AI tools and local models,
explained clearly.
Source-linked reporting, developer-tool guides, and practical analysis that separates documented facts from open questions.
Featured AI Companies
The labs, model providers and startups shaping the AI landscape.
Top Entities
People, products, companies and places driving today's AI coverage.
أحدث المقالات بالعربية
تغطية يومية للذكاء الاصطناعي والتقنية بلغتك.
أخطاء Ollama الشائعة وحلولها (من واقع 6 أشهر استخدام يومي)
تسعة أخطاء تكررت معي أكثر من غيرها، ومعها الحل الذي نجح فعلاً — من نقص الذاكرة إلى مخرجات JSON غير الصالحة.
كيف شغّلت نموذج ذكاء اصطناعي على لابتوب بدون كرت شاشة (تجربتي الكاملة 2026)
معالج i5 و16 جيجا رام وبدون كرت شاشة منفصل: 9 كلمات في الثانية على نموذج 7B. الإعدادات التي غيّرت النتيجة والأخطاء التي أهدرت وقتي.
كيف تبني وكيل ذكاء اصطناعي يعمل على جهازك بدون إنترنت (Ollama + MCP) — جرّبته 3 أسابيع (2026)
بنيت وكيل AI محلي يقرأ ملفاتي وينفّذ أوامر على جهازي بدون إرسال أي بيانات للسحابة. هذه الخطوات الحقيقية، الأخطاء التي واجهتني، والحل الدقيق لكل واحدة.
Latest Stories
Page 2 of 8 · 190 stories
Fixing Painfully Slow Whisper Transcription
Base Whisper is a reference implementation, not a production tool. If your transcriptions take as long as the audio itself, here is the exact stack I use to get 10x to 30x faster speeds locally.
Fix Open WebUI Showing No Models in the Dropdown
If Open WebUI is running but your model dropdown is empty, you likely have a Docker networking mismatch or an unpopulated Ollama instance. Here is the exact fix.
Fix Continue in VS Code Not Connecting to Ollama
If your Continue extension in VS Code is throwing ECONNREFUSED or failing to fetch from Ollama, the fix is usually a mismatch between Node.js IPv6 resolution and your config.json file.
Secure Ollama with Nginx, HTTPS, and a Password
Stop exposing your raw Ollama port to the internet. Here is exactly how I put Ollama behind Nginx with a real SSL certificate and password protection so I can access my local models safely.
How to Stop Ollama From Unloading Models (keep_alive)
Ollama unloads models from VRAM after 5 minutes of inactivity. Here is exactly how to use the keep_alive parameter to force your models to stay loaded so you never wait for a cold start again.
A One-File Docker Compose Stack for Ollama and Open WebUI
Run Ollama and Open WebUI together effortlessly. This single Docker Compose file handles networking, data persistence, and Nvidia GPU passthrough so you can start prompting immediately.
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.
DeepSeek-R1 Repeats Itself or Outputs Gibberish: The 4 Settings That Fixed It
Endless loops, repeated sentences or mojibake from DeepSeek-R1 are almost always sampling or quantization — not a broken model. The exact parameters I now use.
Ollama Filled My Disk: How I Moved the Models Directory Safely
Ollama stores every model on the system drive by default. The exact steps I used to move ~/.ollama/models to a second drive on Linux, Windows and macOS without re-downloading.
Ollama Not Using My NVIDIA GPU in WSL2: The Fix That Finally Worked
Ollama says "no compatible GPUs were discovered" inside WSL2 even though nvidia-smi works. Here is the exact driver, toolkit and verification sequence that fixed it on my RTX 3060.
Ollama Pull Fails with "max retries exceeded" or EOF: How I Get Downloads to Finish
Model downloads that stall at 97% or die with EOF. The five things that fixed it for me: resume behaviour, MTU, DNS, proxy variables and a corrupted blob cache.
Ollama "connection refused on 127.0.0.1:11434": The 5 Causes I Have Actually Hit
Your client cannot reach Ollama on port 11434. Here are the five real causes — service not running, wrong host binding, WSL networking, port conflict, and Docker — with the command that fixes each.
The Daily Pulse
Newsletter delivery is not connected yet. This form only saves your address in this browser; no email is sent.
Get concise, source-linked technology notes without the hype.