diff --git a/index.html b/index.html index f0cfa61..e4bd6f1 100644 --- a/index.html +++ b/index.html @@ -284,8 +284,8 @@

Your schema passed.
  • Flow - Adapters - LangGraph + OpenAI Agents + Trace inputs + LangGraph · OpenAI Agents · local OTLP
  • @@ -305,7 +305,7 @@

    Your schema passed.

    • +
  • Explain Reports @@ -321,25 +321,27 @@

    Your schema passed.
    -

    A report that explains causality

    Not a score.
    A repairable witness.

    Machine-readable JSON and a self-contained HTML report come from one redacted report model. The report works offline.

    +

    A report that explains causality

    Not a score.
    A repairable witness.

    Machine-readable JSON and a self-contained HTML report come from one redacted report model. Coverage is explicit and never presented as correctness.

    -
    report.jsonmachine-readable
    +
    report.jsonmachine-readable excerpt
    {
-  "structural": "pass",
-  "semantic": "breaking",
-  "first_breaking_edge": {
-    "producer": "researcher",
-    "consumer": "verifier",
-    "contract": "verified_requires_opened_supporting_source"
-  },
-  "witness": {
-    "run": "candidate-003",
-    "expected": "opened_sources_count > 0",
-    "observed": 0
-  },
-  "affected": ["verifier", "decision_maker", "publisher"],
-  "repair_before": "verifier"
+  "schema_version": "0.3",
+  "structural": { "status": "PASS" },
+  "semantic": {
+    "status": "FAIL",
+    "first_breaking_edge": "researcher_to_verifier",
+    "coverage": { "summary": {
+      "total_graph_edges": 3,
+      "contracted_and_observed": 3,
+      "observed_contract_coverage_percent": 100.0,
+      "coverage_is_correctness": false
+    } },
+    "findings": [{
+      "contract_id": "research_demo:researcher_to_verifier:verified_requires_opened_supporting_source",
+      "affected_downstream_nodes": ["verifier", "decision_maker", "publisher"]
+    }]
+  }
 }
    @@ -376,14 +378,14 @@

    Your schema passed.

    Measured technical proof

    Small surface.
    Serious checks.

    Current branch measurements, linked to the source that enforces them.

    -
    Tests
    172passing locally
    -
    Coverage
    91.35%90% CI floor
    +
    Tests
    194passing locally
    +
    Coverage
    91.37%90% CI floor
    Python CI
    3.12 · 3.13both required
    Evaluators
    7 typesdeterministic registry
    Runtime
    Local-firstno required API key
    Adapters
    2 maintainedLangGraph + OpenAI Agents
    -

    Benchmarks cover synthetic graphs at 10, 100, and 1,000 nodes. They do not establish production scale. Inspect the methodology.

    +

    Recorded model, prompt, and tool migrations plus a ten-case semantic regression corpus exercise the same comparison engine. The corpus is project test data, not an industry benchmark. Inspect every case. Run the reusable GitHub Action.

    @@ -406,7 +408,7 @@

    Your schema passed. EvaluatorAdd one deterministic contract family.Start with the registry and adversarial tests. AdapterPropose another deterministic trace boundary.Document exact mappings, unsupported cases, and version bounds. - ExampleAdd a focused semantic regression case.Keep the default path deterministic, local, and keyless. + CorpusAdd a focused semantic regression case.Keep the baseline, candidate, contract, and expected finding inspectable.

    Browse all issuesRead contributingOpen a design discussion

    @@ -429,7 +431,7 @@

    Catch the first broken edge.

    diff --git a/proof.json b/proof.json index 79f580a..6d343ae 100644 --- a/proof.json +++ b/proof.json @@ -1,6 +1,6 @@ { - "tests": 172, - "coverage_percent": 91.35, + "tests": 194, + "coverage_percent": 91.37, "python_versions": [ "3.12", "3.13"