arisa 4.1.12 → 4.1.14

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arisa",
-  "version": "4.1.12",
+  "version": "4.1.14",
   "description": "Telegram + Pi Agent modular assistant",
   "type": "module",
   "main": "src/index.js",

package/src/core/agent/agent-manager.js

@@ -57,6 +57,8 @@ async function promptAndThrowOnAssistantError(session, prompt) {
 
 function inferDeliveryMethod(artifact) {
   if (artifact.kind === "audio" || (artifact.mimeType || "").startsWith("audio/")) return "audio";
+  if (artifact.kind === "image" || (artifact.mimeType || "").startsWith("image/")) return "photo";
+  if (artifact.kind === "video" || (artifact.mimeType || "").startsWith("video/")) return "video";
   return "document";
 }
 

package/src/transport/telegram/bot.js

@@ -48,7 +48,23 @@ function getIncomingMessageText(message) {
   return message?.text || message?.caption || "";
 }
 
-function buildPrompt({ ctx, artifact, transcript, toolResult }) {
+function baseMimeType(mimeType = "") {
+  return mimeType.split(";")[0].trim().toLowerCase();
+}
+
+function isInlineTextArtifact(artifact, messageText) {
+  return artifact?.kind === "text"
+    && baseMimeType(artifact.mimeType) === "text/plain"
+    && typeof artifact.text === "string"
+    && artifact.text === messageText;
+}
+
+export function shouldIncludeArtifactReference({ artifact, messageText = "" } = {}) {
+  if (!artifact) return false;
+  return !isInlineTextArtifact(artifact, messageText);
+}
+
+export function buildPrompt({ ctx, artifact, transcript, toolResult }) {
   const parts = [
     `Incoming Telegram message.`,
     `chatId: ${ctx.chat.id}`,
@@ -60,10 +76,12 @@ function buildPrompt({ ctx, artifact, transcript, toolResult }) {
   const messageText = getIncomingMessageText(ctx.message);
   if (messageText) parts.push(`text: ${messageText}`);
   parts.push(...quotedMessageSummary(ctx.message?.reply_to_message));
-  if (artifact?.path) parts.push(`artifactPath: ${artifact.path}`);
-  if (artifact?.id) parts.push(`artifactId: ${artifact.id}`);
-  if (artifact?.mimeType) parts.push(`mimeType: ${artifact.mimeType}`);
-  if (artifact?.kind) parts.push(`kind: ${artifact.kind}`);
+  if (shouldIncludeArtifactReference({ artifact, messageText })) {
+    if (artifact?.path) parts.push(`artifactPath: ${artifact.path}`);
+    if (artifact?.id) parts.push(`artifactId: ${artifact.id}`);
+    if (artifact?.mimeType) parts.push(`mimeType: ${artifact.mimeType}`);
+    if (artifact?.kind) parts.push(`kind: ${artifact.kind}`);
+  }
   if (transcript) {
     parts.push(`transcriptArtifactId: ${transcript.id}`);
     parts.push(`transcriptText: ${transcript.text}`);
@@ -91,21 +109,24 @@ function buildNewSessionPrompt(ctx) {
 }
 
 async function buildAsyncTaskPrompt({ task, artifactStore, toolRegistry, logger }) {
+  const taskText = task.payload.prompt || "";
   const parts = [
     "Scheduled task fired.",
     `taskId: ${task.id}`,
     `chatId: ${task.payload.chatId}`,
-    task.payload.prompt ? `text: ${task.payload.prompt}` : null
+    taskText ? `text: ${taskText}` : null
   ];
 
   if (task.payload.artifactId) {
     const chatArtifactStore = artifactStore.forChat(task.payload.chatId);
     const artifact = await chatArtifactStore.get(task.payload.artifactId);
     if (artifact) {
-      parts.push(`artifactPath: ${artifact.path || ""}`);
-      parts.push(`artifactId: ${artifact.id}`);
-      parts.push(`mimeType: ${artifact.mimeType}`);
-      parts.push(`kind: ${artifact.kind}`);
+      if (shouldIncludeArtifactReference({ artifact, messageText: taskText })) {
+        parts.push(`artifactPath: ${artifact.path || ""}`);
+        parts.push(`artifactId: ${artifact.id}`);
+        parts.push(`mimeType: ${artifact.mimeType}`);
+        parts.push(`kind: ${artifact.kind}`);
+      }
 
       const { normalizedArtifact, toolResult } = await normalizeArtifactForReasoning({
         artifact,
@@ -431,6 +452,8 @@ export async function createTelegramBot({ config, artifactStore, toolRegistry, t
         const input = new InputFile(filePath, filename || undefined);
         if (method === "voice") return bot.api.sendVoice(chatId, input, { caption });
         if (method === "document") return bot.api.sendDocument(chatId, input, { caption });
+        if (method === "photo" || method === "image") return bot.api.sendPhoto(chatId, input, { caption });
+        if (method === "video") return bot.api.sendVideo(chatId, input, { caption });
         return bot.api.sendAudio(chatId, input, { caption });
       }
     };

package/src/transport/telegram/media.js

@@ -128,7 +128,7 @@ export async function captureIncomingArtifact(ctx, artifactStore) {
     return store.createText({
       text: ctx.message.text,
       source: baseSource,
-      metadata: {}
+      metadata: { visibility: "internal", representation: "inline-message" }
     });
   }
 

package/test/telegram-text-artifact.test.js

@@ -0,0 +1,81 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+import { buildPrompt, shouldIncludeArtifactReference } from "../src/transport/telegram/bot.js";
+import { captureIncomingArtifact } from "../src/transport/telegram/media.js";
+
+function createTextContext(text = "hello") {
+  return {
+    chat: { id: 123 },
+    from: { id: 456, username: "martin" },
+    msg: { message_id: 789 },
+    message: {
+      message_id: 789,
+      text
+    }
+  };
+}
+
+test("does not expose inline message text artifacts as prompt attachments", () => {
+  const prompt = buildPrompt({
+    ctx: createTextContext("hello"),
+    artifact: {
+      id: "artifact-1",
+      kind: "text",
+      mimeType: "text/plain",
+      text: "hello"
+    }
+  });
+
+  assert.match(prompt, /text: hello/);
+  assert.doesNotMatch(prompt, /artifactId: artifact-1/);
+  assert.doesNotMatch(prompt, /mimeType: text\/plain/);
+  assert.doesNotMatch(prompt, /kind: text/);
+});
+
+test("keeps distinct artifacts visible to the prompt", () => {
+  assert.equal(
+    shouldIncludeArtifactReference({
+      artifact: {
+        id: "artifact-1",
+        kind: "document",
+        mimeType: "text/plain",
+        path: "/tmp/note.txt"
+      },
+      messageText: "see attached"
+    }),
+    true
+  );
+
+  assert.equal(
+    shouldIncludeArtifactReference({
+      artifact: {
+        id: "artifact-2",
+        kind: "text",
+        mimeType: "text/plain",
+        text: "different text"
+      },
+      messageText: "hello"
+    }),
+    true
+  );
+});
+
+test("marks incoming Telegram text artifacts as internal inline messages", async () => {
+  const calls = [];
+  const artifactStore = {
+    forChat: (chatId) => ({
+      createText: async (request) => {
+        calls.push({ chatId, ...request });
+        return { id: "artifact-1", chatId, kind: "text", mimeType: "text/plain", ...request };
+      }
+    })
+  };
+
+  const artifact = await captureIncomingArtifact(createTextContext("hello"), artifactStore);
+
+  assert.equal(artifact.text, "hello");
+  assert.deepEqual(calls[0].metadata, {
+    visibility: "internal",
+    representation: "inline-message"
+  });
+});