Search for a command to run...
Every module up to this point used the same streaming events without ever reading two fields that were there the entire time. message_start and message_delta carry token usage on every single stream —
Everything in the earlier modules assumed the happy path: the API responds, the model generates valid output, tools return results, the stream closes cleanly. In production, none of that is guaranteed
Phase 4 gave the study system memory of individual problems — when each one is due for review based on SM-2. Phase 5 asks a different question: zoomed out across all topics, what should I actually be
04-structured-output solved the reliability problem: use forced tool use to guarantee valid JSON back from the model. But it waited for the complete response before doing anything with it. For a backe
The first time I ran evals on my LLM playground, the agent scored lower than basic RAG — 7.5 vs 8.9. The agent is more capable: it has live tool data on top of retrieval, it can answer questions basic
The first three phases of my study system handled the session loop: generate a problem, grade the solution, log the result. Useful, but with a fundamental gap — there was no memory of what I struggled
In the basic RAG module, retrieval was unconditional. Every query triggered the same pipeline: embed the query, search the vector store, stuff the top chunks into the prompt, generate. The model had n
Picking up Python as a JavaScript engineer is mostly straightforward because the mental models transfer well. Functions, modules, async patterns, data structures — the concepts are the same. What trip
Prompt engineering only becomes interesting when you're prompting for structured, specific output that has to be reliable enough to feed downstream systems. A chatbot that gives a slightly different a
