Alonso Silva
@alonsosilva
Followers
4K
Following
134K
Media
5K
Statuses
25K
AI Researcher @ Bell Labs | PhD in Physics | LLMs, LLMs, LLMs
Joined July 2009
✅ How to get a 100% Guaranteed Valid Tool Call from an LLM. It's possible, reliable, and surprisingly simple using `litelines`.
Want to visualize the response format constraints on the LLM when working in @ProjectJupyter ? Then you might be interested in my new project `litelines`.
0
4
12
Pequeño pero sustancial post acerca para entender la memoria en los agetnes, por @alonsosilva: https://t.co/v2Z0h8IIP9
0
1
2
huggingface.co
Want to visualize the response format constraints on the LLM when working in @ProjectJupyter ? Then you might be interested in my new project `litelines`.
0
1
3
Jeremy @jtpio and Nicolas @NicolasBrichet_ from @QuantStack presenting JupyterLite AI at @PyDataParis #PyDataParis
0
1
4
Featured in @marimo_io newsletter 🤩
Want to visualize the response format constraints on the LLM when working in @ProjectJupyter ? Then you might be interested in my new project `litelines`.
2
2
12
and don’t get me started on tool_choice auto! literally no support (maybe just for harmony), and it’s the first thing every single agentic library requires!
Tool calling on Open Models is quite a mess, there is no standard, everyone doing what they want. 😖
2
5
13
Doing a slightly better documentation than just the https://t.co/R02eKXHOXC (link to the documentation below). Feeback?
Want to visualize the response format constraints on the LLM when working in @ProjectJupyter ? Then you might be interested in my new project `litelines`.
1
3
8
😍
0
4
13
Here is a @marimo_io molab notebook to get you started with structured generation (link below)
Want to visualize the response format constraints on the LLM when working in @ProjectJupyter ? Then you might be interested in my new project `litelines`.
1
12
20
One of my favorite presentations at @aiDotEngineer was "Inside FLUX: how it really works" by @andi_blatt from @bfl_ml citing @arxiv papers and explaining what’s going on inside FLUX Kontext. #AIEParis
0
0
5
0
0
1
VB @reach_vb from @huggingface presenting the State of Open Weights LLMs - 2025 at @aiDotEngineer #AIEParis
2
5
56
1
5
23