Articles
Articles about Doka and AI agents
How Doka works, why a desktop agent matters, and what is happening with AI. page 2.
AnythingLLM Desktop: What It Does and Where Chatting Over Documents Stops
A breakdown of AnythingLLM's desktop version: workspaces and RAG, agentic skills and MCP, what Desktop lacks compared to the server version, and what privacy actually depends on.
Jan AI: A Local ChatGPT Replacement, and Where It Hits a Ceiling
What Jan can do: local models from Hugging Face, a server on localhost:1337, connecting cloud providers, and MCP. How the app differs from an agent, and where the chat format runs out.
llama.cpp on CPU: Running a Local Model Without a Graphics Card
How to run llama.cpp on CPU only, which model actually works without a GPU, why everything comes down to memory bandwidth, and what speeds to expect.
Embeddings in llama.cpp: Local Search Over Your Own Documents
How to start llama-server in --embeddings mode, why you need a separate model for embeddings, what --reranking does, and why RAG ends up needing two servers.
MCP in llama.cpp: Tools for a Local GGUF Model
How to connect MCP servers in llama-server through --mcp-servers-config, why --jinja is required, why the CORS proxy is risky to enable, and what stays on your computer in this setup.
llama.cpp vs Ollama vs LM Studio: Engine vs Wrapper Explained
Why the 'llama.cpp or Ollama' debate rests on a false premise, how Llama differs from llama.cpp, and how to choose between them for your own home setup.
llama-cpp-python: A Local Model From Python in Ten Minutes
How to install llama-cpp-python, enable a GPU build, load a GGUF model straight from Hugging Face, and start an OpenAI-compatible server with one command.
Running llama.cpp on Windows Without Building From Source
How to run llama.cpp on Windows using prebuilt binaries: which build to pick for NVIDIA, AMD, or CPU-only, how llama-server starts up, and why you shouldn't change --host to 0.0.0.0.
llama-server: A Local OpenAI-Compatible API on Your Own Computer
How to start llama-server, which endpoints it exposes, why you need --api-key and --parallel, and why changing --host to 0.0.0.0 opens your model up to the whole network.
MLX or GGUF on Mac: Which Format to Run a Local Model In
How the MLX format differs from GGUF on Apple Silicon, why mlx-lm won't open a .gguf, what to pick for your MacBook, and how unified memory affects model size.
AI That Controls Your Computer: How It Actually Works
How computer use differs from an agent with access to files and the terminal, why controlling a mouse needs the cloud and screenshots, and which option already works locally today.
Skills, Tools, and MCP in Open WebUI: What's What
How Open WebUI skills differ from tools and functions, how SKILL.md is structured, how MCP servers connect, and where the line of locality actually falls in this setup.