Advertisement. OOO “TIMEWEB.CLOUD”. INN: 7810945525. Erid: 2W5zFJYfkGG Link to the AI agent - AI agents
A special offer:
- The promo code ELTON to double your payment
- To activate the doubled payment, enter the promo code, and then top up your account
Table of contents
Introduction: From limited AI to a personal agent
Every day we interact with artificial intelligence that helps us solve tasks and find answers. However, its capabilities are often limited, and the answers can suffer from "hallucinations" or be too general.
The reason is simple: standard models have no context about us — our projects, personal notes or specific instructions.
By giving the AI all the necessary information, we don't just make it smarter. We create a personal AI agent capable of automating the routine, from personal tasks to complex business processes. Such an agent can:
- Speed up your learning and your search for information.
- Take over communication with clients.
- Manage your projects, saving time and resources.
In this detailed guide we’ll go step by step through how to create such an agent on the Timeweb Cloud platform without the need to write code, as well as how to integrate it into complex workflows using n8n and Docker/Open WebUI.
The technical basis: RAG, vectors and tokens
Before getting down to creating an agent, it’s important to understand three key concepts that underlie its work: RAG, vectors and tokens.
1. RAG (Retrieval Augmented Generation)
RAG (Retrieval Augmented Generation) translates as "Generation Augmented by Retrieval." It's a technology that solves the main problem of large language models (LLMs) — hallucinations.
Instead of making up information, RAG lets the model first find relevant data in your personal knowledge base, and then use it to generate an accurate answer.
Example: If a regular AI might recommend you books by non-existent authors, then RAG, relying on your documents, will either give an accurate answer or honestly say: “I don’t have enough information.”
2. Vectors and semantic search
The main difference between RAG and traditional search (e.g. Google) is that it searches not by keywords, but by meaning. For this, vectors are used.
Vectors can be imagined as points on a huge map of meaning. The closer two points on this map, the closer their semantic meaning. When you upload files into the knowledge base, the RAG service turns them into millions of such vectors and saves them in a special vector database. This process is called indexing.
3. Tokens: the fuel for AI
Tokens are pieces of text that serve as fuel for the AI’s work. One word can consist of several tokens. Tokens are consumed in three main cases:
-
Indexing: Turning your documents into vectors.
-
The query: Sending your question to the model.
-
The answer: The agent generating the final answer.
Your costs for using AI are measured exactly in tokens.
Timeweb Cloud: a platform for your agent
Timeweb Cloud offers a convenient and functional platform for creating and integrating AI agents.
| Advantage | Description |
|---|---|
| Convenience and ecosystem | A simple control panel where in one place you can rent a server, buy a domain and create an AI agent. |
| Automation | Renting a server lets you use tools like n8n and nocodb for free, to create complex automations without programming. |
| Support | A smart assistant is available that helps with the settings, as well as round-the-clock 24/7 technical support. |
| The financial model | Topping up your balance and buying token packages from popular LLM providers happens in rubles with Russian bank cards. The funds are debited as you use them, which is convenient for budget planning. |
A special offer:
Launch a project in the cloud with the promo code ELTON: https://tweb.link/krcf4 To activate the doubled payment, enter the promo code, and then top up your account
Step 1: Creating an AI agent (choosing a model and plan)
The process starts in the Timeweb Cloud control panel in the “AI agents” section.
- Adding an agent: Go to the section and click “Create” or “Add.”
- Choosing a model: Choose a suitable neural-network model.
An important technical nuance:
It will be impossible to change the chosen model after creating the agent It’s recommended to choose a model that’s optimal in terms of the price-quality ratio, for example, GPT 5 mini (spends tokens on reasoning) or gpt-4o-mini (doesn’t spend tokens on reasoning)
- Choosing a plan (tokens): Determine the necessary number of tokens.
An important technical nuance:
You can increase the plan during work, but it will be impossible to decrease it. Tokens can be bought additionally as needed.
- The initial setup: At the first stage you can skip configuring the system prompt and connecting the knowledge base. We’ll return to them later.
- The order: Check the parameters and click “Order.” Launching the agent will take a few minutes.
Step 2: Configuring and integrating the agent
After launching the agent we move on to configuring it and integrating it into your workflows.
2.1. Configuring the system prompt (the Playground)
For the agent to work for your tasks, it needs a clear instruction — a system prompt. This is done in the “Playground” tab.
The system prompt defines the agent’s role, goal, communication style and limitations. A correctly formulated instruction increases the accuracy and relevance of the answers.
| Recommendation | Description | Example |
|---|---|---|
| The agent’s personality | Specify the name, tasks and area of expertise. | ”You are Webby, a virtual assistant who helps users find answers in the Timeweb Cloud documentation.” |
| The agent’s goal | Formulate the main task, key responsibilities and priorities. | ”Your main task is to provide accurate, clear and complete answers based on the Timeweb Cloud documentation.” |
| Expertise | Clearly specify which data it should rely on, and teach it to avoid made-up examples. | ”Use the Timeweb Cloud documentation as the main source of information. Never make up information yourself.” |
| Limitations | Specify the topics that are forbidden, and what to do if the agent doesn’t know the answer. | ”If you don’t know the answer, say: ‘I don’t have enough information to answer this question.‘” |
Fine-tuning the agent
In the “Playground” you can also configure the generation parameters:
| Parameter | Purpose | A recommendation for RAG |
|---|---|---|
| The maximum number of tokens | Sets the maximum length of the answer. | Adjust depending on the required level of detail. |
| Temperature | The degree of the model’s creativity. | Lower it (closer to 0) so that the agent works strictly according to the provided data and avoids hallucinations. |
| Top P | The probability threshold for choosing words. | Lower values will ensure the focus and coherence of the answers. |
| Improve the query for the search | The agent reformulates your query for a more accurate search in the knowledge base. | It’s recommended to enable it to increase the search accuracy. |
2.2. Integrating the agent
Timeweb Cloud offers several ready-made ways to integrate your agent.
Integration into a site (a widget)
This is the ideal solution for company sites or personal blogs.
The agent can act as an online consultant, helping users navigate and advising potential clients about products.
- Configuring the widget: In the “Embed chat” tab you can configure the appearance and specify the allowed domains.
- Getting the code: Timeweb Cloud provides ready-made JS code to insert.
Integration with a Telegram bot
If you don’t have a site, you can create an online consultant in Telegram. For this it’s enough to:
- Get an API key for your bot via BotFather.
- Insert this key into the corresponding field in the agent’s “Integrations” section.
Using the API
Each agent gets its own API endpoint.
Timeweb Cloud supports two types of API:
- An OpenAI-compatible API: Suitable for integration with external libraries and UIs.
- A native API: Simpler for quick use.
API keys are necessary for authorising requests to the private endpoint and let you set up any complex automations via external services, such as n8n.
We’ll talk about this a bit later.
Step 3: Creating and connecting a knowledge base (the RAG service)
The knowledge base is the heart of your personal agent.
It lets you further train the AI model without the help of engineers, using your documents as context.
- Creating a knowledge base: Return to the “AI agents” section and choose the “Knowledge base” tab. Click “Add.”
- The indexing plan: You’ll be offered a plan for indexing (e.g. 10 million tokens).
Important:
This amount (e.g. 450 rubles) is debited immediately and is intended for turning your documents into vectors. Subsequently, tokens from this package will be spent on indexing new files.
- Uploading data: In the “Data source” section you can upload your files.
A limitation:
The total volume of uploaded files should not yet exceed 50 megabytes. Various formats are supported (e.g. Markdown notes from Obsidian).
- Choosing a server: Choose the server with the database on which your vectors will be stored.
Performance:
The speed of the agent’s access to the knowledge base depends on the choice of server. For individual use, a minimal configuration may be enough.
Payment:
Unlike tokens, the server rental is debited by the day, which lets you plan your expenses flexibly.
- Launching and connecting: After the order, the RAG service will be ready in 10–15 minutes. Return to your AI agent, go to “Management” and choose the created knowledge base.
After connecting the knowledge base, be sure to test the agent in the “Playground” by asking it a question, the answer to which is contained only in your new documents.
One of the most powerful ways to use an AI agent is to integrate it into automated workflows.
We’ll look at how to create a Telegram community-manager bot with n8n, which will answer questions in the chat using the Timeweb Cloud knowledge base.
A prerequisite:
You should have your own instance of n8n deployed (e.g. on a Timeweb Cloud server) and have an API key for your AI agent.
1. Setting up the Telegram trigger
The first step is to set up the node that will listen for incoming messages in Telegram.
This node will launch the workflow when any text message is received in the chat or in the bot’s private messages.
2. A query to the AI agent via the HTTP node
To send a request to the Timeweb Cloud AI Agent we use the HTTP Request node. The Timeweb Cloud agent provides an OpenAI-compatible API, which simplifies integration.
At the end of the URL be sure to add (as in the screenshot):
/chat/completions
Then we configure the other settings of the current module:
- We enable Send Headers:
And we fill in
- Name:
Authorization - Value:
Bearer [and your API key]
To find the API key, we open the Timeweb Cloud settings:
We paste the obtained key next to the word Bearer in n8n (the http node):
- now we create one more Header - “Add Parameter”
Content-Type
application/json
-
Now we need to form the body of the request to the AI (that is, our text received from the Telegram node)
-
now into the JSON field we paste this code, in which we specify exactly what to send - json.data.text is the text received from the Telegram node:
{
"model": "gpt-5-mini",
"messages": [
{
"role": "user",
"content": "{{ $json.message.text }}"
}
],
"max_completion_tokens": 3000,
"stream": false
}
3. Processing the answer and sending it to Telegram
After receiving the answer from the AI agent (the HTTP Request node) we need to extract the answer text and send it back to the user.
- and we fill in the following data in this node:
Chat ID:
{{ $('Telegram Trigger').item.json.message.chat.id }}
Text:
{{ $json.choices[0].message.content }}
And that’s it! Our bot is already working!
The n8n workflow scheme
-
Telegram Trigger (Receives a message)
-
HTTP Request (Sends the message to the Timeweb Cloud agent)
-
Set (Extracts the answer text)
-
Telegram (Sends the answer to the user)
This way, you create a fully automated community manager, trained on your data, that works 24/7.
An advanced case: a corporate AI with Open WebUI and Docker
For teamwork and for providing access to the AI agent without a VPN and complex settings, the Docker + Open WebUI combination is ideal.
Open WebUI provides a convenient, ChatGPT-like interface that can be connected to your Timeweb Cloud AI Agent via its OpenAI-compatible API.
1. Deploying Open WebUI using Docker
This approach lets you get a ChatGPT-like interface for teamwork or personal use, without the need to deploy on a remote server.
1. Prerequisites
- Docker Desktop installed: You should download and install Docker Desktop for your operating system.
What is it?
Docker runs programs in separate boxes, so nothing conflicts. And everything works the same way on any computer
- The Timeweb Cloud AI Agent: You should have an AI agent created and configured in the Timeweb Cloud control panel, with an obtained API key and API endpoint (URL).
2. Launching Open WebUI with one command
Open WebUI is distributed as a Docker image, which lets you launch it with one command in the command line (cmd.exe, PowerShell, Terminal).
- You open the search on Windows
- You type cmd.exe
- The command line opens and you paste the Open WebUI installer there
docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name openwebui --restart always ghcr.io/open-webui/open-webui:main
- What does this code do? First, it launches docker
| Parameter | Description |
|---|---|
| -d | Launches the container in the background (detached). |
| -p 3000:8080 | Forwards port 8080 inside the container to port 3000 on your computer. You’ll access the interface at http://localhost:3000. |
| —name openwebui | Gives the container the convenient name openwebui |
| -v open-webui:/app/backend/data | Creates a Docker volume to save data (settings, users, chat history) between restarts. |
| —restart always | Ensures that the container starts automatically after a restart of Docker or the computer. |
2.2. Checking the launch
After running the command you can check the container’s status:
- Via Docker Desktop: As shown in the screenshot, a container named openwebui should appear in the list. Make sure it’s running (or press the Start button if it’s stopped).
A screenshot from the Docker app:
- We launch the container - we press Start (see the screenshot)
- And we open the local page (works only on your device) -
http://localhost:3000/
On the first launch, Open WebUI will offer you to create an administrator account.
Register using your email and password
Open WebUI supports connecting external models compatible with the OpenAI API.
Your Timeweb Cloud AI Agent uses exactly this standard.
-
Go to Settings: In the Open WebUI interface find the Settings icon or your profile.
-
Adding a connection: Go to the Connections section
-
We click the plus.
-
Now we fill in the connection.
- Configuring the OpenAI-compatible API:
- We copy from the Timeweb Cloud site
- and paste into Open WebUI
- it remains to get the API key, which, as you remember, is under “Management” - Public (availability)
- and we add it to Open WebUI
- be sure to press “Save” in the bottom right corner and it’s done!
After saving the settings, your Timeweb Cloud AI Agent, trained on your corporate knowledge base, will become available for selection in the dropdown list of models in the main chat window of Open WebUI.
This provides:
- A unified interface: Employees work in a familiar chat interface.
- Security: Access to the trained AI without the need to set up a VPN.
- Corporate search: The AI agent acts as a smart search across the company’s internal documentation and instructions.
Examples of application: from personal routine to business automation
Personal cases (Telegram bots)
You can create several assistants for different tasks, using the integration with Telegram:
- A mentor: Helps with goals and tasks, motivates, relying on your personal notes and the books you’ve read.
- A community manager (a detailed case with n8n): Saves your time by automatically answering typical questions in chats, using the knowledge base about your project
This automation is easily implemented via n8n, hosted on a Timeweb Cloud server.
It’s created in the same way as described in the previous example - click
Business cases (automation and growth)
- An online consultant on a site:
Implementing a widget with an AI agent, trained on all the information about your products and services, contributes to revenue growth thanks to the instant reaction to client requests.
- An internal corporate AI (Open WebUI):
Using the Open WebUI technology, you can give employees access to AI with a corporate knowledge base.
The interface resembles ChatGPT, but works without additional network settings (VPN), becoming a smart search across the internal documentation and instructions.
- Complex automations via the API: Using the Timeweb Cloud API key and n8n, you can set up:
- Automatic replies to business emails: An agent trained on corporate knowledge can compose drafts or send ready replies.
- Smart mailings: Setting up personalised mailings about promotions, increasing the probability of sales.
Conclusion
Creating a personal AI agent on Timeweb Cloud is a powerful step towards automation and increasing efficiency.
You get the ability to quickly and simply create assistants for business, routine and self-development, without programmers and complex integrations.
A step-by-step explanation of all the actions, including detailed automation scenarios with n8n and Timeweb Cloud, you’ll always find on my site.
Don’t forget:
On your first top-up, use the promo code ELTON to double the amount in your account.
Sources
- The TimeWeb Cloud documentation on AI agents - Link
© 2026 Elton Labs. All rights reserved.