chat-agent-toolkit 1.2.105 → 1.2.107

package/dist/tools/search/titleGeneratorAgent.d.ts

@@ -0,0 +1,7 @@
+import { LanguageModel } from 'ai';
+import { ChatTurnMessage } from './meta-search-types';
+type TitleGeneratorInput = {
+    chat_history: ChatTurnMessage[];
+};
+declare const generateTitle: (input: TitleGeneratorInput, llm: LanguageModel) => Promise<string>;
+export default generateTitle;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "chat-agent-toolkit",
-  "version": "1.2.105",
+  "version": "1.2.107",
   "description": "Multi-provider AI agent toolkit: generate language responses, search the web, extract content, and manage memory across 10+ LLM providers.",
   "author": "vtempest <grokthiscontact@gmail.com>",
   "license": "rights.institute/PROSPER",

package/src/tools/search/titleGeneratorAgent.ts

@@ -0,0 +1,61 @@
+/**
+ * @module research/chains/titleGeneratorAgent
+ * @description Generates a short, human-friendly title summarising a chat
+ * conversation using the Vercel AI SDK. Used to give conversations with
+ * multiple turns a meaningful title instead of the truncated first message.
+ */
+import { generateText, type LanguageModel } from "ai";
+import { formatChatHistoryAsString } from "../../utils";
+import type { ChatTurnMessage } from "./meta-search-types";
+
+const titleGeneratorPrompt = `
+You are a title generator for an AI powered search engine. You will be given a conversation between a user and an AI. Generate a concise, descriptive title that captures the overall topic of the whole conversation.
+
+Rules for the title:
+- Maximum 6 words, ideally 3-5.
+- No surrounding quotes, no trailing punctuation.
+- Use title case.
+- Describe the subject matter, not the fact that it is a conversation (e.g. "Quantum Computing Basics", not "A Chat About Quantum Computing").
+
+Return only the title between the XML tags <title> and </title>. For example:
+
+<title>SpaceX Starship Launch Timeline</title>
+
+Conversation:
+{chat_history}
+`;
+
+type TitleGeneratorInput = {
+  chat_history: ChatTurnMessage[];
+};
+
+/**
+ * Extracts the title from the model output, tolerating responses that omit
+ * the XML tags. Falls back to the trimmed raw text, stripped of surrounding
+ * quotes and truncated to a sane length.
+ */
+const parseTitle = (text: string): string => {
+  const match = text.match(/<title>([\s\S]*?)<\/title>/i);
+  const raw = (match ? match[1] : text).trim();
+  return raw.replace(/^["'\s]+|["'\s]+$/g, "").slice(0, 80);
+};
+
+const generateTitle = async (
+  input: TitleGeneratorInput,
+  llm: LanguageModel,
+): Promise<string> => {
+  const prompt = titleGeneratorPrompt.replace(
+    "{chat_history}",
+    formatChatHistoryAsString(input.chat_history),
+  );
+
+  const { text } = await generateText({
+    model: llm,
+    temperature: 0,
+    prompt,
+  });
+
+  return parseTitle(text);
+};
+
+export default generateTitle;