Run Ollama cloud models from a small VPS
Ollama on a small no-GPU VPS with cloud models doing the inference: signin without a browser, a private API on 11434, quotas, and a local fallback.
#llm
Filtering by topic #llm · clear
How big an LLM fits in your server RAM
Size an LLM for a CPU server without guessing: weights in bytes, KV cache growth with context, runtime overhead, and why fitting is not the same as usable.
Agent harness token overhead: where it goes
Two harnesses on one model can bill an order of magnitude apart. The mechanics behind the gap, and how to measure your own agent's token spend.
Run a Jev-style decision model on a VPS
Jev has no public weights. Build the same typed decisions on your own server with llama.cpp grammars or GLiNER2.5-Decide, and check the confidence.
Build your own AI agent on a VPS
An AI agent is a loop around a language model that can use tools. Learn the concepts behind building one on your own VPS: the loop, tools, MCP, memory.
Host Ollama on a VPS: self-host an LLM safely
A 7B model wants 8 GB of RAM and gives 4 to 10 tokens per second on CPU. Host one on a VPS and call it at 127.0.0.1:11434/v1, with port 11434 kept closed.
Can you self-host Claude? The honest answer
Claude's weights are not public, so no server of yours can run it. Here is what you can self-host instead: open models, a gateway, and Claude Code.
Build your own OpenClaw-style AI agent
OpenClaw is a personal AI agent that runs commands on your own server. Here is how one is built, piece by piece, and why hardening comes first.
Ollama Cloud vs your own server: what changes
Ollama Cloud and a self-hosted Ollama share one CLI and one API. See exactly which settings change, and what leaves your machine on each path.
Self-host Langfuse to trace your AI agents
Run Langfuse on your own VPS: the real resource floor, pinned image tags, TLS, ClickHouse retention before it fills the disk, and backups that work.
Ollama vs vLLM: which LLM server to run
Ollama is a convenience layer for one user, on CPU if needed. vLLM is a throughput engine for GPUs. Pick by workload, with the real commands for both.
Build an AI agent with Claude on a VPS
Use Claude as the brain and your VPS as the body. How the Messages API, tool use, and MCP fit together to build an AI agent you run yourself.
Which AI models can you self-host?
Pick a model by the RAM you actually have. Sizing arithmetic for 4 GB, 16 GB and 64 GB VPS plans, honest CPU token rates, and the hidden cost of context.
Keep an Ollama model loaded in memory
Ollama unloads a model after 5 minutes idle, so the next request waits for a reload. Set keep_alive in a systemd override so it survives reboots, then preload.
Ollama context length: set num_ctx and size RAM
Ollama truncates long prompts at a small default context. Set num_ctx per request or OLLAMA_CONTEXT_LENGTH server-wide; size KV cache RAM before raising it.
Ollama: fix context deadline exceeded
Ollama context deadline exceeded means a timeout fired before the model answered. Find which layer set it: your client, the model load, or nginx.
KV cache vs prompt cache explained
The KV cache is RAM your server pays for on every request. Prompt caching is a bill someone else discounts. Only one can break a model load.
SearXNG JSON API for Open WebUI search
Your SearXNG instance can feed web search to Open WebUI. Enable the JSON format, then hand the other container a URL it can actually reach.
AI Agent vs LLM vs AI Assistant
An LLM is weights that need RAM. An assistant adds a chat surface. An agent adds tools and a loop, and it holds credentials, so it needs a box that stays on.