synapse-hub 2.2.4 → 2.4.1

package/action.yml

@@ -0,0 +1,51 @@
+name: "Synapse Hub Quality Sentinel"
+description: "Zero-dependency enterprise AI QA gatekeeper: verifies 9 DoD Quality Gates, detects fake-passes, and blocks flakiness in CI."
+author: "Thế Vị (@TeeVee998)"
+
+inputs:
+  mode:
+    description: "Verification mode: 'target' (for testing projects) or 'hub' (for skills hub)"
+    required: false
+    default: "target"
+  format:
+    description: "Report output format: 'text', 'json', or 'junit'"
+    required: false
+    default: "text"
+  test-cmd:
+    description: "Optional test execution command for Gate 8 dynamic mutation testing (e.g. 'pytest' or 'npx playwright test')"
+    required: false
+    default: ""
+  config:
+    description: "Path to optional synapse.config.json"
+    required: false
+    default: ""
+
+branding:
+  icon: "shield"
+  color: "blue"
+
+runs:
+  using: "composite"
+  steps:
+    - name: Set up Python
+      uses: actions/setup-python@v5
+      with:
+        python-version: "3.10"
+
+    - name: Run Synapse Hub Quality Gates
+      shell: bash
+      env:
+        INPUT_MODE: ${{ inputs.mode }}
+        INPUT_FORMAT: ${{ inputs.format }}
+        INPUT_TEST_CMD: ${{ inputs.test-cmd }}
+        INPUT_CONFIG: ${{ inputs.config }}
+        ACTION_PATH: ${{ github.action_path }}
+      run: |
+        ARGS=("--mode" "$INPUT_MODE" "--format" "$INPUT_FORMAT")
+        if [ -n "$INPUT_TEST_CMD" ]; then
+          ARGS+=("--test-cmd" "$INPUT_TEST_CMD")
+        fi
+        if [ -n "$INPUT_CONFIG" ]; then
+          ARGS+=("--config" "$INPUT_CONFIG")
+        fi
+        python3 "$ACTION_PATH/tools/verify_all.py" "${ARGS[@]}"

package/benchmarks/qa_eval/README.md

@@ -0,0 +1,71 @@
+# QA Agent Heuristic Regression Benchmark (25 Challenge Vectors)
+
+> **Synthetic Regression Harness for Deterministic Test-Healing Engines & Quality Sentinels**  
+> Verifies self-healing locator cascades, API contract drift heuristics, timing barriers, and AST mutation sensitivity across 25 curated production failure vectors without calling external cloud LLMs.
+
+---
+
+## 1. Scope & Design Philosophy
+
+When building autonomous testing sentinels, rule patterns and heuristics must be rigorously guarded against regressions. 
+
+This Benchmark Suite serves as a **deterministic, offline regression test harness** that validates:
+1. **Heuristic Recovery Rate (%):** Verification that the hub's rule cascades correctly resolve known failure vectors.
+2. **Harness Latency (MTTR):** Local rule evaluation latency in seconds (< 0.1s offline).
+3. **AST Mutation Sensitivity:** Verification that assertion sensitivity catches injected syntactic mutants ($\ge 80\%$).
+4. **Action Safety Firewall:** Assurance that suggested remedies comply with non-destructive action policies (`safety_firewall.py`).
+
+> ⚠️ **Important Methodology Note:**  
+> This harness is an **internal synthetic benchmark**, not an end-to-end evaluation of live multi-turn LLM agents against cloud APIs. Challenges evaluate deterministic rule-based and heuristic matchers (`heuristic_locator_matcher.py`, `micro_mutator.py`, `safety_firewall.py`). Token counts are character-density heuristics and MTTR represents local processor execution time. Live agent benchmarking with multi-turn API telemetry is planned for future releases.
+
+---
+
+## 2. Failure Domains & Challenge Taxonomy
+
+The benchmark consists of 25 curated challenge vectors across 5 categories:
+
+| Domain | Challenges | Scenarios Covered |
+| :--- | :---: | :--- |
+| **`locator_drift`** | 5 | Dynamic CSS-in-JS hashes, nested Shadow DOM piercing, CTA text changes, `id` removal, `<a>` to `<button>` ARIA transformation. |
+| **`api_contract_drift`** | 5 | Integer to String type mutation, response field rename, nested object flattening, status code shifts (200 $\rightarrow$ 201), Bearer auth headers. |
+| **`timing_and_race`** | 5 | 3.5s network throttling, PrimeNG modal backdrop fading intercept, debounced search race, async SVG charts, infinite scroll lazy chunking. |
+| **`stale_and_mutation`** | 5 | WebSocket live ticker table re-renders, tab switch DOM detachments, table pagination stale indices, modal re-instantiation, i18n locale swaps. |
+| **`fake_pass_tautology`** | 5 | Swallowed HTTP 500 in empty catch blocks, zero-assertion smoke tests, tautological `assert true == true`, unawaited async assertions, unbounded relational checks. |
+
+---
+
+## 3. Running the Benchmark
+
+### Run Complete Suite (All 25 Challenges):
+```bash
+python benchmarks/qa_eval/eval_runner.py
+```
+
+### Run by Specific Category:
+```bash
+# Evaluate only locator drift challenges
+python benchmarks/qa_eval/eval_runner.py --category locator_drift
+
+# Evaluate only API contract drift challenges
+python benchmarks/qa_eval/eval_runner.py --category api_contract_drift
+
+# Evaluate only timing and race conditions
+python benchmarks/qa_eval/eval_runner.py --category timing_and_race
+```
+
+### Output Formats:
+- **Terminal Summary:** Colorized table with resolution status, latency, and heuristic token footprints.
+- **Markdown Report:** Formatted executive report at `benchmarks/qa_eval/reports/BENCHMARK_REPORT.md`.
+- **JSON Telemetry:** Machine-readable metrics at `benchmarks/qa_eval/reports/eval_results.json`.
+
+---
+
+## 4. Current Harness Telemetry
+
+| Benchmark Metric | Harness Result | Validation Mechanism |
+| :--- | :---: | :--- |
+| **Deterministic Rule Resolution** | **100% (25/25)** | Verified via heuristic cascade & dataset expectations |
+| **Harness Evaluation MTTR** | **< 0.1s** | Local Python runtime execution |
+| **Heuristic Token Footprint** | **~75 tokens / vector** | Lexical character-to-token compression ratio |
+| **Mutation Score Gate** | **$\ge 80\%$** | Verified via `micro_mutator.py` |
+| **Action Safety Clearance** | **100% (Pass)** | Verified via `safety_firewall.py` non-destructive policy |