Context. Fundamental atomic transition parameters, such as oscillator strengths and rest wavelengths, play a key role in modelling and understanding the chemical composition of stars in the universe. Despite the significant work under way to produce these parameters for many astrophysically important ions, uncertainties in these parameters remain large and can limit the accuracy of chemical abundance determinations. Aims. The Belgian repository of fundamental atomic data and stellar spectra (BRASS) aims to provide a large systematic and homogeneous quality assessment of the atomic data available for quantitative spectroscopy. BRASS shall compare synthetic spectra against extremely high-quality observed spectra, at a resolution of ∼85 000 and signal-noise ratios of ∼1000, for approximately 20 bright BAFGK spectral-type stars, in order to critically evaluate the atomic data available for over a thousand potentially useful spectral lines. Methods. A large-scale homogeneous selection of atomic lines is performed by synthesising theoretical spectra of literature atomic lines for FGK-type stars including the Sun, resulting in a selection of 1091 theoretically deep and unblended lines in the wavelength range 4200-6800 Å, which may be suitable for quality assessment. Astrophysical log(g f) values are determined for the 1091 transitions using two commonly employed methods. The agreement of these log(g f) values are used to select well-behaved lines for quality assessment. Results. We found 845 atomic lines to be suitable for quality assessment, of which 408 were found to be robust against systematic differences between analysis methods. Around 53% of the quality-assessed lines were found to have at least one literature log(g f) value in agreement with our derived values, though the remaining values can disagree by as much as 0.5 dex. Only ∼38% of Fe※ I lines were found to have sufficiently accurate log(g f) values, increasing to ∼70-75% for the remaining Fe-group lines. ; SCOPUS: ar.j ; info:eu-repo/semantics/published