Concepts_25.06 - terrytaylorbonn/auxdrone GitHub Wiki
25.0608
3 main parts
(IN DIAGRAM BELOW IN RED)
- 1 Youtube demos. Hands-on demos (that cover the whole gamut) are the essential first step.
- 2 Youtube demo deployments. Choosing a few good demos from the previous goal to embellish and deploy.
- 3 Hackathons. Using ChatGPT and Copilot to build from scratch. Deep diving into various side topics (like partitions, WSL2, etc).
(IN DIAGRAM BELOW IN BLUE)
"BINARY" means that deterministic AI apps need a main binary algorithm that controls execution. Could be a chat UI (with human input steering the activity) or a python script.
- 4 APIs. Deep diving into the APIs of "agent frameworks" (HuggingFace, PydanticAI, OpenAI, and Mistral).
- 4.1 HF
- 4.2 PydanticAI
- 4.3 OpenAI
- 4.4 Mistral (?)
- 4b Agents 25.0608
(IN DIAGRAM BELOW IN GREEN)
"AI" means
- probablistic token generators based on trained models that provide the AI functionality.
- the "binary wrapper" (my term) for the binary logic that surrounds the token generator that
- controls the chaos of the token generator
- provides API interface (standard I/Fs like OpenAI) that can be called by agents
See
- 4c LLMs. describes
- host LLM (lcoal, HF, etc)
- fine-tune/train LLM
- build LLM