Home/About
About RecallRun
A notebook for engineers who ship AI features to real users: practical questions, code that was run, numbers that were measured.
RecallRun is a notebook for engineers who ship AI features to real users. Every post answers one practical question (when do I need a vector database, why does my RAG miss exact error codes, how do I stop fake citations) with code that has been run and numbers that were measured, not guessed.
How posts are made
- Run it first. Benchmarks and code samples are executed before publishing. Each post says what it was tested on (hardware, library versions, dataset).
- Show the setup. Scripts are included so you can reproduce or disagree with the numbers.
- Corrections are public. The author is responsible for every claim. If something is wrong, tell me and I will fix it and note the correction.
Who writes it
Alby Tomy
Founder, RecallRun
Python/FastAPI Engineer Data Engineer AI Engineer
I build RAG and MCP systems in Python, and RecallRun is where I write up what I measure along the way. If a post has a number in it, I ran the code that produced it, and the script is in the post so you can check it.
Also built: gitcollect, an open-source CLI that groups GitHub and GitLab repos into collections with per-repo access control.
How the site makes money
Some posts link to ShipRAG, a starter kit I built and sell. The site may also show ads. Neither affects what the benchmarks say.