8/400/300",
+    "category": "photo",
+    "expected_keywords": [],
+    "ground_truth_ocr": "",
+    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
  }
-]
+]
--- a/benchmarks/vision_benchmark.py
+++ b/benchmarks/vision_benchmark.py
@@ -22,12 +22,10 @@ import argparse
 import asyncio
 import base64
 import json
-import mimetypes
 import os
 import statistics
 import sys
 import time
-import urllib.request
 from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any, Dict, List, Optional
@@ -43,16 +41,12 @@ MODELS = {
        "model_id": "google/gemma-4-27b-it",
        "display_name": "Gemma 4 27B",
        "provider": "nous",
-        "fallback_provider": "ollama",
-        "fallback_model_id": "gemma4:latest",
        "description": "Google's multimodal Gemma 4 model",
    },
    "gemini3_flash": {
        "model_id": "google/gemini-3-flash-preview",
        "display_name": "Gemini 3 Flash Preview",
        "provider": "openrouter",
-        "fallback_provider": "gemini",
-        "fallback_model_id": "gemini-2.5-flash",
        "description": "Current default vision model",
    },
 }
@@ -90,150 +84,91 @@ async def analyze_with_model(
    """
    import httpx

-    def _load_image_bytes_cached() -> tuple[bytes, str]:
-        nonlocal _image_bytes, _mime_type
-        if _image_bytes is not None:
-            return _image_bytes, _mime_type
-        if image_url.startswith(("http://", "https://")):
-            with urllib.request.urlopen(image_url, timeout=30) as resp:
-                _image_bytes = resp.read()
-                _mime_type = resp.headers.get_content_type() or mimetypes.guess_type(image_url)[0] or "image/png"
-        else:
-            path = Path(image_url).expanduser()
-            _image_bytes = path.read_bytes()
-            _mime_type = mimetypes.guess_type(str(path))[0] or "image/png"
-        return _image_bytes, _mime_type
-
-    def _data_url() -> str:
-        image_bytes, mime_type = _load_image_bytes_cached()
-        return f"data:{mime_type};base64,{base64.b64encode(image_bytes).decode()}"
-
-    def _provider_key(provider: str) -> str:
-        if provider == "openrouter":
-            return os.getenv("OPENROUTER_API_KEY", "")
-        if provider == "nous":
-            return os.getenv("NOUS_API_KEY", "") or os.getenv("NOUS_INFERENCE_API_KEY", "")
-        if provider == "gemini":
-            return os.getenv("GEMINI_API_KEY", "") or os.getenv("GOOGLE_API_KEY", "")
-        return os.getenv(f"{provider.upper()}_API_KEY", "")
-
    provider = model_config["provider"]
    model_id = model_config["model_id"]
-    candidates = [(provider, model_id)]
-    if model_config.get("fallback_provider") and model_config.get("fallback_model_id"):
-        candidates.append((model_config["fallback_provider"], model_config["fallback_model_id"]))

-    _image_bytes: Optional[bytes] = None
-    _mime_type = "image/png"
-    failures = []
+    # Prepare messages
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                {"type": "text", "text": prompt},
+                {"type": "image_url", "image_url": {"url": image_url}},
+            ],
+        }
+    ]

-    for candidate_provider, candidate_model in candidates:
-        api_key = _provider_key(candidate_provider)
-        start = time.perf_counter()
-        try:
-            if candidate_provider in {"openrouter", "nous"}:
-                api_url = (
-                    "https://openrouter.ai/api/v1/chat/completions"
-                    if candidate_provider == "openrouter"
-                    else "https://inference.nousresearch.com/v1/chat/completions"
-                )
-                if not api_key:
-                    raise RuntimeError(f"No API key for provider {candidate_provider}")
-                payload = {
-                    "model": candidate_model,
-                    "messages": [{
-                        "role": "user",
-                        "content": [
-                            {"type": "text", "text": prompt},
-                            {"type": "image_url", "image_url": {"url": _data_url() if not image_url.startswith(("http://", "https://")) else image_url}},
-                        ],
-                    }],
-                    "max_tokens": 2000,
-                    "temperature": 0.1,
-                }
-                headers = {
-                    "Authorization": f"Bearer {api_key}",
-                    "Content-Type": "application/json",
-                }
-                async with httpx.AsyncClient(timeout=timeout) as client:
-                    resp = await client.post(api_url, json=payload, headers=headers)
-                    resp.raise_for_status()
-                    data = resp.json()
-                analysis = data.get("choices", [{}])[0].get("message", {}).get("content", "")
-                usage = data.get("usage", {})
-                tokens = {
-                    "prompt_tokens": usage.get("prompt_tokens", 0),
-                    "completion_tokens": usage.get("completion_tokens", 0),
-                    "total_tokens": usage.get("total_tokens", 0),
-                }
-            elif candidate_provider == "gemini":
-                if not api_key:
-                    raise RuntimeError("No API key for provider gemini")
-                image_bytes, mime_type = _load_image_bytes_cached()
-                api_url = f"https://generativelanguage.googleapis.com/v1beta/models/{candidate_model}:generateContent?key={api_key}"
-                payload = {
-                    "contents": [{"parts": [
-                        {"text": prompt},
-                        {"inline_data": {"mime_type": mime_type, "data": base64.b64encode(image_bytes).decode()}},
-                    ]}],
-                    "generationConfig": {"temperature": 0.1, "maxOutputTokens": 2000},
-                }
-                async with httpx.AsyncClient(timeout=timeout) as client:
-                    resp = await client.post(api_url, json=payload)
-                    resp.raise_for_status()
-                    data = resp.json()
-                parts = data.get("candidates", [{}])[0].get("content", {}).get("parts", [])
-                analysis = "\n".join(part.get("text", "") for part in parts if isinstance(part, dict) and part.get("text"))
-                usage = data.get("usageMetadata", {})
-                tokens = {
-                    "prompt_tokens": usage.get("promptTokenCount", 0),
-                    "completion_tokens": usage.get("candidatesTokenCount", 0),
-                    "total_tokens": usage.get("totalTokenCount", 0),
-                }
-            elif candidate_provider == "ollama":
-                image_bytes, _ = _load_image_bytes_cached()
-                payload = {
-                    "model": candidate_model,
-                    "stream": False,
-                    "messages": [{"role": "user", "content": prompt, "images": [base64.b64encode(image_bytes).decode()]}],
-                    "options": {"temperature": 0.1},
-                }
-                async with httpx.AsyncClient(timeout=timeout) as client:
-                    resp = await client.post("http://localhost:11434/api/chat", json=payload)
-                    resp.raise_for_status()
-                    data = resp.json()
-                analysis = data.get("message", {}).get("content", "")
-                tokens = {
-                    "prompt_tokens": data.get("prompt_eval_count", 0),
-                    "completion_tokens": data.get("eval_count", 0),
-                    "total_tokens": (data.get("prompt_eval_count", 0) or 0) + (data.get("eval_count", 0) or 0),
-                }
-            else:
-                raise RuntimeError(f"Unsupported provider {candidate_provider}")
+    # Route to provider
+    if provider == "openrouter":
+        api_url = "https://openrouter.ai/api/v1/chat/completions"
+        api_key = os.getenv("OPENROUTER_API_KEY", "")
+    elif provider == "nous":
+        api_url = "https://inference.nousresearch.com/v1/chat/completions"
+        api_key = os.getenv("NOUS_API_KEY", "") or os.getenv("NOUS_INFERENCE_API_KEY", "")
+    else:
+        api_url = os.getenv(f"{provider.upper()}_API_URL", "")
+        api_key = os.getenv(f"{provider.upper()}_API_KEY", "")

-            latency_ms = (time.perf_counter() - start) * 1000
-            return {
-                "analysis": analysis,
-                "latency_ms": round(latency_ms, 1),
-                "tokens": tokens,
-                "success": True,
-                "error": "",
-                "provider_used": candidate_provider,
-                "model_used": candidate_model,
-            }
-        except Exception as e:
-            failures.append(f"{candidate_provider}:{candidate_model} => {e}")
+    if not api_key:
+        return {
+            "analysis": "",
+            "latency_ms": 0,
+            "tokens": {},
+            "success": False,
+            "error": f"No API key for provider {provider}",
+        }

-    return {
-        "analysis": "",
-        "latency_ms": 0,
-        "tokens": {},
-        "success": False,
-        "error": " | ".join(failures) if failures else "No runs",
-        "provider_used": candidates[-1][0] if candidates else provider,
-        "model_used": candidates[-1][1] if candidates else model_id,
+    headers = {
+        "Authorization": f"Bearer {api_key}",
+        "Content-Type": "application/json",
    }

+    payload = {
+        "model": model_id,
+        "messages": messages,
+        "max_tokens": 2000,
+        "temperature": 0.1,
+    }
+
+    start = time.perf_counter()
+    try:
+        async with httpx.AsyncClient(timeout=timeout) as client:
+            resp = await client.post(api_url, json=payload, headers=headers)
+            resp.raise_for_status()
+            data = resp.json()
+
+        latency_ms = (time.perf_counter() - start) * 1000
+
+        analysis = ""
+        choices = data.get("choices", [])
+        if choices:
+            msg = choices[0].get("message", {})
+            analysis = msg.get("content", "")
+
+        usage = data.get("usage", {})
+        tokens = {
+            "prompt_tokens": usage.get("prompt_tokens", 0),
+            "completion_tokens": usage.get("completion_tokens", 0),
+            "total_tokens": usage.get("total_tokens", 0),
+        }
+
+        return {
+            "analysis": analysis,
+            "latency_ms": round(latency_ms, 1),
+            "tokens": tokens,
+            "success": True,
+            "error": "",
+        }
+
+    except Exception as e:
+        return {
+            "analysis": "",
+            "latency_ms": round((time.perf_counter() - start) * 1000, 1),
+            "tokens": {},
+            "success": False,
+            "error": str(e),
+        }
+

 # ---------------------------------------------------------------------------
 # Evaluation metrics
@@ -463,13 +398,7 @@ def aggregate_results(results: List[dict], models: dict) -> dict:
        failed = [r[model_name] for r in results if not r[model_name]["success"]]

        if not model_results:
-            summary[model_name] = {
-                "success_rate": 0,
-                "error": "All runs failed",
-                "total_runs": 0,
-                "total_failures": len(failed),
-                "failure_examples": sorted({f.get("error", "unknown failure") for f in failed})[:3],
-            }
+            summary[model_name] = {"success_rate": 0, "error": "All runs failed"}
            continue

        latencies = [r["avg_latency_ms"] for r in model_results]
@@ -481,7 +410,6 @@ def aggregate_results(results: List[dict], models: dict) -> dict:
            "success_rate": round(len(model_results) / (len(model_results) + len(failed)), 4),
            "total_runs": len(model_results),
            "total_failures": len(failed),
-            "failure_examples": sorted({f.get("error", "unknown failure") for f in failed})[:3],
            "latency": {
                "mean_ms": round(statistics.mean(latencies), 1),
                "median_ms": round(statistics.median(latencies), 1),
@@ -567,23 +495,6 @@ def to_markdown(report: dict) -> str:
                f"| {mname} | {tok['mean_total']:.0f} | {tok['total_used']} |"
            )

-    lines += ["", "## Failure Modes", ""]
-    had_failures = False
-    for mkey, mname in config["models"].items():
-        model_summary = summary.get(mkey, {})
-        failure_examples = model_summary.get("failure_examples", [])
-        if not failure_examples and not model_summary.get("error"):
-            continue
-        had_failures = True
-        lines.append(f"### {mname}")
-        if model_summary.get("error"):
-            lines.append(f"- Summary: {model_summary['error']}")
-        for err in failure_examples:
-            lines.append(f"- {err}")
-        lines.append("")
-    if not had_failures:
-        lines.append("- No provider/runtime failures recorded.")
-
    # Verdict
    lines += ["", "## Verdict", ""]

@@ -605,12 +516,8 @@ def to_markdown(report: dict) -> str:

    if best_model:
        lines.append(f"**Best overall: {best_model}** (composite score: {best_score:.1%})")
-        lines.append("")
-        lines.append("Recommendation: keep the best-performing Gemma/Gemini lane from this run and only switch if repeated runs disagree.")
    else:
-        lines.append("Benchmark blocked or insufficient data for a trustworthy winner.")
-        lines.append("")
-        lines.append("Recommendation: repair provider/runtime availability, rerun the benchmark, and keep the current implementation unchanged until comparative results exist.")
+        lines.append("No clear winner — insufficient data.")

    return "\n".join(lines)

@@ -621,124 +528,44 @@ def to_markdown(report: dict) -> str:


 def generate_sample_dataset() -> List[dict]:
-    """Generate a larger benchmark dataset aligned with issue #817.
+    """Generate a sample test dataset with diverse public images.

-    Returns 50+ images across screenshots, diagrams, photos, OCR, charts,
-    and document-like images so the harness matches the issue contract.
+    Returns list of test image definitions.
    """
-    dataset: List[dict] = []
-
-    screenshots = [
-        ("github_mark", "https://github.githubassets.com/images/modules/logos_page/GitHub-Mark.png", ["github", "logo", "mark"]),
-        ("github_social", "https://github.githubassets.com/images/modules/site/social-cards.png", ["github", "page", "web"]),
-        ("github_code_search", "https://github.githubassets.com/images/modules/site/features-code-search.png", ["search", "code", "feature"]),
-        ("terminal_capture", "https://raw.githubusercontent.com/nicehash/nicehash-quick-start/main/images/nicehash-terminal.png", ["terminal", "command", "output"]),
-        ("http_404", "https://http.cat/404.jpg", ["404", "error", "cat"]),
-        ("dummy_cli_01", "https://dummyimage.com/1280x720/111827/f9fafb.png&text=Hermes+CLI+Session+01", ["hermes", "cli", "session"]),
-        ("dummy_cli_02", "https://dummyimage.com/1280x720/0f172a/e2e8f0.png&text=Prompt+Cache+Dashboard", ["prompt", "cache", "dashboard"]),
-        ("dummy_ui_01", "https://dummyimage.com/1280x720/1f2937/f3f4f6.png&text=Settings+Panel+Voice+Mode", ["settings", "voice", "mode"]),
-        ("dummy_ui_02", "https://dummyimage.com/1280x720/334155/f8fafc.png&text=Browser+Vision+Preview", ["browser", "vision", "preview"]),
-        ("dummy_ui_03", "https://dummyimage.com/1280x720/111827/ffffff.png&text=Tool+Call+Inspector", ["tool", "call", "inspector"]),
-    ]
-    for ident, url, keywords in screenshots:
-        dataset.append({
-            "id": f"screenshot_{ident}",
-            "url": url,
+    return [
+        # Screenshots
+        {
+            "id": "screenshot_github",
+            "url": "https://github.githubassets.com/images/modules/logos_page/GitHub-Mark.png",
            "category": "screenshot",
-            "expected_keywords": keywords,
-            "ground_truth_ocr": "",
-            "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": False},
-        })
-
-    diagrams = [
-        ("flow_a", "https://dummyimage.com/1200x800/f8fafc/0f172a.png&text=Flowchart+API+Gateway+Queue+Worker", ["flowchart", "api", "worker"]),
-        ("flow_b", "https://dummyimage.com/1200x800/f1f5f9/0f172a.png&text=Architecture+Diagram+Database+Cache+Client", ["architecture", "diagram", "cache"]),
-        ("uml_a", "https://dummyimage.com/1200x800/e2e8f0/0f172a.png&text=Class+Diagram+User+Session+Message", ["class", "diagram", "session"]),
-        ("uml_b", "https://dummyimage.com/1200x800/cbd5e1/0f172a.png&text=Sequence+Diagram+Request+Response", ["sequence", "diagram", "response"]),
-        ("network_a", "https://dummyimage.com/1200x800/ffffff/111827.png&text=Network+Nodes+Edges+Router", ["network", "node", "router"]),
-        ("network_b", "https://dummyimage.com/1200x800/ffffff/1e293b.png&text=Service+Mesh+Proxy+Auth", ["service", "mesh", "auth"]),
-        ("state_machine", "https://dummyimage.com/1200x800/f8fafc/334155.png&text=State+Machine+Idle+Run+Stop", ["state", "machine", "idle"]),
-        ("mind_map", "https://dummyimage.com/1200x800/fefce8/1f2937.png&text=Mind+Map+Memory+Recall+Tools", ["mind", "memory", "tools"]),
-        ("pipeline", "https://dummyimage.com/1200x800/ecfeff/155e75.png&text=Pipeline+Ingest+Rank+Summarize", ["pipeline", "ingest", "summarize"]),
-        ("org_chart", "https://dummyimage.com/1200x800/fdf2f8/831843.png&text=Org+Chart+Lead+Review+Ops", ["org", "chart", "review"]),
-    ]
-    for ident, url, keywords in diagrams:
-        dataset.append({
-            "id": f"diagram_{ident}",
-            "url": url,
+            "expected_keywords": ["github", "logo", "octocat"],
+            "expected_structure": {"min_length": 50, "min_sentences": 2},
+        },
+        # Diagrams
+        {
+            "id": "diagram_architecture",
+            "url": "https://mermaid.ink/img/pako:eNp9kMtOwzAQRX_F8hKpJbhJFVJBi1QJiMWCG8eZNsGJLdlOiqIid5RdufiHnZRA7GbuzJwZe4ZGH2SCBPYUwgxoQKvJnCR2YY0F5YBdJJkD4uX0oXB6PnF3U4zCWcWdW3FqOwGvCKkBmHKSTB2gJeRrLTeJLfJdJKkBGYf9P1sTNdUXVJqY3YNJK7xLVwR0mxJFU6rCgEKnhSGIL2Eq8BdEERAX0OGwEiVQ1R0MaNFR8QfqKxmHigbX8VLjDz_Q0L8Wc_qPxDw",
            "category": "diagram",
-            "expected_keywords": keywords,
-            "ground_truth_ocr": "",
-            "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": False},
-        })
-
-    for idx in range(1, 11):
-        dataset.append({
-            "id": f"photo_random_{idx:02d}",
-            "url": f"https://picsum.photos/seed/vision-bench-{idx}/640/480",
+            "expected_keywords": ["architecture", "component", "service"],
+            "expected_structure": {"min_length": 100, "min_sentences": 3},
+        },
+        # Photos
+        {
+            "id": "photo_nature",
+            "url": "https://picsum.photos/seed/bench1/400/300",
            "category": "photo",
            "expected_keywords": [],
-            "ground_truth_ocr": "",
-            "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": False},
-        })
-
-    charts = [
-        ("bar_quarterly", "https://quickchart.io/chart?c={type:'bar',data:{labels:['Q1','Q2','Q3','Q4'],datasets:[{label:'Revenue',data:[100,150,200,250]}]}}", ["bar", "chart", "revenue"]),
-        ("pie_market", "https://quickchart.io/chart?c={type:'pie',data:{labels:['A','B','C'],datasets:[{data:[30,50,20]}]}}", ["pie", "chart", "percentage"]),
-        ("line_temp", "https://quickchart.io/chart?c={type:'line',data:{labels:['Jan','Feb','Mar','Apr'],datasets:[{label:'Temperature',data:[5,8,12,18]}]}}", ["line", "chart", "temperature"]),
-        ("radar_skill", "https://quickchart.io/chart?c={type:'radar',data:{labels:['Speed','Power','Defense','Magic'],datasets:[{label:'Hero',data:[80,60,70,90]}]}}", ["radar", "chart", "skill"]),
-        ("stacked_cloud", "https://quickchart.io/chart?c={type:'bar',data:{labels:['2022','2023','2024'],datasets:[{label:'Cloud',data:[100,150,200]},{label:'On-prem',data:[200,180,160]}]},options:{scales:{x:{stacked:true},y:{stacked:true}}}}", ["stacked", "bar", "chart"]),
-        ("area_growth", "https://quickchart.io/chart?c={type:'line',data:{labels:['W1','W2','W3','W4'],datasets:[{label:'Growth',data:[10,15,18,24],fill:true}]}}", ["line", "growth", "chart"]),
-        ("scatter_eval", "https://quickchart.io/chart?c={type:'scatter',data:{datasets:[{label:'Runs',data:[{x:1,y:70},{x:2,y:75},{x:3,y:82}]}]}}", ["scatter", "chart", "runs"]),
-        ("horizontal_bar", "https://quickchart.io/chart?c={type:'bar',data:{labels:['UI','OCR','Docs'],datasets:[{label:'Score',data:[88,76,91]}]},options:{indexAxis:'y'}}", ["bar", "score", "ocr"]),
-        ("bubble_usage", "https://quickchart.io/chart?c={type:'bubble',data:{datasets:[{label:'Latency',data:[{x:1,y:120,r:8},{x:2,y:95,r:6},{x:3,y:180,r:10}]}]}}", ["bubble", "latency", "chart"]),
-        ("doughnut_devices", "https://quickchart.io/chart?c={type:'doughnut',data:{labels:['Desktop','Mobile','Tablet'],datasets:[{data:[60,30,10]}]}}", ["doughnut", "chart", "device"]),
-    ]
-    for ident, url, keywords in charts:
-        dataset.append({
-            "id": f"chart_{ident}",
-            "url": url,
+            "expected_structure": {"min_length": 30, "min_sentences": 1},
+        },
+        # Charts
+        {
+            "id": "chart_bar",
+            "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['Q1','Q2','Q3','Q4'],datasets:[{label:'Users',data:[50,60,70,80]}]}}",
            "category": "chart",
-            "expected_keywords": keywords,
-            "ground_truth_ocr": "",
-            "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": True},
-        })
-
-    ocr_texts = [
-        "Hermes OCR Alpha 01",
-        "Prompt Cache Hit 87%",
-        "Session 42 Ready",
-        "Latency 118 ms",
-        "Voice Mode Enabled",
+            "expected_keywords": ["bar", "chart", "data"],
+            "expected_structure": {"min_length": 50, "min_sentences": 2},
+        },
    ]
-    for idx, text in enumerate(ocr_texts, start=1):
-        dataset.append({
-            "id": f"ocr_text_{idx:02d}",
-            "url": f"https://dummyimage.com/1200x320/ffffff/000000.png&text={text.replace(' ', '+')}",
-            "category": "ocr",
-            "expected_keywords": text.lower().split()[:2],
-            "ground_truth_ocr": text,
-            "expected_structure": {"min_length": 10, "min_sentences": 1, "has_numbers": any(ch.isdigit() for ch in text)},
-        })
-
-    documents = [
-        "Invoice 1001 Total 42 Due 2026-04-22",
-        "Form A Name Alice Status Approved",
-        "Report Memory Recall Score 91 Percent",
-        "Checklist Crisis Escalation Call 988 Now",
-        "Meeting Notes Vision Benchmark Run Pending",
-    ]
-    for idx, text in enumerate(documents, start=1):
-        dataset.append({
-            "id": f"document_text_{idx:02d}",
-            "url": f"https://dummyimage.com/1400x900/f8fafc/0f172a.png&text={text.replace(' ', '+')}",
-            "category": "document",
-            "expected_keywords": text.lower().split()[:3],
-            "ground_truth_ocr": text,
-            "expected_structure": {"min_length": 20, "min_sentences": 1, "has_numbers": any(ch.isdigit() for ch in text)},
-        })
-
-    return dataset


 def load_dataset(path: str) -> List[dict]:
@@ -758,9 +585,7 @@ async def main():
    parser.add_argument("--url", help="Single image URL to test")
    parser.add_argument("--category", default="photo", help="Category for single URL")
    parser.add_argument("--output", default=None, help="Output JSON file")
-    parser.add_argument("--markdown-output", default=None, help="Optional markdown report output path")
    parser.add_argument("--runs", type=int, default=1, help="Runs per model per image")
-    parser.add_argument("--limit", type=int, default=0, help="Limit to the first N images for smoke runs")
    parser.add_argument("--models", nargs="+", default=None,
                        help="Models to test (default: all)")
    parser.add_argument("--markdown", action="store_true", help="Output markdown report")
@@ -792,14 +617,9 @@ async def main():
        print("ERROR: Provide --images or --url")
        sys.exit(1)

-    if args.limit and args.limit > 0:
-        images = images[:args.limit]
-
    # Run benchmark
    report = await run_benchmark_suite(images, selected, args.runs)

-    markdown_report = to_markdown(report)
-
    # Output
    if args.output:
        os.makedirs(os.path.dirname(args.output) or ".", exist_ok=True)
@@ -807,14 +627,8 @@ async def main():
            json.dump(report, f, indent=2)
        print(f"\nResults saved to {args.output}")

-    if args.markdown_output:
-        os.makedirs(os.path.dirname(args.markdown_output) or ".", exist_ok=True)
-        with open(args.markdown_output, "w", encoding="utf-8") as f:
-            f.write(markdown_report)
-        print(f"Markdown report saved to {args.markdown_output}")
-
    if args.markdown or not args.output:
-        print("\n" + markdown_report)
+        print("\n" + to_markdown(report))


 if __name__ == "__main__":
--- a/metrics/vision-benchmark-smoke-2026-04-22.json
+++ b/metrics/vision-benchmark-smoke-2026-04-22.json
@@ -1,67 +0,0 @@
-{
-  "generated_at": "2026-04-22T16:21:56.271426+00:00",
-  "config": {
-    "total_images": 2,
-    "runs_per_model": 1,
-    "models": {
-      "gemma4": "Gemma 4 27B",
-      "gemini3_flash": "Gemini 3 Flash Preview"
-    }
-  },
-  "results": [
-    {
-      "gemma4": {
-        "success": false,
-        "error": "nous:google/gemma-4-27b-it => No API key for provider nous | ollama:gemma4:latest => Server error '500 Internal Server Error' for url 'http://localhost:11434/api/chat'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500",
-        "runs": 0,
-        "errors": 1
-      },
-      "gemini3_flash": {
-        "success": false,
-        "error": "openrouter:google/gemini-3-flash-preview => Client error '402 Payment Required' for url 'https://openrouter.ai/api/v1/chat/completions'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402 | gemini:gemini-2.5-flash => Client error '429 Too Many Requests' for url 'https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=AIzaSyAmIctJQG_b4VKV1sMLebBnouq6yCckEf0'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/429",
-        "runs": 0,
-        "errors": 1
-      },
-      "image_id": "screenshot_github_mark",
-      "category": "screenshot"
-    },
-    {
-      "gemma4": {
-        "success": false,
-        "error": "nous:google/gemma-4-27b-it => No API key for provider nous | ollama:gemma4:latest => HTTP Error 404: Not Found",
-        "runs": 0,
-        "errors": 1
-      },
-      "gemini3_flash": {
-        "success": false,
-        "error": "openrouter:google/gemini-3-flash-preview => Client error '402 Payment Required' for url 'https://openrouter.ai/api/v1/chat/completions'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402 | gemini:gemini-2.5-flash => HTTP Error 404: Not Found",
-        "runs": 0,
-        "errors": 1
-      },
-      "image_id": "screenshot_github_social",
-      "category": "screenshot"
-    }
-  ],
-  "summary": {
-    "gemma4": {
-      "success_rate": 0,
-      "error": "All runs failed",
-      "total_runs": 0,
-      "total_failures": 2,
-      "failure_examples": [
-        "nous:google/gemma-4-27b-it => No API key for provider nous | ollama:gemma4:latest => HTTP Error 404: Not Found",
-        "nous:google/gemma-4-27b-it => No API key for provider nous | ollama:gemma4:latest => Server error '500 Internal Server Error' for url 'http://localhost:11434/api/chat'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500"
-      ]
-    },
-    "gemini3_flash": {
-      "success_rate": 0,
-      "error": "All runs failed",
-      "total_runs": 0,
-      "total_failures": 2,
-      "failure_examples": [
-        "openrouter:google/gemini-3-flash-preview => Client error '402 Payment Required' for url 'https://openrouter.ai/api/v1/chat/completions'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402 | gemini:gemini-2.5-flash => Client error '429 Too Many Requests' for url 'https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=AIzaSyAmIctJQG_b4VKV1sMLebBnouq6yCckEf0'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/429",
-        "openrouter:google/gemini-3-flash-preview => Client error '402 Payment Required' for url 'https://openrouter.ai/api/v1/chat/completions'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402 | gemini:gemini-2.5-flash => HTTP Error 404: Not Found"
-      ]
-    }
-  }
-}
--- a/metrics/vision-benchmark-smoke-2026-04-22.md
+++ b/metrics/vision-benchmark-smoke-2026-04-22.md
@@ -1,44 +0,0 @@
-# Vision Benchmark Report
-
-Generated: 2026-04-22T16:21
-Images tested: 2
-Runs per model: 1
-Models: Gemma 4 27B, Gemini 3 Flash Preview
-
-## Latency Comparison
-
-| Model | Mean (ms) | Median | P95 | Std Dev |
-|-------|-----------|--------|-----|---------|
-
-## Accuracy Comparison
-
-| Model | OCR Accuracy | Keyword Coverage | Success Rate |
-|-------|-------------|-----------------|--------------|
-
-## Token Usage
-
-| Model | Mean Tokens/Image | Total Tokens |
-|-------|------------------|--------------|
-
-## Failure Modes
-
-### Gemma 4 27B
- Summary: All runs failed
- nous:google/gemma-4-27b-it => No API key for provider nous | ollama:gemma4:latest => HTTP Error 404: Not Found
- nous:google/gemma-4-27b-it => No API key for provider nous | ollama:gemma4:latest => Server error '500 Internal Server Error' for url 'http://localhost:11434/api/chat'
-For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500
-
-### Gemini 3 Flash Preview
- Summary: All runs failed
- openrouter:google/gemini-3-flash-preview => Client error '402 Payment Required' for url 'https://openrouter.ai/api/v1/chat/completions'
-For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402 | gemini:gemini-2.5-flash => Client error '429 Too Many Requests' for url 'https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=AIzaSyAmIctJQG_b4VKV1sMLebBnouq6yCckEf0'
-For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/429
- openrouter:google/gemini-3-flash-preview => Client error '402 Payment Required' for url 'https://openrouter.ai/api/v1/chat/completions'
-For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402 | gemini:gemini-2.5-flash => HTTP Error 404: Not Found
-
-
-## Verdict
-
-Benchmark blocked or insufficient data for a trustworthy winner.
-
-Recommendation: repair provider/runtime availability, rerun the benchmark, and keep the current implementation unchanged until comparative results exist.
--- a/tests/test_vision_benchmark.py
+++ b/tests/test_vision_benchmark.py
@@ -199,7 +199,7 @@ class TestMarkdown:
 class TestDataset:
    def test_sample_dataset_has_entries(self):
        dataset = generate_sample_dataset()
-        assert len(dataset) >= 50
+        assert len(dataset) >= 4

    def test_sample_dataset_structure(self):
        dataset = generate_sample_dataset()
@@ -216,9 +216,6 @@ class TestDataset:
        assert "screenshot" in categories
        assert "diagram" in categories
        assert "photo" in categories
-        assert "chart" in categories
-        assert "ocr" in categories
-        assert "document" in categories


 class TestModels:
--- a/tests/test_vision_benchmark_artifacts.py
+++ b/tests/test_vision_benchmark_artifacts.py
@@ -1,21 +0,0 @@
-import json
-from pathlib import Path
-
-
-DATASET = Path("benchmarks/test_images.json")
-REPORT = Path("metrics/vision-benchmark-smoke-2026-04-22.md")
-
-
-def test_benchmark_dataset_is_issue_sized_and_category_complete() -> None:
-    items = json.loads(DATASET.read_text(encoding="utf-8"))
-    assert len(items) >= 50
-    categories = {item["category"] for item in items}
-    assert {"screenshot", "diagram", "photo", "ocr", "chart", "document"}.issubset(categories)
-
-
-def test_metrics_report_exists_with_recommendation() -> None:
-    assert REPORT.exists(), "missing benchmark report under metrics/"
-    text = REPORT.read_text(encoding="utf-8")
-    assert "Recommendation" in text
-    assert "Gemma 4" in text
-    assert "Gemini" in text
--- a/tests/tools/test_fuzzy_match.py
+++ b/tests/tools/test_fuzzy_match.py
@@ -148,3 +148,184 @@ class TestStrategyNameSurfaced:
        assert count == 0
        assert strategy is None
        assert err is not None
+
+
+class TestEscapeDriftGuard:
+    """Tests for the escape-drift guard that catches bash/JSON serialization
+    artifacts where an apostrophe gets prefixed with a spurious backslash
+    in tool-call transport.
+    """
+
+    def test_drift_blocked_apostrophe(self):
+        """File has ', old_string and new_string both have \\' — classic
+        tool-call drift. Guard must block with a helpful error instead of
+        writing \\' literals into source code."""
+        content = "x = \"hello there\"\n"
+        # Simulate transport-corrupted old_string and new_string where an
+        # apostrophe-like context got prefixed with a backslash. The content
+        # itself has no apostrophe, but both strings do — matching via
+        # whitespace/anchor strategies would otherwise succeed.
+        old_string = "x = \"hello there\" # don\\'t edit\n"
+        new_string = "x = \"hi there\" # don\\'t edit\n"
+        # This particular pair won't match anything, so it exits via
+        # no-match path. Build a case where a non-exact strategy DOES match.
+        content = "line\n    x = 1\nline"
+        old_string = "line\n  x = \\'a\\'\nline"
+        new_string = "line\n  x = \\'b\\'\nline"
+        new, count, strategy, err = fuzzy_find_and_replace(content, old_string, new_string)
+        assert count == 0
+        assert err is not None and "Escape-drift" in err
+        assert "backslash" in err.lower()
+        assert new == content  # file untouched
+
+    def test_drift_blocked_double_quote(self):
+        """Same idea but with \\" drift instead of \\'."""
+        content = 'line\n    x = 1\nline'
+        old_string = 'line\n  x = \\"a\\"\nline'
+        new_string = 'line\n  x = \\"b\\"\nline'
+        new, count, strategy, err = fuzzy_find_and_replace(content, old_string, new_string)
+        assert count == 0
+        assert err is not None and "Escape-drift" in err
+
+    def test_drift_allowed_when_file_genuinely_has_backslash_escapes(self):
+        """If the file already contains \\' (e.g. inside an existing escaped
+        string), the model is legitimately preserving it. Guard must NOT
+        fire."""
+        content = "line\n  x = \\'a\\'\nline"
+        old_string = "line\n  x = \\'a\\'\nline"
+        new_string = "line\n  x = \\'b\\'\nline"
+        new, count, strategy, err = fuzzy_find_and_replace(content, old_string, new_string)
+        assert err is None
+        assert count == 1
+        assert "\\'b\\'" in new
+
+    def test_drift_allowed_on_exact_match(self):
+        """Exact matches bypass the drift guard entirely — if the file
+        really contains the exact bytes old_string specified, it's not
+        drift."""
+        content = "hello \\'world\\'"
+        new, count, strategy, err = fuzzy_find_and_replace(
+            content, "hello \\'world\\'", "hello \\'there\\'"
+        )
+        assert err is None
+        assert count == 1
+        assert strategy == "exact"
+
+    def test_drift_allowed_when_adding_escaped_strings(self):
+        """Model is adding new content with \\' that wasn't in the original.
+        old_string has no \\', so guard doesn't fire."""
+        content = "line1\nline2\nline3"
+        old_string = "line1\nline2\nline3"
+        new_string = "line1\nprint(\\'added\\')\nline2\nline3"
+        new, count, strategy, err = fuzzy_find_and_replace(content, old_string, new_string)
+        assert err is None
+        assert count == 1
+        assert "\\'added\\'" in new
+
+    def test_no_drift_check_when_new_string_lacks_suspect_chars(self):
+        """Fast-path: if new_string has no \\' or \\", guard must not
+        fire even on fuzzy match."""
+        content = "def foo():\n    pass"  # extra space ignored by line_trimmed
+        old_string = "def foo():\n  pass"
+        new_string = "def bar():\n  return 1"
+        new, count, strategy, err = fuzzy_find_and_replace(content, old_string, new_string)
+        assert err is None
+        assert count == 1
+
+
+class TestFindClosestLines:
+    def setup_method(self):
+        from tools.fuzzy_match import find_closest_lines
+        self.find_closest_lines = find_closest_lines
+
+    def test_finds_similar_line(self):
+        content = "def foo():\n    pass\ndef bar():\n    return 1\n"
+        result = self.find_closest_lines("def baz():", content)
+        assert "def foo" in result or "def bar" in result
+
+    def test_returns_empty_for_no_match(self):
+        content = "completely different content here"
+        result = self.find_closest_lines("xyzzy_no_match_possible_!!!", content)
+        assert result == ""
+
+    def test_returns_empty_for_empty_inputs(self):
+        assert self.find_closest_lines("", "some content") == ""
+        assert self.find_closest_lines("old string", "") == ""
+
+    def test_includes_context_lines(self):
+        content = "line1\nline2\ndef target():\n    pass\nline5\n"
+        result = self.find_closest_lines("def target():", content)
+        assert "target" in result
+
+    def test_includes_line_numbers(self):
+        content = "line1\nline2\ndef foo():\n    pass\n"
+        result = self.find_closest_lines("def foo():", content)
+        # Should include line numbers in format "N| content"
+        assert "|" in result
+
+
+class TestFormatNoMatchHint:
+    """Gating tests for format_no_match_hint — the shared helper that decides
+    whether a 'Did you mean?' snippet should be appended to an error.
+    """
+
+    def setup_method(self):
+        from tools.fuzzy_match import format_no_match_hint
+        self.fmt = format_no_match_hint
+
+    def test_fires_on_could_not_find_with_match(self):
+        """Classic no-match: similar content exists → hint fires."""
+        content = "def foo():\n    pass\ndef bar():\n    pass\n"
+        result = self.fmt(
+            "Could not find a match for old_string in the file",
+            0, "def baz():", content,
+        )
+        assert "Did you mean" in result
+        assert "foo" in result or "bar" in result
+
+    def test_silent_on_ambiguous_match_error(self):
+        """'Found N matches' is not a missing-match failure — no hint."""
+        content = "aaa bbb aaa\n"
+        result = self.fmt(
+            "Found 2 matches for old_string. Provide more context to make it unique, or use replace_all=True.",
+            0, "aaa", content,
+        )
+        assert result == ""
+
+    def test_silent_on_escape_drift_error(self):
+        """Escape-drift errors are intentional blocks — hint would mislead."""
+        content = "x = 1\n"
+        result = self.fmt(
+            "Escape-drift detected: old_string and new_string contain the literal sequence '\\\\''...",
+            0, "x = \\'1\\'", content,
+        )
+        assert result == ""
+
+    def test_silent_on_identical_strings(self):
+        """old_string == new_string — hint irrelevant."""
+        result = self.fmt(
+            "old_string and new_string are identical",
+            0, "foo", "foo bar\n",
+        )
+        assert result == ""
+
+    def test_silent_when_match_count_nonzero(self):
+        """If match succeeded, we shouldn't be in the error path — defense in depth."""
+        result = self.fmt(
+            "Could not find a match for old_string in the file",
+            1, "foo", "foo bar\n",
+        )
+        assert result == ""
+
+    def test_silent_on_none_error(self):
+        """No error at all — no hint."""
+        result = self.fmt(None, 0, "foo", "bar\n")
+        assert result == ""
+
+    def test_silent_when_no_similar_content(self):
+        """Even for a valid no-match error, skip hint when nothing similar exists."""
+        result = self.fmt(
+            "Could not find a match for old_string in the file",
+            0, "totally_unique_xyzzy_qux", "abc\nxyz\n",
+        )
+        assert result == ""
--- a/tests/tools/test_patch_did_you_mean.py
+++ b/tests/tools/test_patch_did_you_mean.py
@@ -0,0 +1,114 @@
+import json
+import os
+import textwrap
+from pathlib import Path
+
+import tools.skill_manager_tool as skill_manager_tool
+from tools.file_tools import patch_tool
+from tools.skill_manager_tool import _create_skill, _patch_skill
+
+
+def _disable_patch_tool_guards(monkeypatch):
+    monkeypatch.setattr("tools.file_tools._check_sensitive_path", lambda _path: None)
+    monkeypatch.setattr("tools.file_tools._check_file_staleness", lambda _path, _task_id: None)
+    monkeypatch.setattr("tools.file_tools._log_and_check_conflict", lambda _path, _task_id, _action: None)
+
+
+def test_patch_tool_replace_no_match_shows_rich_hint_without_legacy_hint(tmp_path, monkeypatch):
+    _disable_patch_tool_guards(monkeypatch)
+    sample = tmp_path / "sample.py"
+    sample.write_text("def foo():\n    return 1\n\ndef bar():\n    return 2\n", encoding="utf-8")
+
+    raw = patch_tool(
+        mode="replace",
+        path=str(sample),
+        old_string="def barycentric():",
+        new_string="def barycentric_new():",
+        task_id="qa960-replace-rich-hint",
+    )
+
+    result = json.loads(raw)
+    assert result["success"] is False
+    assert "Could not find a match" in result["error"]
+    assert "Did you mean one of these sections?" in result["error"]
+    assert "def bar():" in result["error"] or "def foo():" in result["error"]
+    assert "[Hint:" not in raw
+
+
+def test_patch_tool_replace_ambiguous_error_does_not_show_did_you_mean(tmp_path, monkeypatch):
+    _disable_patch_tool_guards(monkeypatch)
+    sample = tmp_path / "sample.py"
+    sample.write_text("aaa\nbbb\naaa\n", encoding="utf-8")
+
+    raw = patch_tool(
+        mode="replace",
+        path=str(sample),
+        old_string="aaa",
+        new_string="ccc",
+        task_id="qa960-replace-ambiguous",
+    )
+
+    result = json.loads(raw)
+    assert result["success"] is False
+    assert "Found 2 matches" in result["error"]
+    assert "Did you mean one of these sections?" not in result["error"]
+    assert "[Hint:" not in raw
+
+
+def test_patch_tool_v4a_no_match_shows_rich_hint(tmp_path, monkeypatch):
+    _disable_patch_tool_guards(monkeypatch)
+    sample = tmp_path / "sample.py"
+    sample.write_text("def foo():\n    return 1\n", encoding="utf-8")
+
+    patch = textwrap.dedent(
+        f"""\
+        *** Begin Patch
+        *** Update File: {sample}
+        @@
+        -def barycentric():
+        +def barycentric_new():
+        *** End Patch
+        """
+    )
+
+    raw = patch_tool(mode="patch", patch=patch, task_id="qa960-v4a-rich-hint")
+    result = json.loads(raw)
+    assert result["success"] is False
+    assert "Patch validation failed" in result["error"]
+    assert "Did you mean one of these sections?" in result["error"]
+    assert "def foo():" in result["error"]
+
+
+def test_skill_patch_no_match_shows_rich_hint(tmp_path, monkeypatch):
+    monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+    skills_dir = tmp_path / "skills"
+    skills_dir.mkdir(parents=True, exist_ok=True)
+    monkeypatch.setattr(skill_manager_tool, "SKILLS_DIR", skills_dir)
+    monkeypatch.setattr(skill_manager_tool, "_security_scan_skill", lambda _skill_dir: None)
+
+    _create_skill(
+        "qa-skill",
+        textwrap.dedent(
+            """\
+            ---
+            name: qa-skill
+            description: test
+            ---
+
+            Step 1: Do the thing.
+            Step 2: Verify the thing.
+            """
+        ),
+    )
+
+    result = _patch_skill(
+        "qa-skill",
+        "Step 1: Do the production rollout.",
+        "Step 1: Updated.",
+    )
+
+    assert result["success"] is False
+    assert "Could not find a match" in result["error"]
+    assert "Did you mean one of these sections?" in result["error"]
+    assert "Step 1: Do the thing." in result["error"]
+    assert "file_preview" in result
--- a/tools/file_operations.py
+++ b/tools/file_operations.py
@@ -757,12 +757,14 @@ class ShellFileOperations(FileOperations):
            content, old_string, new_string, replace_all
        )
        
-        if error:
-            return PatchResult(error=error)
-        
-        if match_count == 0:
-            return PatchResult(error=f"Could not find match for old_string in {path}")
-        
+        if error or match_count == 0:
+            err_msg = error or f"Could not find match for old_string in {path}"
+            try:
+                from tools.fuzzy_match import format_no_match_hint
+                err_msg += format_no_match_hint(err_msg, match_count, old_string, content)
+            except Exception:
+                pass
+            return PatchResult(error=err_msg)
        # Write back
        write_result = self.write_file(path, new_content)
        if write_result.error:
--- a/tools/file_tools.py
+++ b/tools/file_tools.py
@@ -8,6 +8,7 @@ import os
 import threading
 import time
 from pathlib import Path
+from typing import Any, Dict, Optional
 from tools.binary_extensions import has_binary_extension
 from tools.file_operations import ShellFileOperations
 from agent.redact import redact_sensitive_text
@@ -690,8 +691,11 @@ def patch_tool(mode: str = "replace", path: str = None, old_string: str = None,
        result_json = json.dumps(result_dict, ensure_ascii=False)
        # Hint when old_string not found — saves iterations where the agent
        # retries with stale content instead of re-reading the file.
+        # Suppressed when patch_replace already attached a rich "Did you mean?"
+        # snippet (which is strictly more useful than the generic hint).
        if result_dict.get("error") and "Could not find" in str(result_dict["error"]):
-            result_json += "\n\n[Hint: old_string not found. Use read_file to verify the current content, or search_files to locate the text.]"
+            if "Did you mean one of these sections?" not in str(result_dict["error"]):
+                result_json += "\n\n[Hint: old_string not found. Use read_file to verify the current content, or search_files to locate the text.]"
        return result_json
    except Exception as e:
        return tool_error(str(e))
--- a/tools/fuzzy_match.py
+++ b/tools/fuzzy_match.py
@@ -93,6 +93,21 @@ def fuzzy_find_and_replace(content: str, old_string: str, new_string: str,
                    f"Provide more context to make it unique, or use replace_all=True."
                )

+            # Escape-drift guard: when the matched strategy is NOT `exact`,
+            # we matched via some form of normalization. If new_string
+            # contains shell/JSON-style escape sequences (\\' or \\") that
+            # would be written literally into the file but the matched
+            # region of the file has no such sequences, this is almost
+            # certainly tool-call serialization drift — the model typed
+            # an apostrophe/quote and the transport added a stray
+            # backslash. Writing new_string as-is would corrupt the file.
+            # Block with a helpful error so the model re-reads and retries
+            # instead of the caller silently persisting garbage (or not).
+            if strategy_name != "exact":
+                drift_err = _detect_escape_drift(content, matches, old_string, new_string)
+                if drift_err:
+                    return content, 0, None, drift_err
+
            # Perform replacement
            new_content = _apply_replacements(content, matches, new_string)
            return new_content, len(matches), strategy_name, None
@@ -101,6 +116,46 @@ def fuzzy_find_and_replace(content: str, old_string: str, new_string: str,
    return content, 0, None, "Could not find a match for old_string in the file"


+def _detect_escape_drift(content: str, matches: List[Tuple[int, int]],
+                         old_string: str, new_string: str) -> Optional[str]:
+    """Detect tool-call escape-drift artifacts in new_string.
+
+    Looks for ``\\'`` or ``\\"`` sequences that are present in both
+    old_string and new_string (i.e. the model copy-pasted them as "context"
+    it intended to preserve) but don't exist in the matched region of the
+    file. That pattern indicates the transport layer inserted spurious
+    shell-style escapes around apostrophes or quotes — writing new_string
+    verbatim would literally insert ``\\'`` into source code.
+
+    Returns an error string if drift is detected, None otherwise.
+    """
+    # Cheap pre-check: bail out unless new_string actually contains a
+    # suspect escape sequence. This keeps the guard free for all the
+    # common, correct cases.
+    if "\\'" not in new_string and '\\"' not in new_string:
+        return None
+
+    # Aggregate matched regions of the file — that's what new_string will
+    # replace. If the suspect escapes are present there already, the
+    # model is genuinely preserving them (valid for some languages /
+    # escaped strings); accept the patch.
+    matched_regions = "".join(content[start:end] for start, end in matches)
+
+    for suspect in ("\\'", '\\"'):
+        if suspect in new_string and suspect in old_string and suspect not in matched_regions:
+            plain = suspect[1]  # "'" or '"'
+            return (
+                f"Escape-drift detected: old_string and new_string contain "
+                f"the literal sequence {suspect!r} but the matched region of "
+                f"the file does not. This is almost always a tool-call "
+                f"serialization artifact where an apostrophe or quote got "
+                f"prefixed with a spurious backslash. Re-read the file with "
+                f"read_file and pass old_string/new_string without "
+                f"backslash-escaping {plain!r} characters."
+            )
+    return None
+
+
 def _apply_replacements(content: str, matches: List[Tuple[int, int]], new_string: str) -> str:
    """
    Apply replacements at the given positions.
@@ -564,3 +619,86 @@ def _map_normalized_positions(original: str, normalized: str,
        original_matches.append((orig_start, min(orig_end, len(original))))
    
    return original_matches
+
+
+def find_closest_lines(old_string: str, content: str, context_lines: int = 2, max_results: int = 3) -> str:
+    """Find lines in content most similar to old_string for "did you mean?" feedback.
+
+    Returns a formatted string showing the closest matching lines with context,
+    or empty string if no useful match is found.
+    """
+    if not old_string or not content:
+        return ""
+
+    old_lines = old_string.splitlines()
+    content_lines = content.splitlines()
+
+    if not old_lines or not content_lines:
+        return ""
+
+    # Use first line of old_string as anchor for search
+    anchor = old_lines[0].strip()
+    if not anchor:
+        # Try second line if first is blank
+        candidates = [l.strip() for l in old_lines if l.strip()]
+        if not candidates:
+            return ""
+        anchor = candidates[0]
+
+    # Score each line in content by similarity to anchor
+    scored = []
+    for i, line in enumerate(content_lines):
+        stripped = line.strip()
+        if not stripped:
+            continue
+        ratio = SequenceMatcher(None, anchor, stripped).ratio()
+        if ratio > 0.3:
+            scored.append((ratio, i))
+
+    if not scored:
+        return ""
+
+    # Take top matches
+    scored.sort(key=lambda x: -x[0])
+    top = scored[:max_results]
+
+    parts = []
+    seen_ranges = set()
+    for _, line_idx in top:
+        start = max(0, line_idx - context_lines)
+        end = min(len(content_lines), line_idx + len(old_lines) + context_lines)
+        key = (start, end)
+        if key in seen_ranges:
+            continue
+        seen_ranges.add(key)
+        snippet = "\n".join(
+            f"{start + j + 1:4d}| {content_lines[start + j]}"
+            for j in range(end - start)
+        )
+        parts.append(snippet)
+
+    if not parts:
+        return ""
+
+    return "\n---\n".join(parts)
+
+
+def format_no_match_hint(error: Optional[str], match_count: int,
+                         old_string: str, content: str) -> str:
+    """Return a '\\n\\nDid you mean...' snippet for plain no-match errors.
+
+    Gated so the hint only fires for actual "old_string not found" failures.
+    Ambiguous-match ("Found N matches"), escape-drift, and identical-strings
+    errors all have ``match_count == 0`` but a "did you mean?" snippet would
+    be misleading — those failed for unrelated reasons.
+
+    Returns an empty string when there's nothing useful to append.
+    """
+    if match_count != 0:
+        return ""
+    if not error or not error.startswith("Could not find"):
+        return ""
+    hint = find_closest_lines(old_string, content)
+    if not hint:
+        return ""
+    return "\n\nDid you mean one of these sections?\n" + hint
--- a/tools/patch_parser.py
+++ b/tools/patch_parser.py
@@ -290,10 +290,16 @@ def _validate_operations(
                )
                if count == 0:
                    label = f"'{hunk.context_hint}'" if hunk.context_hint else "(no hint)"
-                    errors.append(
+                    msg = (
                        f"{op.file_path}: hunk {label} not found"
                        + (f" — {match_error}" if match_error else "")
                    )
+                    try:
+                        from tools.fuzzy_match import format_no_match_hint
+                        msg += format_no_match_hint(match_error, count, search_pattern, simulated)
+                    except Exception:
+                        pass
+                    errors.append(msg)
                else:
                    # Advance simulation so subsequent hunks validate correctly.
                    # Reuse the result from the call above — no second fuzzy run.
@@ -537,7 +543,13 @@ def _apply_update(op: PatchOperation, file_ops: Any) -> Tuple[bool, str]:
                            error = None
                
                if error:
-                    return False, f"Could not apply hunk: {error}"
+                    err_msg = f"Could not apply hunk: {error}"
+                    try:
+                        from tools.fuzzy_match import format_no_match_hint
+                        err_msg += format_no_match_hint(error, 0, search_pattern, new_content)
+                    except Exception:
+                        pass
+                    return False, err_msg
        else:
            # Addition-only hunk (no context or removed lines).
            # Insert at the location indicated by the context hint, or at end of file.
--- a/tools/skill_manager_tool.py
+++ b/tools/skill_manager_tool.py
@@ -575,9 +575,15 @@ def _patch_skill(
    if match_error:
        # Show a short preview of the file so the model can self-correct
        preview = content[:500] + ("..." if len(content) > 500 else "")
+        err_msg = match_error
+        try:
+            from tools.fuzzy_match import format_no_match_hint
+            err_msg += format_no_match_hint(match_error, match_count, old_string, content)
+        except Exception:
+            pass
        return {
            "success": False,
-            "error": match_error,
+            "error": err_msg,
            "file_preview": preview,
        }
Author	SHA1	Message	Date
Alexander Whitestone	e28d16b324	test: cover patch did-you-mean end-to-end (#960 ) All checks were successful Lint / lint (pull_request) Successful in 18s Details - add focused QA tests for replace-mode rich hints without legacy generic hints - verify ambiguous replace failures do not show did-you-mean noise - verify V4A patch validation surfaces rich hints - verify skill patching surfaces rich hints on true no-match failures	2026-04-22 10:43:34 -04:00
Alexander Whitestone	bc32047610	wip: fix patch did-you-mean dependencies (#960 ) - restore escape-drift guard needed by current fuzzy-match tests - import missing typing symbols in file_tools after porting patch hint logic	2026-04-22 10:41:29 -04:00
Teknium	3a24420d7d	fix(patch): gate 'did you mean?' to no-match + extend to v4a/skill_manage Follow-ups on top of @teyrebaz33's cherry-picked commit: 1. New shared helper format_no_match_hint() in fuzzy_match.py with a startswith('Could not find') gate so the snippet only appends to genuine no-match errors — not to 'Found N matches' (ambiguous), 'Escape-drift detected', or 'identical strings' errors, which would all mislead the model. 2. file_tools.patch_tool suppresses the legacy generic '[Hint: old_string not found...]' string when the rich 'Did you mean?' snippet is already attached — no more double-hint. 3. Wire the same helper into patch_parser.py (V4A patch mode, both _validate_operations and _apply_update) and skill_manager_tool.py so all three fuzzy callers surface the hint consistently. Tests: 7 new gating tests in TestFormatNoMatchHint cover every error class (ambiguous, drift, identical, non-zero match count, None error, no similar content, happy path). 34/34 test_fuzzy_match, 96/96 test_file_tools + test_patch_parser + test_skill_manager_tool pass. E2E verified across all four scenarios: no-match-with-similar, no-match-no-similar, ambiguous, success. V4A mode confirmed end-to-end with a non-matching hunk. (cherry picked from commit `5e6427a42c`)	2026-04-22 10:30:00 -04:00
teyrebaz33	d14c1c5a56	feat(patch): add 'did you mean?' feedback when patch fails to match When patch_replace() cannot find old_string in a file, the error message now includes the closest matching lines from the file with line numbers and context. This helps the LLM self-correct without a separate read_file call. Implements Phase 1 of #536: enhanced patch error feedback with no architectural changes. - tools/fuzzy_match.py: new find_closest_lines() using SequenceMatcher - tools/file_operations.py: attach closest-lines hint to patch errors - tests/tools/test_fuzzy_match.py: 5 new tests for find_closest_lines (cherry picked from commit `15abf4ed8f`)	2026-04-22 10:28:40 -04:00