Setting up the Claude bot or an alternative for our tasks

Vitalik, hello

I will develop everything you wish and even more, I am free for work.
I can start everything today.

Preliminary consultation 2500 UAH.

Show original

Translate

Hello.

I carefully studied the technical specifications. Essentially, this is not just about simple configuration of Claude or Ollama, but about building a full-fledged autonomous AI system with memory, RAG, browser management, access to Windows, and the ability to execute scripts independently.

For the past few years, I have been developing AI solutions, automating business processes, and integrating LLM models into real working environments. I have worked with Ollama, Qwen, DeepSeek, Llama, LangGraph, CrewAI, ChromaDB, Qdrant, Playwright, Selenium, Open Interpreter, and automation systems on Windows.

What I can implement within your specifications:

✔ Deployment and configuration of a remote Windows server;
… ✔ Installation and configuration of Ollama with local models;
✔ Long-term memory for the agent through RAG and vector databases;
✔ Loading and training on PDF, TXT, Markdown, video, and other materials;
✔ Browser management via Playwright/Selenium;
✔ Performing actions in Windows, working with files and applications;
✔ Automatic execution of scripts and tasks on a schedule;
✔ Telegram/Web interface for agent management;
✔ Auto-start after reboot and system status monitoring;
✔ Backup and basic security mechanisms.

I particularly like your approach to architecture: agent memory, local models, and autonomous operation without constant human involvement. This allows for the creation of a truly useful system, rather than just another chatbot with a fancy name.

I am ready to propose a working architecture, select the optimal stack for your budget, and deploy the system turnkey with instructions for further operation.

I am ready to start immediately after discussing the details.

Show original

Translate

Hello.

Your project interests me. I can help with the development of an autonomous AI agent for a remote Windows PC with the ability to work with materials, a browser, files, applications, and execute specified scenarios.

I would like to note that such a system is better built in stages: first a secure MVP with a limited set of tasks and action control, then expanding autonomy after testing. This is important for stable operation 24/7 and protection against erroneous or undesirable actions of the agent.

Before providing an accurate estimate, I need to clarify:

— what specific tasks the agent should perform first;
… — which websites, services, and applications it should use;
— whether a fully local launch is required or a hybrid scheme is acceptable;
— whether there is already a remote Windows PC or if one needs to be selected and configured;
— whether a management interface is needed: Telegram, web panel, or another option;
— what actions the agent should perform only after human confirmation;
— how much material needs to be uploaded to the system at the start.

I can propose a phased implementation: first an MVP with basic autonomous operation, memory, material processing, task management, and test scenarios. After checking stability, we can expand functionality and the level of autonomy.

I am ready to discuss the details and propose a realistic launch plan.

Show original

Translate

Hello, Vitaly!

I already have a similar autonomous agent working for myself, so most of the points in your technical specification are already implemented — I can show it on the finished product.

What has already been done and works:

- Autonomous 24/7 with auto-start after reboot (Windows Scheduled Task — I work on Windows, no need to port from Linux);
- Ingestion of materials: PDF, text, YouTube → transcription via Whisper → summary → stored in the agent's memory;
- Browser control using Playwright, plus a Telegram bot and a web dashboard for logs and management.
…
Regarding security: the agent will have full access to Windows, PowerShell, and files, so unrestricted autonomy is dangerous. I have implemented it this way: before a potentially dangerous action (deleting files, system commands), the agent asks for confirmation instead of executing silently, and each of its steps is logged — you can see what it did. I will apply the same approach here.

Stack according to your preferences: Ollama (Qwen/DeepSeek/Llama), LangGraph/CrewAI, Qdrant/ChromaDB, Playwright/Browser Use, Open Interpreter/PyAutoGUI, Whisper, FastAPI.

I suggest starting with a small first phase so you can see the results before making major investments. I can show you my agent — screenshots or a short demo. What tasks are a priority for the agent? I will respond with specifics on implementation.

Show original

Translate

Good day! We can assist with deploying a standalone AI agent on a remote Windows PC / Windows Server. The project is clear: it is necessary not just to install the model, but to build a functioning agent system with memory, RAG, browser automation, access to the Windows environment, and the ability to perform tasks 24/7. We can implement:
— configuration of remote Windows PC / RDP
— installation of Ollama and local models
— building AI-agent workflow through LangGraph / CrewAI
— long-term memory through ChromaDB / Qdrant
— RAG pipeline for PDFs, texts, courses, and videos
— browser automation through Playwright / Selenium
— desktop automation through PyAutoGUI / Windows tools
— auto-start after reboot
… — Telegram / web interface for agent management
— logs, notifications, backup, and basic security constraints.
To start, we propose an MVP:
Windows VPS → Ollama → agent workflow → vector memory → browser automation → material ingestion → Telegram management.
After this, the system can be scaled for more complex workflows, local LLMs, a scheduler, and a full dashboard.
We are ready to discuss the agent's tasks, required autonomy, hardware/GPU requirements, and access level to Windows.

Show original

Translate

Hello, write to me in private messages and we will discuss all the details, but as colleagues mentioned, it's better on Linux, but we will figure something out.

Show original

Translate

Your Technical Assignment is composed very competently. The task of deploying a standalone OS/Browser-Use AI agent 24/7 on a remote Windows system is fully understood by us. At Lumvex studio, we specialize in complex AI integrations, the development of AI agents, and automation pipelines.

We understand the key pitfalls of such systems (for example, the loss of GUI context in Windows when closing the RDP session, VRAM requirements for local LLMs, and RAG memory synchronization in LangGraph).

Our technical solution for your points:
Remote PC Infrastructure:

We will select and configure a dedicated server (we recommend Hetzner / AWS with GPU, such as NVIDIA RTX 3090/4090 or A10G for comfortable operation of models with 14B/32B parameters).

… We will set up a virtual display (Virtual Display Driver) so that when disconnected from RDP, the agent's screen does not "disappear" and GUI automation (PyAutoGUI / Open Interpreter) continues to see the Windows interface 24/7.

We will configure the automatic start of all services via Windows Task Scheduler / NSSM (Non-Sucking Service Manager).

Agent Brain and Memory (Ollama + CrewAI/LangGraph + Qdrant):

We will deploy Ollama with Qwen2.5-Coder / DeepSeek-R1-Distill models (optimal models for Tool Calling and writing automation scripts).

The architecture of the agent will be built on LangGraph. This will allow the agent to be cyclical and controllable (State Management).

Memory will be implemented through a vector database Qdrant or ChromaDB. We will implement a RAG pipeline: downloadable PDFs/texts will be chunked, embedded (via a local embedding model in Ollama), and stored in the database.

Browser and System Management (Browser-Use & Open Interpreter):

For web surfing, we will implement a combination of Browser-Use + Playwright. The agent will be able to see the DOM tree and screenshots of pages, click, fill out forms, and bypass basic protections.

For managing the Windows OS, we will use a custom layer based on Open Interpreter and PowerShell/PyAutoGUI with strict system limitations (Safe Guardrails) to prevent the agent from damaging system files.

Training (Ingestion Pipeline) & Interface:

We will integrate a local Whisper for transcribing YouTube videos and audio materials.

To manage the agent, upload training files, and view logs in real-time, we will deploy a convenient Web-UI (Streamlit / FastHTML) or create a closed Telegram bot (with commands /start_task, /upload_doc, /view_logs).

Show original

Translate

Good afternoon. I read the technical specifications — an autonomous AI agent is needed on a remote Windows server with long-term memory through RAG, browser and system management, an ingestion pipeline for PDF/MD/YouTube videos, and a Telegram control interface. Your stack choice is already close to optimal — I would like to make a few clarifications on the components based on my experience with similar setups.

Regarding the LLM engine: Ollama is a good choice for easy model deployment, but for long sessions with tool use, vLLM or llama.cpp server directly is more stable — Ollama can have unpredictable delays with large prompts. For models, if a GPU is available, Qwen2.5-Coder-32B or DeepSeek-Coder-V2 works well; on CPU only — Qwen2.5-7B-Instruct at most. For the agent framework, LangGraph is more convenient than CrewAI for long-lived agents because it has a clear state machine — this is critical for 24/7 operation and recovery after restarts.

Memory: I would choose Qdrant instead of ChromaDB for long-term storage — it scales better and has a decent hybrid search (dense + sparse). RAG pipeline on LlamaIndex plus custom ingestion handlers for each format: PDF through unstructured.io, YouTube through yt-dlp + Whisper, Markdown directly. I perform vectorization through sentence-transformers locally to avoid dependency on OpenAI embeddings.

Browser automation — Playwright is better than Selenium because it has built-in auto-wait and is much more reliable on dynamic content. Browser Use on top of Playwright works but adds instability — for production, I would prefer Playwright directly plus my own narrow API from tools.

System access: Open Interpreter is convenient for prototyping, but in a long-running agent, it’s better to provide a Python execution environment through RestrictedPython or Docker-in-Docker. PyAutoGUI should be kept for GUI operations, PowerShell for system commands via subprocess.
…
Telegram interface — aiogram plus webhook on built-in FastAPI; for GUI on top — a simple Streamlit/Gradio dashboard that sees logs, agent memory, and current tasks.

For remote Windows: it’s easiest to set up a dedicated Windows VPS with GPU if necessary from Hetzner/Contabo or OVH. If the budget doesn’t allow for a dedicated GPU — I recommend sticking to CPU models and connecting an external Claude/GPT API as a fallback for heavy tasks, which often turns out to be cheaper than keeping a GPU 24/7.

Regarding security: a whitelist of allowed commands, a mandatory confirm-step for file system writes outside the working directory, and separate restrictions on network requests.

Relevant production experience — a voice AI agent in production on RAG plus Qdrant plus LLM orchestration for veterinary clinics, an AI platform for Telegram on MTProto, and a current side project with MCP servers and parallel agent orchestration on 50+ tasks per session.

If you want, I can send a preliminary scheme of components with boundaries of responsibility before submitting the final estimate. What materials are expected to be uploaded to RAG at the start — PDF documentation plus video courses, or any other formats?

Show original

Translate

Hello, Vitaly!
The task fully matches my profile: I have experience in creating autonomous AI agents, integrating LLM, and automating operating systems.
I am ready to implement the project "turnkey" based on the stack you proposed:
• Infrastructure: Deployment of Windows Server/Cloud PC with GPU, setting up 24/7 auto-start and stable RDP.
• Agent's brain: Local models (Qwen/DeepSeek/Llama) via Ollama + orchestration of logic on LangGraph / CrewAI.
• Memory (RAG): Storing knowledge in the vector database ChromaDB / Qdrant with a pipeline for processing PDF, text, and YouTube (Whisper).
• OS & Browser Automation: Using Browser Use / Playwright for web sessions and Open Interpreter / PyAutoGUI for full control of Windows (GUI, PowerShell, CMD).
• Interface: User-friendly Web UI or Telegram bot for monitoring logs, managing memory, and launching scripts.
I will ensure the security of command execution, set up backups, and provide you with a fully ready, tested system with step-by-step instructions.
… I would be happy to discuss the details, architecture of the solution, and timelines in a call or chat.

Show original

Translate

I worked on UVWeb (https://ou-uv.com) — a B2B system on Flask/Python with automated data flows, REST API integrations, and event-driven logic for CodeZero Group.

I read the entire project brief. I understand the scope: Windows VPS + local LLM (Ollama/Qwen/DeepSeek) + vector memory (Qdrant/ChromaDB) + Playwright for browser control + desktop automation (PyAutoGUI/Open Interpreter) + ingestion pipeline for PDF/video/YouTube + 24/7 scheduler with auto-restart. This is a complete MVP of an agency system, not a simple configuration.

What I will do:
- Setup Windows VPS + Remote Desktop + autostart services after restart
- Implement Ollama with the chosen model (Qwen2.5:7b or DeepSeek)
- LangGraph agent with long-term memory (Qdrant + RAG pipeline)
- Playwright — browser control, clicking, reading data, forms
… - PyAutoGUI + Open Interpreter — full Windows GUI automation
- Ingestion pipeline: PDF, TXT/Markdown, YouTube transcription (Whisper)
- Task scheduler + Telegram bot for agent management
- Workflow testing + user manual

--- OPTIONS ---

- Option A (MVP): 1000 PLN (7 days) — Ollama + LangGraph + ChromaDB + Playwright + PDF/TXT ingestion + autostart + manual
- Option B (Full): 1600 PLN (7 days) — Option A + Qdrant + PyAutoGUI + YouTube/Whisper + Telegram bot + backup — best scope/price ratio
- Option C (Enterprise): 2080 PLN (7 days) — Option B + FastAPI web dashboard + monitoring/alerts + architecture documentation + 30 days support

Completion time: 7 days. I need RDP access to the VPS (or provider info) and a list of initial tasks for the agent.

Portfolio:
- https://ou-uv.com — B2B Flask/Python system, automated data flows, REST API
- https://poseidon.codezerogroup.com — web app Python/React, enterprise integrations
- https://codezerogroup.com — B2B platform, custom CMS, API integrations

5 years in Python/automation — from simple scripts to RAG agency systems with LLM 24/7.

Ready to start upon confirmation — when do we begin?

Since I am new to the freelancehunt service and want to quickly gain a few initial projects for my portfolio, I am offering a 15% discount for the first 5 clients. The offer is valid until 5 orders are received.

Show original

Translate

Hello. I have experience deploying autonomous AI agents on Windows/VPS with Ollama, RAG (Qdrant/ChromaDB), LangGraph/CrewAI, browser automation (Playwright), and desktop automation (PyAutoGUI/Open Interpreter). We can deploy a system with 24/7 operation, memory layer, ingestion pipeline (PDF/YouTube/texts), management via web or Telegram, and auto-start on Windows Server. We are ready to discuss architecture, timelines, and budget after agreeing on the details of the technical specifications.

Show original

Translate

Good afternoon. I am ready to complete this project and have extensive experience in developing various applications.

Show original

Translate

Hello! Your project looks very promising. I am ready to start working and complete it at the highest level.

Show original

Translate

I will help today
I will help today
I will help today
I will help today
I will help today
I will help today

Show original

Translate

Ready to take on the project. The architecture and stack are clear: Ollama, LangGraph/CrewAI, RAG, vector DB, browser + desktop automation, ingestion pipeline, and autonomous workflows on Windows.

But to be honest right away: this is not a "simple agent setup," but a full-fledged MVP of an autonomous AI system with infrastructure, memory, automation, and 24/7 operation.

Regarding timelines:

A working MVP/prototype of this level — approximately 2-3 weeks (14-20 days)
A fully stable system with recovery, security, proper memory, ingestion pipeline, monitoring, and resilient workflows — about 1-2 months.

… Regarding budget:

A working MVP starts at around $3000.
A production-like system of this level — closer to $5000+.

It is optimal to do it in stages:

Basic agent + memory + browser
Windows automation
Ingestion/RAG pipeline
Dashboard / Telegram
Hardening, recovery, and stability 24/7.

Show original

Translate