Angelo D'Ambrosio bakaburg1

Public Health Doctor, ECDC Expert, Infectious Disease Modeling PhD student.

bakaburg1 / req_perform_parallel_promises.R

Created August 8, 2025 17:43

httr2 based function to perform parallel requests using promises

	#' Parallel LLM helpers (promises-based)
	#'
	#' Provides a reimplementation of the parallel chat stack using a
	#' promises/curl-multi engine with per-request retry, caching, and
	#' progress logging. Top-level functions mirror the behaviour of
	#' `ellmer::parallel_chat()` and `ellmer:::parallel_turns()`.

	#' Compute backoff wait time (rate limit headers first)
	#'
	#' @noRd

bakaburg1 / parallel_llm.R

Created August 8, 2025 17:41

ellmer based helpers to perform promises based parallel requests

	#' Helper: extract model-level error object from JSON (top-level or first
	#' choice)
	#' @noRd
	.json_extract_error <- function(json) {
	if (!is.list(json)) return(NULL)
	`%\|\|%` <- rlang::`%\|\|%`

	# Helper to extract error from first choice if available
	choices_first <- function(j) {
	ch <- tryCatch(j$choices, error = \(e) NULL)

bakaburg1 / figure_caption_filter.lua

Last active September 30, 2024 21:46

A filter to store image captions into the Meta fields

	function RawBlock (raw)
	local caption = raw.text:match("\\caption%{[^%}]+%}")
	if caption then
	captions:insert(caption)
	end
	end

	function Pandoc (doc)
	-- append collected captions at the end
	doc.meta.figures = captions -- This works but latex is escaped by \textbackslash

bakaburg1 / gist:817be0e276d94032f1b654afb34452bb

Last active December 27, 2021 19:26

Scrape unbound globals which creates warnings with R CMD CHECK

	library(stringr)
	library(dplyr)

	check_logfile <- ...

	readLines(check_logfile) %>% paste(collapse = '\n') %>%
	str_extract_all("[\\w_]+ ?:( <anonymous>:)?[\\s\\n]+no[\\s\\n]+visible[\\s\\n]+binding[\\s\\n]+for[\\s\\n]+global[\\s\\n]+variable[\\s\\n]+‘.*’") %>%
	lapply(function(x) {
	fun <- str_extract(x, '^[\\w_]+')

bakaburg1 / app.R

Created November 2, 2021 20:26

Child Julia process from shiny app do not survive parent process


	library(shiny)
	print(paste('Inline Julia pID:', sys::exec_background('nohup', c("julia", "-e sleep(3000)"))))
	print(paste('Julia script pID:', sys::exec_background('nohup', c("julia", "test.jl", "&"))))

	# Define UI for application that draws a histogram
	ui <- fluidPage(

	# Application title
	titlePanel("Old Faithful Geyser Data"),

bakaburg1 / ctree_rules.R

Last active October 1, 2020 23:18

Small helper functions to interact with partykit::ctree().

	#' Extract ctree rules.
	#'
	#' Extract the tree rules, optionally formatting them in order to be ready to
	#' use for data filtering.
	#'
	#' @param tree An object produced by \code{partikit::ctree()}.
	#' @param rule.as.text Whether to collapse rule conditions into a string or
	#' leave them as a character vector.
	#' @param eval.ready Whether to format rules in order to be easily eval() for
	#' data filtering. Turns \code{rule.as.text} automatically on.

bakaburg1 / gist:b0a0a354930faa719767163618d321fc

Created July 18, 2019 20:38

	brms.mod.nprest.nb <<- brm(bf(N.prest ~ Mesi + (Mesi\|Context), family = negbinomial()), data = TdA, prior = c(
	set_prior('normal(0, 2.5)', class = 'Intercept'),
	set_prior("cauchy(0, 2.5)", class = 'b')#,
	#set_prior("normal(0, 2.5)", class = 'sd')
	), init_r = 0.1, cores = 8, chains = 8, iter = 8000, warmup = 2000, control = list(max_treedepth = 15, adapt_delta = .90))

bakaburg1 / gist:e89bbc366bf998670c8c93217e175ada

Created July 18, 2019 17:36

Error producing brm code

	brms.mod.attesa.prior <- brm(bf(Attesa ~ log(N.prest) + (log(N.prest)\|Context), shape ~ log(N.prest) + (log(N.prest)\|Context), family = Gamma('log')), data = TdA, prior = c(
	set_prior('normal(0, 2.5)', class = 'Intercept'),
	set_prior("cauchy(0, 2.5)", class = 'b'),
	set_prior("normal(0, 1.5)", class = 'sd', group = 'Context'),
	set_prior("normal(0, 1.5)", class = 'sd'),
	set_prior("student_t(3, 0, 2.5)", class = 'sd', dpar = 'shape'),
	set_prior("student_t(3, 0, 2.5)", class = 'sd', group = 'Context', dpar = 'shape')
	), init_r = 0.1, cores = 8, chains = 8, iter = 8000, warmup = 2000, control = list(max_treedepth = 15, adapt_delta = .90))

	Error in stan_base_prior(prior) : length(base_prior) == 1L is not TRUE

bakaburg1 / mcdo.R

Last active March 17, 2019 12:24

Very simple parallelized implementation of dplyr::do() using mclapply.

bakaburg1 / gist:6345023

Last active December 21, 2015 18:08

Horizontal rule with text. #css #webdesign