wazap-mcp 0.18.0 → 0.18.2

package/dist/messages.js

@@ -399,9 +399,9 @@ export function searchableText(raw, transcript) {
     if (type === "reaction" || type === "deleted" || type === "system")
         return null;
     const own = rule.text?.(node)?.trim() || rule.caption?.(node)?.trim() || rule.detail?.(node)?.trim() || "";
-    // No letter or digit means no words to embed — a "🥰🥰" or "..." only adds
-    // noise that outranks real hits on short queries.
-    if (!/[\p{L}\p{N}]/u.test(own) && spoken === undefined)
+    // Under five letters or digits there is no meaning to embed — a "🥰🥰", a
+    // "Da" or a "..." only adds noise that outranks real hits on short queries.
+    if ((own.match(/[\p{L}\p{N}]/gu)?.length ?? 0) < 5 && spoken === undefined)
         return null;
     return viewText(raw, transcript);
 }

package/dist/recall/engine.js

@@ -211,13 +211,15 @@ class UrlBackend {
  */
 export class EmbedEngine {
     target;
-    constructor(target) {
+    spec;
+    constructor(target, spec) {
         this.target = target;
+        this.spec = spec;
     }
     static async start(settings, spec, onLog = () => { }) {
         if (settings.embedUrl !== null) {
             const base = settings.embedUrl.replace(/\/+$/, "");
-            return new EmbedEngine(new UrlBackend(base));
+            return new EmbedEngine(new UrlBackend(base), spec);
         }
         const bin = findLlama(settings);
         if (bin === null) {
@@ -225,18 +227,24 @@ export class EmbedEngine {
         }
         const sidecar = new LlamaSidecar(bin, embedModelPath(settings.modelsDir, spec), onLog);
         await sidecar.start();
-        return new EmbedEngine(sidecar);
+        return new EmbedEngine(sidecar, spec);
     }
-    async embed(texts) {
+    /**
+     * `kind` picks the model's task prefix: the index holds "document" texts,
+     * searches embed "query". The prefix is the model's side of a retrieval
+     * pair, not part of what the index stores.
+     */
+    async embed(texts, kind) {
         if (texts.length === 0)
             return [];
         await this.target.waitReady();
+        const input = texts.map((text) => `${this.spec.prompts[kind]}${text}`);
         let response;
         try {
             response = await fetch(`${this.target.base}${EMBED_PATH}`, {
                 method: "POST",
                 headers: { "content-type": "application/json" },
-                body: JSON.stringify({ content: texts.length === 1 ? texts[0] : texts }),
+                body: JSON.stringify({ content: input.length === 1 ? input[0] : input }),
                 signal: AbortSignal.timeout(REQUEST_TIMEOUT_MS),
             });
         }
@@ -251,7 +259,7 @@ export class EmbedEngine {
         if (!Array.isArray(reply)) {
             throw new WazapError("RECALL_FAILED", `embedding server refused: ${reply.error?.message ?? "bad reply"}`);
         }
-        if (reply.length !== texts.length) {
+        if (reply.length !== input.length) {
             throw new WazapError("RECALL_FAILED", `embedding server returned ${reply.length} vectors for ${texts.length} texts`);
         }
         return reply.map((item, i) => {

package/dist/recall/models.js

@@ -22,6 +22,8 @@ export const EMBED_MODELS = {
         sha256: "b5ce9d77a3fc4b3b39ccb5643c36777911cc4eb46a66962eadfa3f5f60490d63",
         dims: 768,
         url: "https://huggingface.co/ggml-org/embeddinggemma-300M-GGUF/resolve/main/embeddinggemma-300M-Q8_0.gguf",
+        // EmbeddingGemma's own retrieval task, from its model card.
+        prompts: { query: "task: search result | query: ", document: "title: none | text: " },
     },
     "e5-base-multilingual": {
         alias: "e5-base-multilingual",
@@ -30,6 +32,8 @@ export const EMBED_MODELS = {
         sha256: "548c31b068947aa26b86c8bbfc1f2fabe5233f6d0e1241319832b20a01e5968a",
         dims: 768,
         url: "https://huggingface.co/dinab/multilingual-e5-base-Q8_0-GGUF/resolve/main/multilingual-e5-base-q8_0.gguf",
+        // e5's documented asymmetric prefixes.
+        prompts: { query: "query: ", document: "passage: " },
     },
 };
 export function embedModelSpec(name) {

package/dist/recall/settings.js

@@ -11,6 +11,13 @@ const ON = new Set(["local", "on", "1", "yes", "true"]);
 const MODEL_ALIASES = ["embeddinggemma-300m", "e5-base-multilingual"];
 const DEFAULT_MAX_ROWS = 50_000;
 const MIN_MAX_ROWS = 100;
+/**
+ * Cosine floor for embeddinggemma-300m under its task prompts, measured on a
+ * real index: noise tops out ~0.31, real paraphrases start ~0.35. The prompts
+ * widened the noise/signal gap enough for the floor to mean something.
+ * e5-base needs its own calibration.
+ */
+const DEFAULT_MIN_SIMILARITY = 0.35;
 function parseEnabled(raw) {
     const value = stripPasted(raw ?? "").toLowerCase();
     if (OFF.has(value))
@@ -36,6 +43,15 @@ function parseMaxRows(raw) {
         return n;
     throw new WazapError("INVALID_ID", `WAZAP_RECALL_MAX must be a number >= ${MIN_MAX_ROWS}, got "${value}".`, "Fix WAZAP_RECALL_MAX or remove it");
 }
+function parseMinSimilarity(raw) {
+    const value = stripPasted(raw ?? "");
+    if (value === "")
+        return DEFAULT_MIN_SIMILARITY;
+    const n = Number(value);
+    if (Number.isFinite(n) && n >= 0 && n <= 1)
+        return n;
+    throw new WazapError("INVALID_ID", `WAZAP_RECALL_MIN_SIMILARITY must be a number between 0 and 1, got "${value}".`, "Fix WAZAP_RECALL_MIN_SIMILARITY or remove it");
+}
 function parseUrl(raw) {
     const value = stripPasted(raw ?? "");
     if (value === "")
@@ -59,5 +75,6 @@ export function readRecallSettings(env, dataDir) {
         embedUrl: parseUrl(env.WAZAP_EMBED_URL),
         modelsDir: join(dataDir, "models"),
         maxRows: parseMaxRows(env.WAZAP_RECALL_MAX),
+        minSimilarity: parseMinSimilarity(env.WAZAP_RECALL_MIN_SIMILARITY),
     };
 }

package/dist/recall/store.js

@@ -21,7 +21,8 @@ import { WazapError } from "../errors.js";
 export const TEXT_CAP = 2048;
 /** Rewrite meta+vectors when more than this share of rows is dead. */
 const COMPACT_DEAD_RATIO = 0.3;
-const STATE_VERSION = 1;
+/** v2: model task prompts — a raw-embedded index belongs to a different world. */
+const STATE_VERSION = 2;
 const QUANT = "int8";
 /** The index holds message text; it gets history's permissions, not the defaults. */
 const DIR_MODE = 0o700;
@@ -334,6 +335,7 @@ export class RecallStore {
      */
     query(q, nowMs = Date.now()) {
         const unit = normalize(q.vector);
+        const floor = q.minSimilarity ?? 0;
         const hits = [];
         for (const record of this.live.values()) {
             if (q.chatId !== undefined && record.jid !== q.chatId)
@@ -349,7 +351,7 @@ export class RecallStore {
             for (let i = 0; i < this.spec.dims; i++)
                 dot += unit[i] * this.vectors[offset + i];
             const similarity = dot / 127;
-            if (similarity <= 0)
+            if (similarity <= 0 || similarity < floor)
                 continue;
             hits.push({ record, similarity, score: similarity * recencyDecay(nowMs - record.ts) });
         }

package/dist/tools.js

@@ -503,6 +503,8 @@ Each result carries its date and a score: semantic similarity scaled by
 recency, so fresh matches rank first. chat_id, since, until and from narrow
 the search exactly like search_messages. A hit marked "index only" lives in
 the index alone: quote it, but get_message and download_media cannot see it.
+Results under the similarity floor are dropped rather than listed; when only
+weak matches survive, the output says so — do not present them as found facts.
 
 RECALL_UNAVAILABLE means recall is off or the embedding setup is missing; the
 fix names the command the user has to run. Do not retry it.`,
@@ -1121,7 +1123,13 @@ function renderRecall(title, answer) {
     if (hits.length === 0) {
         return `${title}: no messages found.${catchingUp ? ` ${catchingUp}` : ""}`;
     }
+    // Under ~0.55 cosine, embeddinggemma matches are usually coincidental — the
+    // agent must not present them as found facts.
+    const weak = Math.max(...hits.map((h) => h.similarity)) < 0.55;
     const lines = [`# ${title} (${hits.length})`, ""];
+    if (weak) {
+        lines.push(`Weak matches only (best similarity ${Math.max(...hits.map((h) => h.similarity)).toFixed(2)}): the query may have no real answer — treat these as guesses.`, "");
+    }
     if (catchingUp)
         lines.push(catchingUp, "");
     const introduced = new Set();

package/dist/whatsapp.js

@@ -624,9 +624,10 @@ export class WhatsAppService {
             const store = this.readyRecall();
             const scope = chatId === undefined ? undefined : this.resolveId(chatId);
             const from = opts.from === undefined ? undefined : opts.from === "me" ? this.ownJid() : this.resolveId(opts.from);
-            const [vector] = await this.recallEmbed([query]);
+            const [vector] = await this.recallEmbed([query], "query");
+            const minSimilarity = this.recallEnv instanceof WazapError ? undefined : this.recallEnv.minSimilarity;
             const hits = store
-                .query({ vector: vector, chatId: scope, sinceMs: opts.sinceMs, untilMs: opts.untilMs, from, limit })
+                .query({ vector: vector, chatId: scope, sinceMs: opts.sinceMs, untilMs: opts.untilMs, from, minSimilarity, limit })
                 .map((hit) => {
                 const live = this.store.messages.has(hit.record.sid);
                 return {
@@ -1236,7 +1237,7 @@ export class WhatsAppService {
         try {
             const spec = EMBED_MODELS[this.recallEnv.model];
             this.recallStore = await RecallStore.open(join(this.paths.root, "recall"), spec, this.recallEnv.maxRows);
-            this.recallQueue = new RecallQueue(this.recallStore, (texts) => this.recallEmbed(texts));
+            this.recallQueue = new RecallQueue(this.recallStore, (texts) => this.recallEmbed(texts, "document"));
             if (this.recallStore.count > 0)
                 log(`recall index: ${this.recallStore.count} messages`);
         }
@@ -1269,9 +1270,9 @@ export class WhatsAppService {
         }
         return this.recallEngineP;
     }
-    async recallEmbed(texts) {
+    async recallEmbed(texts, kind) {
         const engine = await this.recallEngine();
-        return engine.embed(texts);
+        return engine.embed(texts, kind);
     }
     /**
      * The ops one raw message turns into: the tombstone a revoke carries for the

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wazap-mcp",
-  "version": "0.18.0",
+  "version": "0.18.2",
   "mcpName": "io.github.razvangirgiz/wazap",
   "description": "WhatsApp for your AI agent. MCP server over Baileys: pairing-code login, several accounts, 33 tools, stdio or HTTP.",
   "license": "MIT",