https://datatweets.com/courses/generative-ai/working-with-llms-in-python/Recent content in Working with LLMs in Python on DATATWEETSHugoenCopyright (c) 2025 DatatweetsSat, 27 Jun 2026 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-1-how-llms-work/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-1-how-llms-work/Before writing any code, build a correct mental model of an LLM: it predicts the next token from a probability distribution, works in tokens rather than words, sees a limited context window, and samples its output.https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-2-your-first-claude-call/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-2-your-first-claude-call/Get from zero to a working API call: install the SDK, keep your key out of your code with an environment variable, call the Messages API with claude-haiku-4-5, and learn to read the response object — id, content, stop_reason, and usage.https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-3-system-prompts-and-roles/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-3-system-prompts-and-roles/Messages to Claude are a list of role-tagged turns, and a separate system prompt sets the model’s persona and rules. See the same question answered two ways, then learn to shape behavior on purpose.https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-4-multi-turn-conversations/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-4-multi-turn-conversations/The model remembers nothing between API calls. To build a chatbot that recalls earlier turns, you keep a messages list that grows — appending each user turn and the model’s reply — and resend it every request.https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-5-tokens-cost-and-streaming/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-5-tokens-cost-and-streaming/Every Claude response reports how many tokens it used. Learn to read those numbers, turn them into a dollar cost, estimate a request’s size before you send it, and stream long answers so they print progressively.https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-6-controlling-the-output/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-6-controlling-the-output/Four practical controls let you shape what comes back from the API: cap the length with max_tokens, end generation early with stop_sequences, dial randomness with temperature, and choose the model that fits your budget and difficulty.https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-7-guided-project-command-line-assistant/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/working-with-llms-in-python/lesson-7-guided-project-command-line-assistant/The module capstone: assemble a real command-line LLM app step by step — a system persona, a remembered conversation, streaming replies, per-turn token and cost tracking, and a clean REPL loop with /exit and /reset.