LLMs

Within days of the release, three clients asked the same question: should we pause the project and rebuild around this? The answer is almost never. Here is why.

Token Guard is a GitHub Action that counts tokens in your agent context files and enforces limits in CI. Here’s why we check agent context into our repos, and why keeping it lean matters for team collaboration.

Tired of parsing flaky JSON from LLM responses? OpenAI’s Structured Outputs feature guarantees your responses match your schema exactly. Here’s how to use it with Pydantic, when to choose it over function calling, and the gotchas you’ll encounter in production.

RAG is the most common pattern for putting an LLM in front of your own data, and the most commonly misunderstood. Here is what it is, when it is the right tool, and how the pieces fit together.

Benchmark scores don’t tell you whether a model will work for your business. Here’s how to evaluate LLMs on the three axes that actually matter: quality, throughput, and cost.

The team that brought type safety to Python web development with Pydantic has just unveiled their take on AI development: Pydantic.ai. This new framework reimagines how we build AI applications by bringing Pydantic’s legendary validation capabilities to the world of Large Language Models.

Large Language Models like GPT-4 are revolutionizing AI, but their power demands rigorous assessment. How do we ensure these marvels perform as intended? Welcome to the crucial world of LLM evaluation.