diff --git a/lapis2-docs/src/components/TsvExample.astro b/lapis2-docs/src/components/TsvExample.astro
new file mode 100644
index 00000000..4e3d497a
--- /dev/null
+++ b/lapis2-docs/src/components/TsvExample.astro
@@ -0,0 +1,8 @@
+---
+
+---
+
+{/*prettier-ignore -- prettier will remove the tabs*/}
+primaryKey pango_lineage region age qc_value insertions aaInsertions
+sequence001 B.1.1.7 Europe 46 0.98 segment1:123:AAA,segment2:456:GTT gene1:123:EPE
+sequence002 B.1.1.7 46 0.98 gene2:123:EPE,gene2:125:EPE
diff --git a/lapis2-docs/src/content/docs/maintainer-docs/references/preprocessing.mdx b/lapis2-docs/src/content/docs/maintainer-docs/references/preprocessing.mdx
index 0ebc0e21..1925a9e6 100644
--- a/lapis2-docs/src/content/docs/maintainer-docs/references/preprocessing.mdx
+++ b/lapis2-docs/src/content/docs/maintainer-docs/references/preprocessing.mdx
@@ -3,9 +3,298 @@ title: Preprocessing
description: Reference on the SILO preprocessing
---
-TODO #565
+import TsvExample from '../../../../components/TsvExample.astro';
-- preprocessing config. Has defaults set in Docker. Only set what you need, leave the rest to the defaults
-- input format of data
+:::tip[Why preprocessing?]
+SILO contains an in-memory database.
+Building this database from the raw input data is computation intensive,
+thus this is done before starting SILO.
+This is called "preprocessing".
+The result is a serialized version of the database that can be loaded into SILO in a much shorter time.
+:::
-## Preprocessing config
+The SILO preprocessing accepts input data in two formats:
+
+- `NDJSON`: a single [NDJSON](https://ndjson.org/) file containing all the data,
+- `TSV/FASTA`: a directory containing
+- a TSV file with the metadata
+- FASTA files with the sequences
+
+The preprocessing configuration file determines which format should be used.
+
+## Preprocessing Configuration
+
+The preprocessing configuration file is a YAML file that allows the keys shown in the table below.
+All keys are optional and have default values.
+Some keys are relevant only for one of the two input file formats.
+
+:::tip
+When using the Docker image, you can adhere to the defaults and mount the files to the correct locations.
+You only need to specify `ndjsonInputFilename` or `pangoLineageDefinitionFilename`
+if you wish to use the corresponding features.
+:::
+
+| Key | Input Format | Default | Default in Docker Image |
+| -------------------------------- | ------------ | -------------------------------- | ------------------------ |
+| `inputDirectory` | both | `./` (current working directory) | `/preprocessing/input/` |
+| `outputDirectory` | both | `./output/` | `/preprocessing/output/` |
+| `intermediateResultsDirectory` | both | `./temp/` | `/preprocessing/temp/` |
+| `preprocessingDatabaseLocation` | both | (absent) | |
+| `ndjsonInputFilename` | `NDJSON` | (absent) | |
+| `metadataFilename` | `TSV/FASTA` | `metadata.tsv` | |
+| `pangoLineageDefinitionFilename` | both | (absent) | |
+| `referenceGenomeFilename` | both | `reference_genomes.json` | |
+| `nucleotideSequencePrefix` | `TSV/FASTA` | `nuc_` | |
+| `genePrefix` | `TSV/FASTA` | `gene_` | |
+
+:::note
+All filenames are relative to the `inputDirectory`.
+:::
+
+:::caution
+`ndjsonInputFilename` and `metadataFilename` must not be specified simultaneously as they determine the format.
+:::
+
+### Description of Keys for Both Formats
+
+- `inputDirectory`:
+ The directory where input files are located.
+- `outputDirectory`:
+ The directory where output files will be placed.
+- `intermediateResultsDirectory`:
+ The directory for storing intermediate results not relevant to the end user, mainly for debugging.
+- `preprocessingDatabaseLocation`:
+ The file for storing internal, intermediate database states for debugging.
+- `pangoLineageDefinitionFilename`:
+ The file with Pango lineage definitions, relative to the inputDirectory.
+ See the section on the [Pango Lineage Definition File below](#the-pango-lineage-definition-file) for details.
+- `referenceGenomeFilename`:
+ The file with [reference genomes](/maintainer-docs/references/reference-genomes), relative to the inputDirectory.
+
+## `NDJSON` Format
+
+SILO will initiate preprocessing in the `NDJSON` format
+if `ndjsonInputFilename` is specified in the preprocessing configuration.
+
+Each line in the NDJSON file must be a JSON object with the following keys:
+
+| Key | Type | Description |
+| ---------------------------- | -------- | ---------------------------------------------------------------------------- |
+| metadata | `object` | An object containing all metadata as key-value pairs. |
+| unalignedNucleotideSequences | `object` | A [sequences object](#sequences-object) with unaligned nucleotide sequences. |
+| alignedNucleotideSequences | `object` | A [sequences object](#sequences-object) with aligned nucleotide sequences. |
+| alignedAminoAcidSequences | `object` | A [sequences object](#sequences-object) with aligned amino acid sequences. |
+| aminoAcidInsertions | `object` | An [insertions object](#insertions-object) with amino acid insertions. |
+| nucleotideInsertions | `object` | An [insertions object](#insertions-object) with nucleotide insertions. |
+
+:::note
+You must configure two metadata columns for insertions in the
+[database configuration](/maintainer-docs/references/database-configuration)
+with the exact names and types as in this snippet:
+
+```yaml
+schema:
+ metadata:
+ - name: nucleotideInsertions
+ type: insertion
+ - name: aminoAcidInsertions
+ type: aaInsertion
+```
+
+Otherwise, SILO will not recognize insertions in the NDJSON format.
+:::
+
+#### Sequences Object
+
+The sequences object contains sequences for each segment or gene.
+It must include all `nucleotideSequences` (or `genes`, respectively) specified in the
+[reference genomes](/maintainer-docs/references/reference-genomes)
+as keys.
+Its values are the sequences as strings of
+[valid symbols](/references/nucleotide-and-amino-acid-symbols)
+or `null`.
+
+#### Insertions Object
+
+The insertions object contains a list of insertions for each segment or gene.
+It must include all `nucleotideSequences` (or `genes`, respectively) specified in the
+[reference genomes](/maintainer-docs/references/reference-genomes)
+as keys.
+Its values are arrays of strings in the format `:`.
+The insertions must consist of [valid symbols](/references/nucleotide-and-amino-acid-symbols).
+
+#### Example of the Schema
+
+```json
+{
+ "metadata": {
+ "primaryKey": "sequence001",
+ "pango_lineage": "B.1.1.7",
+ "region": null,
+ "age": 46,
+ "qc_value": 0.98
+ },
+ "unalignedNucleotideSequences": {
+ "segment1": "CGATA",
+ "segment2": "ACG"
+ },
+ "alignedNucleotideSequences": {
+ "segment1": "CGATAAT",
+ "segment2": "ACGT"
+ },
+ "alignedAminoAcidSequences": {
+ "gene1": "MYSLV*",
+ "gene2": "MADVQ*",
+ "gene3": "MSLYVQ*"
+ },
+ "nucleotideInsertions": {
+ "segment1": ["3:G", "4:A"],
+ "segment2": ["2:GTT"]
+ },
+ "aminoAcidInsertions": {
+ "gene1": ["3:EPE", "4:Q"],
+ "gene2": [],
+ "gene3": []
+ }
+}
+```
+
+:::caution
+For better readability, the example is displayed on multiple lines.
+A real NDJSON file must not contain line breaks and should look as follows:
+
+```
+{"metadata": {"primaryKey": "sequence001", /*...*/ }, "aminoAcidInsertions": /*...*/ }
+{"metadata": {"primaryKey": "sequence002", /*...*/ }, "aminoAcidInsertions": /*...*/ }
+```
+
+:::
+
+## `TSV/FASTA` Format
+
+SILO will initiate preprocessing in the `TSV/FASTA` format
+if `metadataFilename` is specified in the preprocessing configuration.
+
+SILO expects the following files in the `inputDirectory`:
+
+- a TSV file with the metadata named as configured in `metadataFilename`,
+- FASTA files with the sequences.
+
+### Metadata File
+
+The metadata file must be a TSV (tab-separated values) file.
+Its columns must correspond to the metadata fields specified in the [database configuration](/maintainer-docs/references/database-configuration).
+Empty values will be interpreted as `null`.
+
+#### Example
+
+Given the following database configuration:
+
+```yaml
+schema:
+ metadata:
+ - name: primaryKey
+ type: string
+ - name: pango_lineage
+ type: pango_lineage
+ - name: region
+ type: string
+ - name: age
+ type: int
+ - name: qc_value
+ type: float
+ - name: insertions
+ type: insertion
+ - name: aaInsertions
+ type: aaInsertion
+ # other configuration keys ...
+```
+
+The metadata file might look as follows:
+
+
+
+### Sequence Files
+
+In the `TSV/FASTA` format, sequences must be stored in separate FASTA files.
+The filenames must follow this pattern:
+
+- aligned nucleotide sequences: `nuc_.fasta`.
+ The `nuc_` prefix is configurable in the preprocessing configuration via `nucleotideSequencePrefix`.
+- unaligned nucleotide sequences:
+ TODO (https://github.com/GenSpectrum/LAPIS/issues/581) when https://github.com/GenSpectrum/LAPIS-SILO/issues/131 is resolved.
+- aligned amino acid sequences: `gene_.fasta`.
+ The `gene_` prefix is configurable in the preprocessing configuration via `genePrefix`.
+
+There must be one corresponding file for every segment and gene defined in the
+[reference genomes](/maintainer-docs/references/reference-genomes).
+
+The header in the FASTA files must match the `primaryKey` column in the metadata file.
+There must be a one-to-one correspondence between entries in the metadata file and sequences in the FASTA files.
+
+#### Example
+
+Given the reference genomes:
+
+```json
+{
+ "segments": [
+ { "name": "segment1", "sequence": "/*...*/" },
+ { "name": "segment2", "sequence": "/*...*/" }
+ ],
+ "genes": [
+ { "name": "gene1", "sequence": "/*...*/" },
+ { "name": "gene2", "sequence": "/*...*/" },
+ { "name": "gene3", "sequence": "/*...*/" }
+ ]
+}
+```
+
+the input directory should contain the following files:
+
+```
+input/
+├── gene_gene1.fasta
+├── gene_gene2.fasta
+├── gene_gene3.fasta
+├── nuc_segment1.fasta
+├── nuc_segment2.fasta
+└── /* other files... */
+```
+
+The file `nuc_segment1.fasta` might look as follows—
+assuming that the metadata file also has two entries with the primary keys `sequence001` and `sequence002`:
+
+```
+>sequence001
+CGATAAT
+>sequence002
+CGATAAT
+```
+
+## The Pango Lineage Definition File
+
+This file is relevant only if your data includes Pango Lineages.
+
+The Pango lineage definition file is a JSON file mapping Pango lineage names to their aliases.
+It is used to reconstruct the lineage tree structure.
+SILO requires this to properly group sequences into partitions to fully benefit from partitioning.
+
+The file contains a JSON object with alias names as keys and:
+
+- an empty string if the alias is a root node,
+- the name of the parent node if the alias is a child node,
+- an array of parent nodes if the alias is a recombinant.
+
+Here is a minimal example:
+
+```json
+{
+ "A": "",
+ "B": "A.1.1.1",
+ "XA": ["B.1.2", "B.1.42"]
+}
+```
+
+A complete example can be found here:
+https://github.com/cov-lineages/pango-designation/blob/master/pango_designation/alias_key.json
diff --git a/lapis2-docs/src/content/docs/maintainer-docs/references/starting-silo-and-lapis.mdx b/lapis2-docs/src/content/docs/maintainer-docs/references/starting-silo-and-lapis.mdx
index a3c49a26..0f13f1ba 100644
--- a/lapis2-docs/src/content/docs/maintainer-docs/references/starting-silo-and-lapis.mdx
+++ b/lapis2-docs/src/content/docs/maintainer-docs/references/starting-silo-and-lapis.mdx
@@ -35,6 +35,8 @@ Optionally, you can specify:
e.g. `~/LAPIS/database_config.yaml`.
If those parameters are not provided, SILO will try to read the files from its current working directory.
+See the [preprocessing reference](/maintainer-docs/references/preprocessing#preprocessing-config)
+for how to provide the input files.
#### Starting the SILO preprocessing with Docker
diff --git a/lapis2-docs/src/content/docs/maintainer-docs/tutorials/start-lapis-and-silo.mdx b/lapis2-docs/src/content/docs/maintainer-docs/tutorials/start-lapis-and-silo.mdx
index a5e69d4d..68e52e07 100644
--- a/lapis2-docs/src/content/docs/maintainer-docs/tutorials/start-lapis-and-silo.mdx
+++ b/lapis2-docs/src/content/docs/maintainer-docs/tutorials/start-lapis-and-silo.mdx
@@ -72,11 +72,10 @@ See the [config reference](/maintainer-docs/references/database-configuration) f
### Starting SILO Preprocessing
-SILO contains an in-memory database.
-Building this database from the raw input data is computation intensive,
-thus this is done before starting SILO.
-This is called "preprocessing".
-The result is a serialized version of the database that can be loaded into SILO in a much shorter time.
+:::note
+Refer to [Preprocessing](/maintainer-docs/references/preprocessing) for more explanation
+and a full reference of the preprocessing.
+:::
Download the example dataset from the [end-to-end tests](https://github.com/GenSpectrum/LAPIS/tree/main/siloLapisTests/testData):