All stories
250 stories found
Build Custom Ollama Models with a Modelfile
Stop pasting your system prompt into every single new chat. Learn how to write a custom Ollama Modelfile, bake in custom behaviors and parameters, and run perfectly tailored local AI models on your own hardware tonight.
Local LLMs on a Raspberry Pi 5: Honest Numbers & Setup
Wondering if a Raspberry Pi 5 can actually run a local LLM? I spent the weekend testing Llama 3, Phi-3, and Qwen on an 8GB Pi 5 to give you exact commands, thermal limits, and real tokens-per-second benchmarks.
Access Ollama From Another Computer on Your LAN (Safely)
If you are getting connection refused when trying to reach your Ollama server from another machine, the fix requires changing the binding address and opening port 11434. Here is how to do it safely.
Local Embedding Models Compared: Nomic, BGE, E5, mxbai
I spent a week testing Nomic, BGE, E5, and mxbai to fix garbage RAG retrieval. Here is exactly which local embedding model you should use and how to avoid dimension mismatch errors.
Fix LM Studio "Failed to load model" Error
If you are seeing the "Failed to load model" error in LM Studio, you likely have a VRAM shortage, a corrupted GGUF file, or a backend API mismatch. Here is how to fix it right now.
Q4 vs Q5 vs Q8 Quantization: Which GGUF to Actually Download
Staring at 15 different GGUF files on Hugging Face? Here is exactly how Q4, Q5, and Q8 compare in VRAM usage, generation speed, and actual response quality on local hardware.
Fix ComfyUI "Torch Not Compiled With CUDA Enabled" Error
If you are seeing "AssertionError: Torch not compiled with CUDA enabled" in ComfyUI, your Python environment pulled the CPU-only version of PyTorch. Here is exactly how to fix it.
Fix Stable Diffusion Out Of Memory on a 6GB VRAM GPU
Stop hitting CUDA OOM errors. Here are the exact command-line arguments, cross-attention settings, and VRAM limits to run Stable Diffusion smoothly on a 6GB graphics card.
Connect Obsidian to Ollama: 100% Local AI Notes
Learn how to connect Obsidian to Ollama for private, offline AI notes. Tested with Smart Connections and Text Generator on a local vault.
Fix Open WebUI Docker Connection Refused to Ollama
Stop Docker networking errors between Open WebUI and Ollama. Clear connection refused errors with exact IP, CORS, and container configs.
Ollama Answers Get Cut Off Mid-Sentence: num_ctx vs num_predict Explained
Truncated replies and models that "forget" the start of a long document come from two different settings. Here is how I tell them apart and the values I use.
Ollama Error "model requires more system memory": How I Fixed It in 10 Minutes
Ollama refuses to load a model and says it needs more system memory than you have. Here is exactly what the number means and the four fixes that worked on my 32 GB / RTX 3060 box.