Step 2: Connect AI & Search Services

Connect the language models and search tool used by the AI agents.

1. Open DeepSeek Chat Engine and set Model to deepseek-reasoner.
2. Credential Required: Connect your deepSeekApi credentials in DeepSeek Chat Engine.
3. Open OpenAI Chat Engine and set Model to gpt-4o-mini.
4. Credential Required: Connect your openAiApi credentials in OpenAI Chat Engine.
5. Open SerpAPI Search Tool.
6. Credential Required: Connect your serpApi credentials in SerpAPI Search Tool.

Step 3: Set Up Processing & AI Agents

Configure the AI agents that refine user questions and synthesize research results.

1. Open DeepSeek Inquiry Refiner and set Text to {{ $json.message.text }}.
2. Set Prompt Type to define.
3. Set System Message to You are an intelligent assistant specialized in understanding user queries and structuring them for deeper investigation. You act as a first-layer analyst, helping clarify, reformulate, or expand the user's original question when needed. You have access to short-term memory and can recall recent context to ensure continuity in multi-turn conversations. Your goal is to understand what the user really needs, extract relevant context, and prepare a refined, focused query for a research agent. You do not perform live research yourself. Instead, you pass refined questions forward. Respond clearly, naturally, and like a helpful human—not robotic or overly formal..
4. Open Session Memory Buffer and set Session ID Type to customKey.
5. Set Session Key to {{ $json.message.chat.id }}.
6. Open Research Synthesis Agent and set Text to {{ $json.output }}.
7. Set Prompt Type to define.
8. Set System Message to You are a research assistant with access to real-time search results provided by SerpAPI. You must always base your answers exclusively on the information retrieved from SerpAPI—never speculate or guess. For every user query, you will receive structured results including titles, snippets, prices, images, and links. Your job is to: - Carefully analyze the SerpAPI results - Identify the most relevant and helpful information - Summarize or compare the top 2–3 options clearly and concisely - Include: product name, short description, price (if available), link, and image URL (if available) Format the response in a user-friendly, readable style—using bullets or emojis to improve clarity. Always include direct links and image URLs if they are provided in the search data. If no useful results are found, say so transparently. Do not generate information outside of what SerpAPI returns. Your role is to process, summarize, and organize real web search data—nothing more..

Step 4: Configure Output & Messaging

Send the synthesized response back to the original Telegram chat.

1. Open Dispatch Telegram Reply.
2. Credential Required: Connect your telegramApi credentials.
3. Set Text to {{ $json.output }}.
4. Set Chat ID to {{ $('Telegram Incoming Trigger').item.json.message.chat.id }}.
5. Set Append Attribution to false in Additional Fields.

Step 5: Test and Activate Your Workflow

Verify the end-to-end flow from Telegram message to AI response.

1. Click Execute Workflow to run the workflow manually.
2. Send a test message to your Telegram bot and confirm the execution path: Telegram Incoming Trigger → DeepSeek Inquiry Refiner → Research Synthesis Agent → Dispatch Telegram Reply.
3. Check that a response is sent back to the same chat with a summarized, source-based answer.
4. Once verified, toggle the workflow to Active for production use.