fix(confidence-engine): resolve model for episode reasoning

tests/graph/episode-reasoning-seam.test.js

@@ -155,4 +155,42 @@ describe("episode reasoning seam (reconsiderCompletedEpisode)", () => {
     expect(TEST_EPISODE.situationGraph.nodes.length).toBe(1);
   });
 
+  it("resolves configured model via assertConfig fallback when no deps model is injected", async () => {
+    // Set minimal env vars for assertConfig to succeed in this specific test
+    const prevBaseUrl = process.env.OLLAMA_BASE_URL;
+    const prevModel = process.env.OLLAMA_MODEL;
+    process.env.OLLAMA_BASE_URL = "http://localhost:11434";
+    process.env.OLLAMA_MODEL = "test-model";
+
+    const modelCaptured = [];
+    const provider = {
+      generateReconstruction: vi.fn().mockImplementation((prompt, model) => {
+        modelCaptured.push(model);
+        return makeProposalResponse();
+      }),
+    };
+
+    // Reset modules so assertConfig reads fresh env vars
+    await vi.resetModules();
+
+    const { reconsiderCompletedEpisode } = await import("@/lib/graph/orchestrator.js");
+    // Empty deps — matches route.js call: reconsiderCompletedEpisode(prepared)
+    // The fix ensures assertConfig() provides a model instead of leaving it null
+    const result = await reconsiderCompletedEpisode(
+      { ...TEST_EPISODE, provider },
+      {},
+    );
+
+    expect(result.success).toBe(true);
+    expect(modelCaptured.length).toBe(1);
+    expect(modelCaptured[0]).not.toBeNull();
+    expect(typeof modelCaptured[0]).toBe("string");
+
+    // Restore env
+    if (prevBaseUrl == null) delete process.env.OLLAMA_BASE_URL;
+    else process.env.OLLAMA_BASE_URL = prevBaseUrl;
+    if (prevModel == null) delete process.env.OLLAMA_MODEL;
+    else process.env.OLLAMA_MODEL = prevModel;
+  });
+
 });