feat(memory.search): optional minScore RRF threshold

Closes the long-standing backlog item from abfa5463. memory.search now
accepts an optional `minScore` parameter (Zod range 0..1) that drops hits
below the given Reciprocal Rank Fusion score. Default behavior is
unchanged — when minScore is unset, every fused result is returned, same
as today.

The original design memo suggested defaulting to 0.020, but that would
exclude valid pure-semantic matches (one ranker at rank 1 = 1/61 ≈
0.0164). Real-world Phase 2 testing surfaced exactly that case (the
"expose TLS" → HAProxy memory hit). Shipping unfiltered-by-default and
exposing the knob lets specific callers opt into stricter filtering
(e.g. ~0.025 to require two rankers to fire at rank 1) without
penalising legitimate semantic-only hits for the rest.

Tool description updated; per-source rank breakdown remains the primary
confidence signal for the model.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

apps/web/lib/mcp/tools.ts

@@ -736,6 +736,13 @@ const memorySearch: ToolDef = {
       scope: { type: "string", enum: ["project", "user"] },
       tags: { type: "array", items: { type: "string" }, description: "Boost results with these tags." },
       limit: { type: "integer", minimum: 1, maximum: 50, default: 10 },
+      minScore: {
+        type: "number",
+        minimum: 0,
+        maximum: 1,
+        description:
+          "Optional minimum Reciprocal Rank Fusion score for a hit to be returned. Default unset = no extra filter (every fused result returned). Set ~0.025 to require at least two rankers (vector + FTS, or +tag) to fire at rank 1, filtering out weak vector-only matches. The per-source ranks in each result are still the primary way to judge confidence.",
+      },
     },
     required: ["query"],
   },
@@ -743,7 +750,7 @@ const memorySearch: ToolDef = {
     const parsed = MemorySearchInput.safeParse(withDefaultProject(args, ctx));
     if (!parsed.success) return err(parsed.error.message);
 
-    const { query, scope, tags, limit } = parsed.data;
+    const { query, scope, tags, limit, minScore } = parsed.data;
     const requestedKey = projectKeyOrDefault(ctx, parsed.data.project);
     const projectId = requestedKey ? await resolveProjectId(ctx, requestedKey) : null;
     if (requestedKey && !projectId) {
@@ -753,7 +760,7 @@ const memorySearch: ToolDef = {
     const result = await searchMemories(
       ctx.userId,
       query,
-      { scope, projectKey: requestedKey, tags, groupNames: ctx.groups },
+      { scope, projectKey: requestedKey, tags, groupNames: ctx.groups, minScore },
       limit,
     );
 

apps/web/lib/memories.ts

@@ -29,6 +29,12 @@ export interface SearchFilters {
    * shared visibility) — pass through `UserContext.groups`.
    */
   groupNames?: string[];
+  /**
+   * Minimum RRF score a hit must clear. Default `undefined` = no extra
+   * filter (current behavior — every fused result is returned). Set to
+   * e.g. 0.025 to require at least two rankers to fire at rank 1.
+   */
+  minScore?: number;
 }
 
 export interface SearchHit {
@@ -93,7 +99,7 @@ export async function searchMemories(
   filters: SearchFilters = {},
   limit = 20,
 ): Promise<SearchResult> {
-  const { scope, projectKey, tags, groupNames = [] } = filters;
+  const { scope, projectKey, tags, groupNames = [], minScore } = filters;
   const projectId = projectKey
     ? await resolveProjectIdForKey(userId, groupNames, projectKey)
     : null;
@@ -178,7 +184,11 @@ export async function searchMemories(
   fts.forEach((h, i) => accum(h.id, i + 1, "ftsRank"));
   tag.forEach((h, i) => accum(h.id, i + 1, "tagRank"));
 
-  const hits = [...scores.entries()]
+  let entries = [...scores.entries()];
+  if (typeof minScore === "number" && minScore > 0) {
+    entries = entries.filter(([, r]) => r.rrfScore >= minScore);
+  }
+  const hits = entries
     .sort(([, a], [, b]) => b.rrfScore - a.rrfScore)
     .slice(0, limit)
     .map(([id, rank]) => ({