nlm-memory 0.4.2 → 0.5.1

package/tests/integration/backfill-facts.test.ts

@@ -1,362 +0,0 @@
-/**
- * Phase B.5 — backfillFacts integration. Seeds a real SQLite store with
- * sessions that have no facts, runs the backfill module against a stub
- * classifier, and verifies facts land + supersedence fires + state file
- * resumes correctly.
- */
-
-import { mkdtempSync, rmSync, existsSync, readFileSync } from "node:fs";
-import { tmpdir } from "node:os";
-import { join, resolve } from "node:path";
-import { afterEach, beforeEach, describe, expect, it } from "vitest";
-import { backfillFacts } from "../../src/core/facts/backfill-facts.js";
-import { SqliteFactStore } from "../../src/core/storage/sqlite-fact-store.js";
-import { SqliteSessionStore } from "../../src/core/storage/sqlite-session-store.js";
-import type {
-  ClassifyResult,
-  EmbedResult,
-  ExtractedFact,
-  LLMClient,
-} from "../../src/ports/llm-client.js";
-import { LLMUnreachableError } from "../../src/ports/llm-client.js";
-import { makeSession } from "../fixtures/sessions.js";
-
-const MIGRATIONS_DIR = resolve(__dirname, "../../migrations");
-
-class ScriptedClassifier implements LLMClient {
-  calls: string[] = [];
-  constructor(
-    private readonly results: Map<string, ClassifyResult>,
-    private readonly errorIds: Set<string> = new Set(),
-  ) {}
-  async embed(): Promise<EmbedResult> {
-    throw new Error("not used");
-  }
-  async classify(transcript: string): Promise<ClassifyResult> {
-    this.calls.push(transcript);
-    if (this.errorIds.has(transcript)) {
-      throw new LLMUnreachableError("test-stub");
-    }
-    const result = this.results.get(transcript);
-    if (!result) throw new Error(`no scripted result for transcript: ${transcript.slice(0, 60)}`);
-    return result;
-  }
-}
-
-function classifyResult(
-  facts: ExtractedFact[],
-  confidence = 0.9,
-): ClassifyResult {
-  return {
-    label: "L",
-    summary: "S",
-    entities: [],
-    decisions: [],
-    open: [],
-    confidence,
-    facts,
-  };
-}
-
-describe("backfillFacts", () => {
-  let tmp: string;
-  let store: SqliteSessionStore;
-  let factStore: SqliteFactStore;
-  let statePath: string;
-
-  beforeEach(() => {
-    tmp = mkdtempSync(join(tmpdir(), "nlm-b5-"));
-    store = new SqliteSessionStore({
-      dbPath: join(tmp, "canonical.sqlite"),
-      migrationsDir: MIGRATIONS_DIR,
-    });
-    factStore = new SqliteFactStore(store.rawDb());
-    statePath = join(tmp, "backfill_facts.state");
-  });
-
-  afterEach(() => {
-    store.close();
-    rmSync(tmp, { recursive: true, force: true });
-  });
-
-  it("writes facts for sessions that have none, skips sessions with existing facts", async () => {
-    // Two sessions need backfill; one already has a fact.
-    store.insertSessionForTest(makeSession({
-      id: "sess_old1", body: "BODY-OLD-1", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    store.insertSessionForTest(makeSession({
-      id: "sess_old2", body: "BODY-OLD-2", startedAt: "2026-05-17T11:00:00Z",
-    }));
-    store.insertSessionForTest(makeSession({
-      id: "sess_done", body: "BODY-DONE", startedAt: "2026-05-17T12:00:00Z",
-    }));
-    // Pre-existing fact on sess_done — backfill should skip it.
-    await factStore.insert({
-      id: "f_pre", kind: "decision", subject: "x", predicate: "framework",
-      value: "v", sourceSessionId: "sess_done", sourceQuote: null,
-      createdAt: "2026-05-17T12:00:00Z", supersededBy: null, confidence: 0.9,
-    });
-
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-OLD-1", classifyResult([
-        { kind: "decision", subject: "nlm-memory-ts", predicate: "framework", value: "Hono" },
-      ])],
-      ["BODY-OLD-2", classifyResult([
-        { kind: "attribute", subject: "mac-pro", predicate: "endpoint", value: ":8080" },
-        { kind: "attribute", subject: "mac-pro", predicate: "model", value: "qwen2.5-3b" },
-      ])],
-    ]));
-
-    const report = await backfillFacts({
-      store, factStore, classifier, statePath,
-    });
-
-    expect(report.total).toBe(2); // sess_done excluded by the NOT EXISTS clause
-    expect(report.processed).toBe(2);
-    expect(report.factsWritten).toBe(3);
-    expect(classifier.calls).toHaveLength(2);
-
-    expect(await factStore.listBySession("sess_old1")).toHaveLength(1);
-    expect(await factStore.listBySession("sess_old2")).toHaveLength(2);
-    expect(await factStore.listBySession("sess_done")).toHaveLength(1); // untouched
-  });
-
-  it("supersedence fires across backfill iterations (B.4 in the backfill path)", async () => {
-    // Two sessions, both assert framework= for the same subject — newer wins.
-    store.insertSessionForTest(makeSession({
-      id: "sess_early", body: "BODY-EARLY", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    store.insertSessionForTest(makeSession({
-      id: "sess_late", body: "BODY-LATE", startedAt: "2026-05-18T10:00:00Z",
-    }));
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-EARLY", classifyResult([
-        { kind: "decision", subject: "x", predicate: "framework", value: "Fastify" },
-      ])],
-      ["BODY-LATE", classifyResult([
-        { kind: "decision", subject: "x", predicate: "framework", value: "Hono" },
-      ])],
-    ]));
-
-    await backfillFacts({ store, factStore, classifier, statePath });
-
-    const current = await factStore.findCurrent("x", "framework");
-    expect(current?.value).toBe("Hono");
-    expect(current?.sourceSessionId).toBe("sess_late");
-
-    const chains = await factStore.getHistory("x", "framework");
-    expect(chains[0]?.history.map((f) => f.value)).toEqual(["Hono", "Fastify"]);
-  });
-
-  it("dry-run reports counts without writing", async () => {
-    store.insertSessionForTest(makeSession({
-      id: "sess_dry", body: "BODY-DRY", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-DRY", classifyResult([
-        { kind: "decision", subject: "x", predicate: "framework", value: "v" },
-      ])],
-    ]));
-
-    const report = await backfillFacts({
-      store, factStore, classifier, statePath, dryRun: true,
-    });
-
-    expect(report.processed).toBe(1);
-    expect(report.factsWritten).toBe(1);
-    expect(await factStore.listBySession("sess_dry")).toHaveLength(0); // not written
-    expect(existsSync(statePath)).toBe(false); // dry-run never touches state
-  });
-
-  it("state file is written after the run and used to skip done ids on reprocess re-runs", async () => {
-    store.insertSessionForTest(makeSession({
-      id: "sess_a", body: "BODY-A", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    store.insertSessionForTest(makeSession({
-      id: "sess_b", body: "BODY-B", startedAt: "2026-05-17T11:00:00Z",
-    }));
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-A", classifyResult([
-        { kind: "decision", subject: "x", predicate: "framework", value: "v" },
-      ])],
-      ["BODY-B", classifyResult([
-        { kind: "decision", subject: "y", predicate: "framework", value: "v" },
-      ])],
-    ]));
-
-    const r1 = await backfillFacts({ store, factStore, classifier, statePath });
-    expect(r1.processed).toBe(2);
-    expect(JSON.parse(readFileSync(statePath, "utf8")).done.sort()).toEqual([
-      "sess_a", "sess_b",
-    ]);
-
-    // Without reprocess, the SQL eligibility filter excludes both — the
-    // happy-path "resume" is implicit (rows already have facts).
-    const r2 = await backfillFacts({
-      store, factStore, classifier: new ScriptedClassifier(new Map()), statePath,
-    });
-    expect(r2.total).toBe(0);
-
-    // With reprocess, eligibility drops the NOT-EXISTS clause; the state
-    // file is what keeps a resumed run from re-classifying done ids. Under
-    // the post-fix semantics, state-file ids are filtered out BEFORE the
-    // work queue is built, so `total` is 0 (empty work queue) and
-    // `skippedAlreadyDone` reports the pre-filter count.
-    const r3 = await backfillFacts({
-      store, factStore, classifier: new ScriptedClassifier(new Map()), statePath,
-      reprocess: true,
-    });
-    expect(r3.total).toBe(0);
-    expect(r3.skippedAlreadyDone).toBe(2);
-    expect(r3.processed).toBe(0);
-  });
-
-  it("--from skips sessions with id <= cutoff", async () => {
-    store.insertSessionForTest(makeSession({
-      id: "sess_aaa", body: "BODY-A", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    store.insertSessionForTest(makeSession({
-      id: "sess_zzz", body: "BODY-Z", startedAt: "2026-05-17T11:00:00Z",
-    }));
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-Z", classifyResult([
-        { kind: "decision", subject: "z", predicate: "framework", value: "v" },
-      ])],
-    ]));
-    const report = await backfillFacts({
-      store, factStore, classifier, statePath, from: "sess_aaa",
-    });
-    expect(report.total).toBe(1);
-    expect(classifier.calls).toEqual(["BODY-Z"]);
-  });
-
-  it("limit caps the batch size", async () => {
-    for (let i = 0; i < 5; i++) {
-      store.insertSessionForTest(makeSession({
-        id: `sess_${i}`, body: `BODY-${i}`, startedAt: `2026-05-17T10:0${i}:00Z`,
-      }));
-    }
-    const map = new Map<string, ClassifyResult>();
-    for (let i = 0; i < 5; i++) {
-      map.set(`BODY-${i}`, classifyResult([
-        { kind: "decision", subject: `s${i}`, predicate: "framework", value: "v" },
-      ]));
-    }
-    const classifier = new ScriptedClassifier(map);
-    const report = await backfillFacts({
-      store, factStore, classifier, statePath, limit: 2,
-    });
-    expect(report.total).toBe(2);
-    expect(report.processed).toBe(2);
-    expect(classifier.calls).toHaveLength(2);
-  });
-
-  it("limit counts processable sessions, not raw SQL rows (filters state-file done BEFORE limit)", async () => {
-    // 5 sessions in the corpus. 3 are already done in the state file (e.g.
-    // previously hit low-confidence). With --limit 2, the OLD behavior would
-    // slice the first 2 SQL rows (both done) and process 0; the NEW behavior
-    // filters out the 3 done ids and then processes the next 2 untouched
-    // sessions — actually doing 2 sessions worth of work as the operator
-    // expects.
-    for (let i = 0; i < 5; i++) {
-      store.insertSessionForTest(makeSession({
-        id: `sess_${i}`, body: `BODY-${i}`, startedAt: `2026-05-17T10:0${i}:00Z`,
-      }));
-    }
-    // Pre-populate state file as if sess_0, sess_1, sess_2 already done
-    // (e.g. via prior low-confidence runs).
-    const fs = await import("node:fs");
-    fs.writeFileSync(statePath, JSON.stringify({ done: ["sess_0", "sess_1", "sess_2"] }));
-
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-3", classifyResult([
-        { kind: "decision", subject: "s3", predicate: "framework", value: "v" },
-      ])],
-      ["BODY-4", classifyResult([
-        { kind: "decision", subject: "s4", predicate: "framework", value: "v" },
-      ])],
-    ]));
-    const report = await backfillFacts({
-      store, factStore, classifier, statePath, limit: 2,
-    });
-    expect(report.skippedAlreadyDone).toBe(3); // pre-filter count
-    expect(report.total).toBe(2);              // work queue after pre-filter
-    expect(report.processed).toBe(2);          // both processed
-    expect(classifier.calls.sort()).toEqual(["BODY-3", "BODY-4"]);
-  });
-
-  it("low-confidence sessions get marked done so a re-run doesn't re-classify them", async () => {
-    store.insertSessionForTest(makeSession({
-      id: "sess_low", body: "BODY-LOW", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-LOW", classifyResult(
-        [{ kind: "decision", subject: "x", predicate: "framework", value: "v" }],
-        0.2,
-      )],
-    ]));
-    const r1 = await backfillFacts({ store, factStore, classifier, statePath });
-    expect(r1.skippedLowConfidence).toBe(1);
-    expect(r1.factsWritten).toBe(0);
-
-    // Re-run uses the state file to skip rather than retrying.
-    const r2 = await backfillFacts({
-      store, factStore, classifier: new ScriptedClassifier(new Map()), statePath,
-    });
-    expect(r2.skippedAlreadyDone).toBe(1);
-  });
-
-  it("stops the whole run when classifier reports LLMUnreachable (don't burn API)", async () => {
-    for (let i = 0; i < 3; i++) {
-      store.insertSessionForTest(makeSession({
-        id: `sess_${i}`, body: `BODY-${i}`, startedAt: `2026-05-17T10:0${i}:00Z`,
-      }));
-    }
-    const classifier = new ScriptedClassifier(
-      new Map(),
-      new Set(["BODY-0"]), // first session immediately errors
-    );
-    const report = await backfillFacts({ store, factStore, classifier, statePath });
-    expect(report.classifyFailures).toBe(1);
-    expect(report.processed).toBe(0);
-    // Should NOT call the classifier on the remaining 2 sessions.
-    expect(classifier.calls).toEqual(["BODY-0"]);
-  });
-
-  it("excludes sessions started at or after the script's cutoff (race with live ingest)", async () => {
-    // Insert a session with a startedAt in the future relative to NOW.
-    const future = new Date(Date.now() + 60 * 60 * 1000).toISOString();
-    store.insertSessionForTest(makeSession({
-      id: "sess_future", body: "BODY-FUTURE", startedAt: future,
-    }));
-    const classifier = new ScriptedClassifier(new Map());
-    const report = await backfillFacts({ store, factStore, classifier, statePath });
-    expect(report.total).toBe(0);
-  });
-
-  it("reprocess=true re-classifies sessions that already have facts", async () => {
-    store.insertSessionForTest(makeSession({
-      id: "sess_repro", body: "BODY-REPRO", startedAt: "2026-05-17T10:00:00Z",
-    }));
-    await factStore.insert({
-      id: "f_existing", kind: "decision", subject: "x", predicate: "framework",
-      value: "old", sourceSessionId: "sess_repro", sourceQuote: null,
-      createdAt: "2026-05-17T10:00:00Z", supersededBy: null, confidence: 0.9,
-    });
-
-    const classifier = new ScriptedClassifier(new Map([
-      ["BODY-REPRO", classifyResult([
-        { kind: "decision", subject: "x", predicate: "framework", value: "new" },
-      ])],
-    ]));
-    const report = await backfillFacts({
-      store, factStore, classifier, statePath, reprocess: true,
-    });
-    expect(report.processed).toBe(1);
-
-    // The DELETE+insertManyInTxn pattern in applyFactsInTxn wipes the old
-    // fact (same source_session_id) and writes the new one.
-    const all = await factStore.listBySession("sess_repro");
-    expect(all.map((f) => f.value)).toEqual(["new"]);
-  });
-});

package/tests/integration/citation-explicit.test.ts

@@ -1,111 +0,0 @@
-/**
- * POST /api/citation/explicit endpoint integration. Verifies that the
- * cite_session MCP tool's daemon POST path writes to the citation log
- * with kind "tool_use" and source "mcp_tool".
- */
-
-import { existsSync, mkdtempSync, readFileSync, rmSync } from "node:fs";
-import { tmpdir } from "node:os";
-import { join, resolve } from "node:path";
-import { afterEach, beforeEach, describe, expect, it } from "vitest";
-import type { Hono } from "hono";
-import { RecallService } from "../../src/core/recall/recall-service.js";
-import { SqliteSessionStore } from "../../src/core/storage/sqlite-session-store.js";
-import { createApp } from "../../src/http/app.js";
-import type { EmbedResult, LLMClient } from "../../src/ports/llm-client.js";
-
-const MIGRATIONS_DIR = resolve(__dirname, "../../migrations");
-
-class FixedEmbedder implements LLMClient {
-  async embed(): Promise<EmbedResult> {
-    return { vector: new Float32Array(768), model: "fixed-test" };
-  }
-  async classify(): Promise<never> {
-    throw new Error("not used");
-  }
-}
-
-describe("POST /api/citation/explicit", () => {
-  let tmp: string;
-  let store: SqliteSessionStore;
-  let app: Hono;
-  let citationLogPath: string;
-
-  beforeEach(() => {
-    tmp = mkdtempSync(join(tmpdir(), "nlm-citation-explicit-"));
-    store = new SqliteSessionStore({
-      dbPath: join(tmp, "canonical.sqlite"),
-      migrationsDir: MIGRATIONS_DIR,
-    });
-    const recall = new RecallService({ store, llm: new FixedEmbedder() });
-    citationLogPath = join(tmp, "citation-log.jsonl");
-    app = createApp({ recall, store, liveStore: store, citationLogPath });
-  });
-
-  afterEach(() => {
-    store.close();
-    rmSync(tmp, { recursive: true, force: true });
-  });
-
-  it("logs a citation entry and returns logged:true", async () => {
-    const res = await app.request("/api/citation/explicit", {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({
-        id: "cc_sub_a139f4ab7ca5aa909",
-        conversation_id: "conv_test_001",
-      }),
-    });
-    expect(res.status).toBe(200);
-    const json = (await res.json()) as Record<string, unknown>;
-    expect(json["logged"]).toBe(true);
-    expect(json["id"]).toBe("cc_sub_a139f4ab7ca5aa909");
-    expect(json["source"]).toBe("mcp_tool");
-  });
-
-  it("writes to the citation log with kind tool_use", async () => {
-    await app.request("/api/citation/explicit", {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({
-        id: "cc_sub_a139f4ab7ca5aa909",
-        reason: "Confirmed FTS5 decision.",
-      }),
-    });
-    expect(existsSync(citationLogPath)).toBe(true);
-    const lines = readFileSync(citationLogPath, "utf8").trim().split("\n");
-    const entry = JSON.parse(lines[0]!) as Record<string, unknown>;
-    expect(entry["cited_id"]).toBe("cc_sub_a139f4ab7ca5aa909");
-    expect(entry["kind"]).toBe("tool_use");
-    expect(entry["response_preview"]).toBe("Confirmed FTS5 decision.");
-  });
-
-  it("returns 400 when id is missing", async () => {
-    const res = await app.request("/api/citation/explicit", {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ conversation_id: "conv_001" }),
-    });
-    expect(res.status).toBe(400);
-  });
-
-  it("returns 400 when body is not JSON", async () => {
-    const res = await app.request("/api/citation/explicit", {
-      method: "POST",
-      headers: { "Content-Type": "text/plain" },
-      body: "not json",
-    });
-    expect(res.status).toBe(400);
-  });
-
-  it("defaults conversation_id to mcp_tool when absent", async () => {
-    await app.request("/api/citation/explicit", {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ id: "cc_sub_a139f4ab7ca5aa909" }),
-    });
-    const lines = readFileSync(citationLogPath, "utf8").trim().split("\n");
-    const entry = JSON.parse(lines[0]!) as Record<string, unknown>;
-    expect(entry["conversation_id"]).toBe("mcp_tool");
-  });
-});

package/tests/integration/cite-event.test.ts

@@ -1,169 +0,0 @@
-/**
- * /api/recall/cite-event endpoint integration. Exercises the citation log
- * append + readback via Hono app.request() against a real RecallService.
- */
-
-import { existsSync, mkdtempSync, readFileSync, rmSync } from "node:fs";
-import { tmpdir } from "node:os";
-import { join, resolve } from "node:path";
-import { afterEach, beforeEach, describe, expect, it } from "vitest";
-import type { Hono } from "hono";
-import { RecallService } from "../../src/core/recall/recall-service.js";
-import { SqliteSessionStore } from "../../src/core/storage/sqlite-session-store.js";
-import { createApp } from "../../src/http/app.js";
-import type { EmbedResult, LLMClient } from "../../src/ports/llm-client.js";
-
-const MIGRATIONS_DIR = resolve(__dirname, "../../migrations");
-
-class FixedEmbedder implements LLMClient {
-  async embed(): Promise<EmbedResult> {
-    return { vector: new Float32Array(768), model: "fixed-test" };
-  }
-  async classify(): Promise<never> {
-    throw new Error("not used");
-  }
-}
-
-describe("POST /api/recall/cite-event", () => {
-  let tmp: string;
-  let store: SqliteSessionStore;
-  let app: Hono;
-  let citationLogPath: string;
-
-  beforeEach(() => {
-    tmp = mkdtempSync(join(tmpdir(), "nlm-cite-"));
-    store = new SqliteSessionStore({
-      dbPath: join(tmp, "canonical.sqlite"),
-      migrationsDir: MIGRATIONS_DIR,
-    });
-    const recall = new RecallService({ store, llm: new FixedEmbedder() });
-    citationLogPath = join(tmp, "citation-log.jsonl");
-    app = createApp({ recall, store, liveStore: store, citationLogPath });
-  });
-
-  afterEach(() => {
-    store.close();
-    rmSync(tmp, { recursive: true, force: true });
-  });
-
-  it("appends a citation entry and returns ok", async () => {
-    const res = await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({
-        conversation_id: "conv-x",
-        cited_id: "cc_sub_a139f4ab7ca5aa909",
-        response_preview: "Per cc_sub_a139f4ab7ca5aa909 we chose FTS5.",
-      }),
-    });
-    expect(res.status).toBe(200);
-    const body = (await res.json()) as { ok: boolean };
-    expect(body.ok).toBe(true);
-
-    expect(existsSync(citationLogPath)).toBe(true);
-    const lines = readFileSync(citationLogPath, "utf8").trim().split("\n");
-    expect(lines).toHaveLength(1);
-    const entry = JSON.parse(lines[0] ?? "{}") as Record<string, unknown>;
-    expect(entry["conversation_id"]).toBe("conv-x");
-    expect(entry["cited_id"]).toBe("cc_sub_a139f4ab7ca5aa909");
-    expect(entry["response_preview"]).toBe(
-      "Per cc_sub_a139f4ab7ca5aa909 we chose FTS5.",
-    );
-    expect(typeof entry["ts"]).toBe("string");
-  });
-
-  it("persists the kind field when provided (tool_use)", async () => {
-    const res = await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({
-        conversation_id: "conv-mcp",
-        cited_id: "cc_sub_a139f4ab7ca5aa909",
-        kind: "tool_use",
-      }),
-    });
-    expect(res.status).toBe(200);
-    const lines = readFileSync(citationLogPath, "utf8").trim().split("\n");
-    const entry = JSON.parse(lines[lines.length - 1] ?? "{}") as Record<string, unknown>;
-    expect(entry["kind"]).toBe("tool_use");
-  });
-
-  it("rejects missing conversation_id", async () => {
-    const res = await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({ cited_id: "cc_sub_x" }),
-    });
-    expect(res.status).toBe(400);
-  });
-
-  it("rejects missing cited_id", async () => {
-    const res = await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({ conversation_id: "conv-x" }),
-    });
-    expect(res.status).toBe(400);
-  });
-
-  it("rejects non-JSON body", async () => {
-    const res = await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: "not json",
-    });
-    expect(res.status).toBe(400);
-  });
-
-  it("GET /api/recall/cite-stats aggregates appended citations", async () => {
-    await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({
-        conversation_id: "c1",
-        cited_id: "cc_sub_aaa111",
-      }),
-    });
-    await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({
-        conversation_id: "c2",
-        cited_id: "cc_sub_aaa111",
-      }),
-    });
-    await app.request("/api/recall/cite-event", {
-      method: "POST",
-      headers: { "content-type": "application/json" },
-      body: JSON.stringify({
-        conversation_id: "c3",
-        cited_id: "cc_sub_bbb222",
-      }),
-    });
-
-    const res = await app.request("/api/recall/cite-stats?days=7");
-    expect(res.status).toBe(200);
-    const stats = (await res.json()) as {
-      total: number;
-      distinct_ids: number;
-      top_ids: { id: string; count: number }[];
-      log_present: boolean;
-    };
-    expect(stats.total).toBe(3);
-    expect(stats.distinct_ids).toBe(2);
-    expect(stats.log_present).toBe(true);
-    expect(stats.top_ids[0]?.id).toBe("cc_sub_aaa111");
-    expect(stats.top_ids[0]?.count).toBe(2);
-  });
-
-  it("GET /api/recall/cite-stats returns zero-totals when log is absent", async () => {
-    const res = await app.request("/api/recall/cite-stats?days=14");
-    expect(res.status).toBe(200);
-    const stats = (await res.json()) as {
-      total: number;
-      log_present: boolean;
-    };
-    expect(stats.total).toBe(0);
-    expect(stats.log_present).toBe(false);
-  });
-});