Great work. Will try it out sometime ahead. My notes from few months back: """ Pharo can offer new possibilities . LLM : processing of intent into code sent to pharo agents. All of non deterministic/ deterministic logic that is transformed to code to operate on input data Pharo agents: live environment that compiles in fraction of seconds.. perhaps milliseconds. Runs instant unit tests. Distribute verified runtime patch to other agents .. rather than messy library for everything like java/ python scaffold everything from its memory Object datastore: zero impedance non relational exactly mapped object data store with all of ACID properties if needed """ On Fri, 26 Jun, 2026, 03:20 omar abedelkader, <omar.abedelkader@inria.fr> wrote:

Hello everyone, for the first time, we have the first prototype of an inference engine without any connection to any of the external provider like Ollama, LMStudio, or vLLM.

[image: pharo-infer.png]

pharo-llm/pharo-infer: pharo-infer is a inference engine inside Pharo <https://github.com/pharo-llm/pharo-infer> github.com <https://github.com/pharo-llm/pharo-infer> <https://github.com/pharo-llm/pharo-infer>

[image: Capture d’écran 2026-06-25 à 23.40.41.png] — OA