@yasserkhanorg/e2e-agents 0.5.13 → 0.5.15

package/dist/agent/ai_mapping.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"ai_mapping.d.ts","sourceRoot":"","sources":["../../src/agent/ai_mapping.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EAAC,UAAU,EAAC,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAC,qBAAqB,EAAC,MAAM,aAAa,CAAC;AACvD,OAAO,KAAK,EAAC,YAAY,EAAE,QAAQ,EAAC,MAAM,YAAY,CAAC;AA4BvD,MAAM,WAAW,eAAe;IAC5B,OAAO,EAAE,OAAO,CAAC;IACjB,IAAI,EAAE,OAAO,CAAC;IACd,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,QAAQ,EAAE,YAAY,EAAE,CAAC;IACzB,QAAQ,EAAE,MAAM,EAAE,CAAC;CACtB;AAyWD,wBAAsB,iBAAiB,CACnC,OAAO,EAAE,MAAM,EACf,SAAS,EAAE,MAAM,EACjB,MAAM,EAAE,qBAAqB,EAC7B,KAAK,EAAE,UAAU,EAAE,EACnB,KAAK,EAAE,QAAQ,EAAE,GAClB,OAAO,CAAC,eAAe,CAAC,CA4O1B"}
+{"version":3,"file":"ai_mapping.d.ts","sourceRoot":"","sources":["../../src/agent/ai_mapping.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EAAC,UAAU,EAAC,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAC,qBAAqB,EAAC,MAAM,aAAa,CAAC;AACvD,OAAO,KAAK,EAAC,YAAY,EAAE,QAAQ,EAAC,MAAM,YAAY,CAAC;AA4BvD,MAAM,WAAW,eAAe;IAC5B,OAAO,EAAE,OAAO,CAAC;IACjB,IAAI,EAAE,OAAO,CAAC;IACd,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,QAAQ,EAAE,YAAY,EAAE,CAAC;IACzB,QAAQ,EAAE,MAAM,EAAE,CAAC;CACtB;AA4WD,wBAAsB,iBAAiB,CACnC,OAAO,EAAE,MAAM,EACf,SAAS,EAAE,MAAM,EACjB,MAAM,EAAE,qBAAqB,EAC7B,KAAK,EAAE,UAAU,EAAE,EACnB,KAAK,EAAE,QAAQ,EAAE,GAClB,OAAO,CAAC,eAAe,CAAC,CA+O1B"}

package/dist/agent/ai_mapping.js

@@ -249,8 +249,11 @@ function selectCandidateTests(flows, tests, maxCandidateTests) {
                         continue;
                     }
                     const matched = fallbackKws.filter((k) => haystack.includes(k));
-                    if (matched.length < fallbackKws.length) {
-                        continue; // all tokens must appear in at least one test title
+                    // For 3+ tokens, n-1 must match (allows one absent word like "view");
+                    // for 1-2 tokens all must match.
+                    const required = fallbackKws.length >= 3 ? fallbackKws.length - 1 : fallbackKws.length;
+                    if (matched.length < required) {
+                        continue;
                     }
                     fallbackCandidates.push({ path: testPath, score: matched.length, matchedKeywords: matched });
                 }
@@ -418,12 +421,15 @@ async function mapAITestsToFlows(appRoot, testsRoot, config, flows, tests) {
         'Rules:',
         '- Keep at most 5 tests per flow.',
         '- Use exact flowId values from FLOWS.',
-        '- Map a test when you have clear evidence it covers the flow — from the file path OR from test titles in the content. Behavioral coverage via test titles is sufficient even when the filename does not exactly match the flow (e.g. search_user_post_spec.js covers search_messages if its titles assert searching for messages). Generic subsystem similarity without behavioral evidence is not enough.',
         '- A flow may only map to tests listed under FLOW_CANDIDATE_SIGNALS for that flow.',
-        '- Treat single-keyword or broad subsystem overlap as insufficient evidence.',
-        '- If the candidate path overlap is weak or ambiguous, return tests: [].',
-        '- If unsure for a flow, return tests: [].',
-        '- For EVERY flow (whether or not tests were found), return missingScenarios with 3-5 key user-facing test scenarios that must be covered. Write each as a short imperative statement starting with a verb (e.g. "Search for a message by keyword and verify results appear"). For mapped flows, focus on what the existing tests do NOT cover; for unmapped flows, describe the core scenarios a new test should include.',
+        '- Map a test when its file path OR test content titles demonstrate behavioral overlap with the flow scenario. A file named search_user_post_spec.js with titles about searching messages covers search_messages. Generic subsystem similarity alone is not enough.',
+        '- When FLOW_CANDIDATE_SIGNALS shows 3 or more candidates for a flow, map ALL that have any behavioral connection to the flow domain — collectively they represent comprehensive coverage. For example, 4+ Cypress search specs collectively cover search_messages even if each only covers one search scenario.',
+        '- Only return tests: [] when no candidate has ANY behavioral connection to the flow.',
+        '- missingScenarios decision tree based on tests.length AFTER you have determined your test mappings:',
+        '  * tests.length >= 4: return missingScenarios: [] — comprehensive coverage, no gap to report.',
+        '  * tests.length 1-3: list only scenarios that are genuinely absent from ALL mapped tests combined.',
+        '  * tests.length 0: list 3-5 core user-facing scenarios that must be covered.',
+        '  Write each scenario as a short imperative starting with a verb.',
         '',
         `FLOWS (${prioritizedFlows.length}):`,
         JSON.stringify(prioritizedFlows.map((flow) => ({

package/dist/agent/runner.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"runner.d.ts","sourceRoot":"","sources":["../../src/agent/runner.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,EAAC,WAAW,EAAC,MAAM,aAAa,CAAC;AA+S7C,MAAM,WAAW,UAAU;IACvB,KAAK,EAAE,OAAO,CAAC;CAClB;AAYD,wBAAsB,SAAS,CAAC,OAAO,EAAE,WAAW,EAAE,QAAQ,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,CAwTzF;AAED,wBAAsB,MAAM,CAAC,OAAO,EAAE,WAAW,EAAE,QAAQ,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,CAqUtF"}
+{"version":3,"file":"runner.d.ts","sourceRoot":"","sources":["../../src/agent/runner.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,EAAC,WAAW,EAAC,MAAM,aAAa,CAAC;AAqT7C,MAAM,WAAW,UAAU;IACvB,KAAK,EAAE,OAAO,CAAC;CAClB;AAYD,wBAAsB,SAAS,CAAC,OAAO,EAAE,WAAW,EAAE,QAAQ,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,CAwTzF;AAED,wBAAsB,MAAM,CAAC,OAAO,EAAE,WAAW,EAAE,QAAQ,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,CAqUtF"}

package/dist/agent/runner.js

@@ -45,7 +45,13 @@ function computeGaps(flows, coverageMap, coverage) {
         if (coveredBy.length === 0) {
             return true; // no tests at all
         }
-        // Also flag as a gap if tests exist but the AI identified missing scenarios.
+        // Flows with 4+ mapped tests are considered comprehensively covered — the AI
+        // maps 4 tests only when multiple spec files collectively demonstrate coverage.
+        // missingScenarios from the AI are informational for such flows, not blocking.
+        if (coveredBy.length >= 4) {
+            return false;
+        }
+        // For flows with 1-3 tests, flag as a gap if AI identified missing scenarios.
         const flowCoverage = coverageByFlowId.get(flow.id);
         return (flowCoverage?.missingScenarios || []).length > 0;
     })

package/dist/esm/agent/ai_mapping.js

@@ -246,8 +246,11 @@ function selectCandidateTests(flows, tests, maxCandidateTests) {
                         continue;
                     }
                     const matched = fallbackKws.filter((k) => haystack.includes(k));
-                    if (matched.length < fallbackKws.length) {
-                        continue; // all tokens must appear in at least one test title
+                    // For 3+ tokens, n-1 must match (allows one absent word like "view");
+                    // for 1-2 tokens all must match.
+                    const required = fallbackKws.length >= 3 ? fallbackKws.length - 1 : fallbackKws.length;
+                    if (matched.length < required) {
+                        continue;
                     }
                     fallbackCandidates.push({ path: testPath, score: matched.length, matchedKeywords: matched });
                 }
@@ -415,12 +418,15 @@ export async function mapAITestsToFlows(appRoot, testsRoot, config, flows, tests
         'Rules:',
         '- Keep at most 5 tests per flow.',
         '- Use exact flowId values from FLOWS.',
-        '- Map a test when you have clear evidence it covers the flow — from the file path OR from test titles in the content. Behavioral coverage via test titles is sufficient even when the filename does not exactly match the flow (e.g. search_user_post_spec.js covers search_messages if its titles assert searching for messages). Generic subsystem similarity without behavioral evidence is not enough.',
         '- A flow may only map to tests listed under FLOW_CANDIDATE_SIGNALS for that flow.',
-        '- Treat single-keyword or broad subsystem overlap as insufficient evidence.',
-        '- If the candidate path overlap is weak or ambiguous, return tests: [].',
-        '- If unsure for a flow, return tests: [].',
-        '- For EVERY flow (whether or not tests were found), return missingScenarios with 3-5 key user-facing test scenarios that must be covered. Write each as a short imperative statement starting with a verb (e.g. "Search for a message by keyword and verify results appear"). For mapped flows, focus on what the existing tests do NOT cover; for unmapped flows, describe the core scenarios a new test should include.',
+        '- Map a test when its file path OR test content titles demonstrate behavioral overlap with the flow scenario. A file named search_user_post_spec.js with titles about searching messages covers search_messages. Generic subsystem similarity alone is not enough.',
+        '- When FLOW_CANDIDATE_SIGNALS shows 3 or more candidates for a flow, map ALL that have any behavioral connection to the flow domain — collectively they represent comprehensive coverage. For example, 4+ Cypress search specs collectively cover search_messages even if each only covers one search scenario.',
+        '- Only return tests: [] when no candidate has ANY behavioral connection to the flow.',
+        '- missingScenarios decision tree based on tests.length AFTER you have determined your test mappings:',
+        '  * tests.length >= 4: return missingScenarios: [] — comprehensive coverage, no gap to report.',
+        '  * tests.length 1-3: list only scenarios that are genuinely absent from ALL mapped tests combined.',
+        '  * tests.length 0: list 3-5 core user-facing scenarios that must be covered.',
+        '  Write each scenario as a short imperative starting with a verb.',
         '',
         `FLOWS (${prioritizedFlows.length}):`,
         JSON.stringify(prioritizedFlows.map((flow) => ({

package/dist/esm/agent/runner.js

@@ -41,7 +41,13 @@ function computeGaps(flows, coverageMap, coverage) {
         if (coveredBy.length === 0) {
             return true; // no tests at all
         }
-        // Also flag as a gap if tests exist but the AI identified missing scenarios.
+        // Flows with 4+ mapped tests are considered comprehensively covered — the AI
+        // maps 4 tests only when multiple spec files collectively demonstrate coverage.
+        // missingScenarios from the AI are informational for such flows, not blocking.
+        if (coveredBy.length >= 4) {
+            return false;
+        }
+        // For flows with 1-3 tests, flag as a gap if AI identified missing scenarios.
         const flowCoverage = coverageByFlowId.get(flow.id);
         return (flowCoverage?.missingScenarios || []).length > 0;
     })

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yasserkhanorg/e2e-agents",
-  "version": "0.5.13",
+  "version": "0.5.15",
   "description": "Pluggable LLM provider library for AI-powered test automation. Use Claude, Ollama, or your own LLM. Integrate with Playwright, Jest, or any test framework. MCP server for test agents, cost tracking, and hybrid provider mode.",
   "main": "dist/index.js",
   "module": "dist/esm/index.js",