alecor.net.
  • AI
  • Home
  • About
  • Contact
  • Projects
  • Tags
  • Categories
  • Archives
🇬🇧 🇦🇷 🇧🇷 🇵🇱 🇷🇺
  • 2025-11-12

    LLM Observability Cheatsheet: Tracing, Tokens, Cost and OpenTelemetry GenAI

    A practical cheatsheet for observing LLM apps and agents in production: the OpenTelemetry GenAI semantic conventions (gen_ai.* attributes, span names and kinds, metrics, and their development...

    AI · cheatsheets AI LLMs agents observability OpenTelemetry tracing MLOps English
  • 2025-09-17

    Prompt Injection and Agent Security Cheatsheet

    A practical cheatsheet on prompt injection and agent security: direct vs. indirect injection, Simon Willison's lethal trifecta, OWASP Top 10 for LLM Applications 2025 (LLM01 and friends), and the...

    AI · cheatsheets AI LLMs agents security prompt injection OWASP English
  • 2025-08-20

    Running Open-Weight LLMs Locally Cheatsheet: Quantization, Ollama, llama.cpp, vLLM

    A practical cheatsheet for running open-weight LLMs on your own hardware: quantization formats (GGUF k-quants such as Q4_K_M, GPTQ, AWQ, MXFP4), VRAM math for weights and KV cache with a Python...

    AI · cheatsheets AI LLMs open-weight models local LLMs quantization Ollama llama.cpp vLLM GPU English
  • 2025-06-30

    Context Engineering Cheatsheet

    A practical cheatsheet on context engineering for LLM apps and agents: what goes into the context window, the write/select/compress/isolate strategies, how long contexts fail (poisoning,...

    AI · cheatsheets AI LLMs agents context engineering prompt engineering prompt caching English
  • 2025-05-14

    RAG Cheatsheet: Chunking, Hybrid Search, Reranking

    A practical cheatsheet for the retrieval half of RAG: chunking strategies and sizes, metadata, choosing embeddings, hybrid search with BM25 and reciprocal rank fusion, cross-encoder reranking,...

    AI · cheatsheets AI LLMs RAG retrieval embeddings hybrid search reranking English
← Previous Page 2 of 3 Next →

Nothing you read here should be considered advice or recommendation. Everything is purely for informational purposes.

Atom Feed