Based on @dsull's answer.

The smart-seq3 analysis is handled by the kallisto-bustools pipeline.

Create a conda environment if needed. Create the conda environment file and save it as env_kb.yml.

# conda environment for kallisto bustools pipeline

name: kb

channels:
 - bioconda
 - conda-forge

dependencies:
 - python=3.8
 - kallisto=0.48.0
 - bustools=0.42.0
 - pip
 - pip:
    - kb-python==0.27.3

Create the conda environment and activate it.

mamba env create -f env_kb.yml
conda activate kb

Kallisto bustools tools should be now available.

Create index:

kb ref \
  -i index.idx \
  -g t2g.txt \
  -f1 transcriptome.fa \
  /path/genome.fa \
  /path/genes.gtf \
  --kallisto kallisto \
  --bustools bustools

Quantify:

kb count \
  -i index.idx \
  -g t2g.txt \
  -t 20 \
  -m 100G \
  -o out/ \
  -x SMARTSEQ3 \
  --gene-names \
  --h5ad \
  --kallisto kallisto \
  --bustools bustools \
  -w whitelisted_barcodes.txt \
  /path/K45_9001_S1_I1_001.fastq.gz \
  /path/K45_9001_S1_I2_001.fastq.gz \
  /path/K45_9001_S1_R1_001.fastq.gz \
  /path/K45_9001_S1_R2_001.fastq.gz

The four fastq files must be specified in that same order: I1, I2, R1 and R2.