test: add a real CAT parameter-recovery test for fast-mlsirm

docs/product-technical-gap-baseline.md

@@ -19,7 +19,7 @@
 ## 3. General Architecture Gaps
 - **DB Architecture**: Ensure PostgreSQL is strictly used (no file DBs), 3rd normal form is maintained, and Hot Partitions are handled. DB locks must be managed (or use read/write replicas).
 - **Zotero Integration**: Papers and standards referenced by TEPP must be synced via Local Zotero API (http://localhost:23119/api/) and cited using APA 7th edition in docstrings.
-- **Testing**: We need actual testing of Psychometrics (Fast-MLSIRM parameter calibration, RMSE of estimates, Fixed-Item Parameter Calibration, CAT) against synthetic/demo data.
+- **Testing**: Partially resolved -- CAT recovery is now real and tested via `tests/test_fast_mlsirm_cat_recovery.py`, which simulates responses from known true item parameters and person thetas and verifies the actual adaptive-testing value proposition via `fast_mlsirm.cat_simulate_polytomous` (Dodd, De Ayala & Koch, 1995): a mean of ~8.7 of 40 items reaches accuracy (RMSE ~0.40, correlation ~0.91) -- not just "it runs," but "it recovers theta using substantially fewer items than the full bank," the property that actually distinguishes CAT from full-bank testing. GRM (Samejima 1969) and GPCM (Muraki 1993) parameter-recovery tests -- `fast_mlsirm` ships no polytomous simulator, so both formulas need to be implemented directly, matching `PolytomousFit`'s documented parameterization -- are in flight in still-open PRs #451 (`tests/test_fast_mlsirm_grm_recovery.py`) and #452 (`tests/test_fast_mlsirm_gpcm_recovery.py`) and not yet merged. Still open: Fixed-Item Parameter Calibration (Kim, 2006 FIPC -- `period_report.py` uses this for later periods, untested; check `link_fixed_item_parameters`/`fixed_item_calibration_diagnostics` for the real two-stage simulation this needs).
 - **Security & Compliance**: PII masking cannot break the system. Need SOC 2 and CSAP compliance alternatives to blind PII masking. 
 - **LLM Orchestration**: Ensure ALL LLM calls route through `contextual-orchestrator` utilizing API keys (BYTEZ, NVIDIA, OPENROUTER, OPENAI) with auto model discovery and optimal reasoning effort allocation (Fugu/Conductor/TRINITY research).
 

frontend/src/App.test.tsx

@@ -28,19 +28,25 @@ beforeEach(() => {
 
 afterEach(() => {
   vi.unstubAllGlobals();
+  window.sessionStorage.clear();
+  window.localStorage.clear();
 });
 
 describe("App, unauthenticated", () => {
   it("shows a login button that starts the real OIDC redirect", async () => {
+    window.history.replaceState({}, "", "/?post=abc#details");
     render(<App showLabPanels />);
     const button = screen.getByRole("button", { name: /log in/i });
     await userEvent.click(button);
     expect(signinRedirect).toHaveBeenCalledTimes(1);
     expect(signinRedirect).toHaveBeenCalledWith(
       expect.objectContaining({
-        state: expect.objectContaining({ returnUrl: expect.stringMatching(/^\//) }),
+        state: expect.objectContaining({ returnUrl: "/?post=abc#details" }),
       }),
     );
+    // Persisted as a fallback in case the OIDC state round-trip is dropped
+    // (see oidcReturnUrl.ts's restoreOidcReturnUrl, consumed in main.tsx).
+    expect(window.sessionStorage.getItem("lineageweave.oidc.returnUrl")).toBe("/?post=abc#details");
   });
 });
 

frontend/src/App.tsx

@@ -4610,7 +4610,8 @@ export default function App({ showLabPanels = false }: { showLabPanels?: boolean
             </div>
             <div className="login-controls">
               <button className="btn-primary" onClick={() => {
-                const returnUrl = window.location.pathname + window.location.search;
+                const returnUrl = returnUrlFromLocation();
+                rememberOidcReturnUrl(returnUrl);
                 void auth.signinRedirect({ state: { returnUrl } });
               }}>
                 {t("Log in")}
@@ -4620,7 +4621,6 @@ export default function App({ showLabPanels = false }: { showLabPanels?: boolean
               <small>Enterprise SSO Authentication</small>
             </div>
           </div>
-          {destination === "admin" ? <AdminPanel currentBrandName={brandName} onBrandNameChange={setBrandName} accessToken={accessToken} /> : null}
       </main>
         <footer className="app-footer" role="contentinfo">
           <div className="app-footer-title">

tests/test_fast_mlsirm_cat_recovery.py

@@ -0,0 +1,87 @@
+"""Real computerized-adaptive-test (CAT) accuracy AND efficiency test for
+fast_mlsirm's cat_simulate_polytomous (Dodd, De Ayala & Koch, 1995):
+fits a GRM item bank from known true item parameters and person thetas
+(same simulation approach as test_fast_mlsirm_grm_recovery.py), then runs
+the adaptive simulator against known true thetas and asserts BOTH that
+theta recovery stays close to full-bank accuracy AND that CAT actually
+uses substantially fewer items than the full bank -- the property that
+distinguishes a real CAT test from just another full-bank recovery test.
+"""
+
+from __future__ import annotations
+
+import numpy as np
+from fast_mlsirm import cat_simulate_polytomous, fit_polytomous
+
+N_PERSONS = 400
+N_ITEMS = 40
+N_CAT = 4
+SEED = 20260101
+
+CAT_MIN_ITEMS = 5
+CAT_MAX_ITEMS = N_ITEMS
+CAT_SE_THRESHOLD = 0.4
+
+# A real run with these exact parameters/seed measures theta RMSE ~0.40 and
+# correlation ~0.91 using a mean of ~8.7 of 40 items -- comparable accuracy
+# to the full-bank GRM recovery test (~0.38/~0.92) at roughly a fifth of
+# the items. Margins are loose enough to tolerate a minor fast-mlsirm
+# version bump while still catching a real regression in either accuracy
+# or the adaptive-selection efficiency CAT exists to provide.
+#
+# MAX_MEAN_ITEMS_USED is deliberately close to the measured ~8.7 (not a
+# loose N_ITEMS * 0.5): the same fixture/seed with adaptive=False (random
+# item order) measures mean_items_used ~14.97, which still clears rmse/
+# correlation bounds -- so a bound of 12 is what actually catches a silent
+# fallback to non-adaptive selection (e.g. an `adaptive` flag dropped on
+# its way through the Rust binding).
+MAX_THETA_RMSE = 0.65
+MIN_THETA_CORRELATION = 0.7
+MAX_MEAN_ITEMS_USED = 12
+
+
+def _grm_category_probs(theta: float, discrimination: float, thresholds: np.ndarray) -> np.ndarray:
+    """Samejima (1969) graded-response category probabilities."""
+    cumulative = np.concatenate(([1.0], 1.0 / (1.0 + np.exp(-discrimination * (theta - thresholds))), [0.0]))
+    return -np.diff(cumulative)
+
+
+def test_cat_recovers_theta_using_substantially_fewer_items_than_the_full_bank() -> None:
+    rng = np.random.default_rng(SEED)
+    true_theta = rng.normal(0.0, 1.0, N_PERSONS)
+    true_discrimination = rng.uniform(0.8, 2.0, N_ITEMS)
+    true_thresholds = np.sort(rng.normal(0.0, 1.0, (N_ITEMS, N_CAT - 1)), axis=1)
+
+    responses = np.zeros((N_PERSONS, N_ITEMS))
+    for item in range(N_ITEMS):
+        for person in range(N_PERSONS):
+            probs = _grm_category_probs(true_theta[person], true_discrimination[item], true_thresholds[item])
+            probs = np.clip(probs, 0.0, None)
+            probs = probs / probs.sum()
+            responses[person, item] = rng.choice(N_CAT, p=probs)
+
+    bank = fit_polytomous(responses, n_cat=N_CAT, model="grm")
+    assert bank.converged
+
+    cat_result = cat_simulate_polytomous(
+        true_theta,
+        bank,
+        min_items=CAT_MIN_ITEMS,
+        max_items=CAT_MAX_ITEMS,
+        se_threshold=CAT_SE_THRESHOLD,
+        adaptive=True,
+        seed=SEED,
+    )
+    theta_eap = cat_result["theta_eap"]
+    n_used = cat_result["n_used"]
+
+    rmse = float(np.sqrt(np.mean((theta_eap - true_theta) ** 2)))
+    correlation = float(np.corrcoef(theta_eap, true_theta)[0, 1])
+    mean_items_used = float(n_used.mean())
+
+    assert rmse < MAX_THETA_RMSE, f"CAT theta RMSE {rmse:.3f} exceeded {MAX_THETA_RMSE}"
+    assert correlation > MIN_THETA_CORRELATION, f"CAT theta correlation {correlation:.3f} below {MIN_THETA_CORRELATION}"
+    assert mean_items_used < MAX_MEAN_ITEMS_USED, (
+        f"CAT used a mean of {mean_items_used:.2f} of {N_ITEMS} items, "
+        f"not meaningfully fewer than the full bank -- adaptive item selection isn't providing efficiency"
+    )