diff --git a/milli/src/index.rs b/milli/src/index.rs
index e2ab51a1c7..c8af5aba33 100644
--- a/milli/src/index.rs
+++ b/milli/src/index.rs
@@ -808,7 +808,7 @@ pub(crate) mod tests {
     use maplit::btreemap;
     use tempfile::TempDir;
 
-    use crate::update::{IndexDocuments, UpdateFormat};
+    use crate::update::{IndexDocuments};
     use crate::Index;
 
     pub(crate) struct TempIndex {
@@ -844,13 +844,12 @@ pub(crate) mod tests {
         let index = Index::new(options, &path).unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
+        let content = documents!([
             { "id": 1, "name": "kevin" },
             { "id": 2, "name": "bob", "age": 20 },
             { "id": 2, "name": "bob", "age": 20 }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -869,8 +868,12 @@ pub(crate) mod tests {
         // we add all the documents a second time. we are supposed to get the same
         // field_distribution in the end
         let mut wtxn = index.write_txn().unwrap();
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
+        let content = documents!([
+            { "id": 1, "name": "kevin" },
+            { "id": 2, "name": "bob", "age": 20 },
+            { "id": 2, "name": "bob", "age": 20 }
+        ]);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -887,13 +890,12 @@ pub(crate) mod tests {
         );
 
         // then we update a document by removing one field and another by adding one field
-        let content = &br#"[
+        let content = documents!([
             { "id": 1, "name": "kevin", "has_dog": true },
             { "id": 2, "name": "bob" }
-        ]"#[..];
+        ]);
         let mut wtxn = index.write_txn().unwrap();
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
diff --git a/milli/src/lib.rs b/milli/src/lib.rs
index 8381b799c3..99bc29fe49 100644
--- a/milli/src/lib.rs
+++ b/milli/src/lib.rs
@@ -1,6 +1,9 @@
 #[macro_use]
 extern crate pest_derive;
 
+#[macro_use]
+pub mod documents;
+
 mod criterion;
 mod error;
 mod external_documents_ids;
@@ -12,7 +15,6 @@ pub mod proximity;
 mod search;
 pub mod tree_level;
 pub mod update;
-pub mod documents;
 
 use std::borrow::Cow;
 use std::collections::{BTreeMap, HashMap};
diff --git a/milli/src/search/distinct/mod.rs b/milli/src/search/distinct/mod.rs
index e7dc52a82c..3f576bbc5b 100644
--- a/milli/src/search/distinct/mod.rs
+++ b/milli/src/search/distinct/mod.rs
@@ -27,21 +27,24 @@ pub trait Distinct {
 #[cfg(test)]
 mod test {
     use std::collections::HashSet;
+    use std::io::Cursor;
 
+    use bimap::BiHashMap;
     use once_cell::sync::Lazy;
     use rand::seq::SliceRandom;
     use rand::Rng;
     use roaring::RoaringBitmap;
     use serde_json::{json, Value};
 
+    use crate::documents::{DocumentsBuilder, DocumentsReader};
     use crate::index::tests::TempIndex;
     use crate::index::Index;
-    use crate::update::{IndexDocumentsMethod, UpdateBuilder, UpdateFormat};
+    use crate::update::{IndexDocumentsMethod, UpdateBuilder};
     use crate::{DocumentId, FieldId, BEU32};
 
-    static JSON: Lazy<Value> = Lazy::new(generate_json);
+    static JSON: Lazy<Vec<u8>> = Lazy::new(generate_json);
 
-    fn generate_json() -> Value {
+    fn generate_json() -> Vec<u8> {
         let mut rng = rand::thread_rng();
         let num_docs = rng.gen_range(10..30);
 
@@ -69,7 +72,13 @@ mod test {
             documents.push(doc);
         }
 
-        Value::Array(documents)
+        let mut cursor = Cursor::new(Vec::new());
+        let mut builder = DocumentsBuilder::new(&mut cursor, BiHashMap::new()).unwrap();
+
+        builder.add_documents(documents).unwrap();
+        builder.finish().unwrap();
+
+        cursor.into_inner()
     }
 
     /// Returns a temporary index populated with random test documents, the FieldId for the
@@ -89,13 +98,14 @@ mod test {
         let mut addition = builder.index_documents(&mut txn, &index);
 
         addition.index_documents_method(IndexDocumentsMethod::ReplaceDocuments);
-        addition.update_format(UpdateFormat::Json);
-        addition.execute(JSON.to_string().as_bytes(), |_, _| ()).unwrap();
+        let reader = crate::documents::DocumentsReader::from_reader(Cursor::new(&*JSON)).unwrap();
+        addition.execute(reader, |_, _| ()).unwrap();
 
         let fields_map = index.fields_ids_map(&txn).unwrap();
         let fid = fields_map.id(&distinct).unwrap();
 
-        let map = (0..JSON.as_array().unwrap().len() as u32).collect();
+        let documents = DocumentsReader::from_reader(Cursor::new(&*JSON)).unwrap();
+        let map = (0..documents.len() as u32).collect();
 
         txn.commit().unwrap();
 
diff --git a/milli/src/update/clear_documents.rs b/milli/src/update/clear_documents.rs
index 789970a8ea..0f74c1e2de 100644
--- a/milli/src/update/clear_documents.rs
+++ b/milli/src/update/clear_documents.rs
@@ -80,7 +80,7 @@ mod tests {
     use heed::EnvOpenOptions;
 
     use super::*;
-    use crate::update::{IndexDocuments, UpdateFormat};
+    use crate::update::{IndexDocuments};
 
     #[test]
     fn clear_documents() {
@@ -90,13 +90,12 @@ mod tests {
         let index = Index::new(options, &path).unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
+        let content = documents!([
             { "id": 0, "name": "kevin", "age": 20 },
             { "id": 1, "name": "kevina" },
             { "id": 2, "name": "benoit", "country": "France" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         // Clear all documents from the database.
diff --git a/milli/src/update/delete_documents.rs b/milli/src/update/delete_documents.rs
index e18c6bbd1d..3dd0ab303b 100644
--- a/milli/src/update/delete_documents.rs
+++ b/milli/src/update/delete_documents.rs
@@ -548,7 +548,7 @@ mod tests {
     use maplit::hashset;
 
     use super::*;
-    use crate::update::{IndexDocuments, Settings, UpdateFormat};
+    use crate::update::{IndexDocuments, Settings};
     use crate::FilterCondition;
 
     #[test]
@@ -559,13 +559,12 @@ mod tests {
         let index = Index::new(options, &path).unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
+        let content = documents!([
             { "id": 0, "name": "kevin", "object": { "key1": "value1", "key2": "value2" } },
             { "id": 1, "name": "kevina", "array": ["I", "am", "fine"] },
             { "id": 2, "name": "benoit", "array_of_object": [{ "wow": "amazing" }] }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         // delete those documents, ids are synchronous therefore 0, 1, and 2.
@@ -590,13 +589,12 @@ mod tests {
         let index = Index::new(options, &path).unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
+        let content = documents!([
             { "mysuperid": 0, "name": "kevin" },
             { "mysuperid": 1, "name": "kevina" },
             { "mysuperid": 2, "name": "benoit" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         // Delete not all of the documents but some of them.
@@ -621,7 +619,7 @@ mod tests {
         builder.set_filterable_fields(hashset! { S("label") });
         builder.execute(|_, _| ()).unwrap();
 
-        let content = &br#"[
+        let content = documents!([
             {"docid":"1_4","label":"sign"},
             {"docid":"1_5","label":"letter"},
             {"docid":"1_7","label":"abstract,cartoon,design,pattern"},
@@ -642,9 +640,8 @@ mod tests {
             {"docid":"1_58","label":"abstract,art,cartoon"},
             {"docid":"1_68","label":"design"},
             {"docid":"1_69","label":"geometry"}
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         // Delete not all of the documents but some of them.
diff --git a/milli/src/update/index_documents/mod.rs b/milli/src/update/index_documents/mod.rs
index b948b0111c..cb9e7785da 100644
--- a/milli/src/update/index_documents/mod.rs
+++ b/milli/src/update/index_documents/mod.rs
@@ -822,9 +822,11 @@ mod tests {
     use std::io::Cursor;
 
     use big_s::S;
+    use bimap::BiHashMap;
     use heed::EnvOpenOptions;
 
     use super::*;
+    use crate::documents::DocumentsBuilder;
     use crate::update::DeleteDocuments;
     use crate::HashMap;
 
@@ -837,9 +839,12 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name\n1,kevin\n2,kevina\n3,benoit\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([
+            { "id": 1, "name": "kevin" },
+            { "id": 2, "name": "kevina" },
+            { "id": 3, "name": "benoit" }
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -851,9 +856,8 @@ mod tests {
 
         // Second we send 1 document with id 1, to erase the previous ones.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name\n1,updated kevin\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([ { "id": 1, "name": "updated kevin" } ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 1);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -865,9 +869,12 @@ mod tests {
 
         // Third we send 3 documents again to replace the existing ones.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name\n1,updated second kevin\n2,updated kevina\n3,updated benoit\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 2);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([
+            { "id": 1, "name": "updated second kevin" },
+            { "id": 2, "name": "updated kevina" },
+            { "id": 3, "name": "updated benoit" }
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 2);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -888,9 +895,12 @@ mod tests {
         // First we send 3 documents with duplicate ids and
         // change the index method to merge documents.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name\n1,kevin\n1,kevina\n1,benoit\n"[..];
+        let content = documents!([
+            { "id": 1, "name": "kevin" },
+            { "id": 1, "name": "kevina" },
+            { "id": 1, "name": "benoit" }
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
         builder.index_documents_method(IndexDocumentsMethod::UpdateDocuments);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
@@ -915,9 +925,8 @@ mod tests {
 
         // Second we send 1 document with id 1, to force it to be merged with the previous one.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,age\n1,25\n"[..];
+        let content = documents!([ { "id": 1, "age": 25 } ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Csv);
         builder.index_documents_method(IndexDocumentsMethod::UpdateDocuments);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
@@ -937,13 +946,13 @@ mod tests {
         let mut doc_iter = doc.iter();
         assert_eq!(doc_iter.next(), Some((0, &br#""1""#[..])));
         assert_eq!(doc_iter.next(), Some((1, &br#""benoit""#[..])));
-        assert_eq!(doc_iter.next(), Some((2, &br#""25""#[..])));
+        assert_eq!(doc_iter.next(), Some((2, &br#"25"#[..])));
         assert_eq!(doc_iter.next(), None);
         drop(rtxn);
     }
 
     #[test]
-    fn not_auto_generated_csv_documents_ids() {
+    fn not_auto_generated_documents_ids() {
         let path = tempfile::tempdir().unwrap();
         let mut options = EnvOpenOptions::new();
         options.map_size(10 * 1024 * 1024); // 10 MB
@@ -951,35 +960,12 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name\nkevin\nkevina\nbenoit\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
-        assert!(builder.execute(content, |_, _| ()).is_err());
-        wtxn.commit().unwrap();
-
-        // Check that there is no document.
-        let rtxn = index.read_txn().unwrap();
-        let count = index.number_of_documents(&rtxn).unwrap();
-        assert_eq!(count, 0);
-        drop(rtxn);
-    }
-
-    #[test]
-    fn not_auto_generated_json_documents_ids() {
-        let path = tempfile::tempdir().unwrap();
-        let mut options = EnvOpenOptions::new();
-        options.map_size(10 * 1024 * 1024); // 10 MB
-        let index = Index::new(options, &path).unwrap();
-
-        // First we send 3 documents and 2 without ids.
-        let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
-            { "name": "kevina", "id": 21 },
+        let content = documents!([
             { "name": "kevin" },
+            { "name": "kevina" },
             { "name": "benoit" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         assert!(builder.execute(content, |_, _| ()).is_err());
         wtxn.commit().unwrap();
 
@@ -999,10 +985,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name\nkevin\nkevina\nbenoit\n"[..];
+        let content = documents!([
+            { "name": "kevin" },
+            { "name": "kevina" },
+            { "name": "benoit" }
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -1018,10 +1007,9 @@ mod tests {
 
         // Second we send 1 document with the generated uuid, to erase the previous ones.
         let mut wtxn = index.write_txn().unwrap();
-        let content = format!("id,name\n{},updated kevin", kevin_uuid);
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Csv);
-        builder.execute(content.as_bytes(), |_, _| ()).unwrap();
+        let content = documents!([ { "name": "updated kevin", "id": kevin_uuid } ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 1);
+        builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
         // Check that there is **always** 3 documents.
@@ -1031,7 +1019,8 @@ mod tests {
 
         let docs = index.documents(&rtxn, vec![0, 1, 2]).unwrap();
         let (kevin_id, _) =
-            docs.iter().find(|(_, d)| d.get(0).unwrap() == br#""updated kevin""#).unwrap();
+            docs.iter().find(|(_, d)| d.get(0).unwrap() == br#""updated kevin""#)
+            .unwrap();
         let (id, doc) = docs[*kevin_id as usize];
         assert_eq!(id, *kevin_id);
 
@@ -1052,9 +1041,12 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name\n1,kevin\n2,kevina\n3,benoit\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([
+            { "id": 1, "name": "kevin" },
+            { "id": 2, "name": "kevina" },
+            { "id": 3, "name": "benoit" }
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -1066,9 +1058,9 @@ mod tests {
 
         // Second we send 1 document without specifying the id.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name\nnew kevin"[..];
+        let content = documents!([ { "name": "new kevin" } ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Csv);
+        builder.enable_autogenerate_docids();
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -1080,7 +1072,7 @@ mod tests {
     }
 
     #[test]
-    fn empty_csv_update() {
+    fn empty_update() {
         let path = tempfile::tempdir().unwrap();
         let mut options = EnvOpenOptions::new();
         options.map_size(10 * 1024 * 1024); // 10 MB
@@ -1088,9 +1080,8 @@ mod tests {
 
         // First we send 0 documents and only headers.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -1101,83 +1092,6 @@ mod tests {
         drop(rtxn);
     }
 
-    #[test]
-    fn json_documents() {
-        let path = tempfile::tempdir().unwrap();
-        let mut options = EnvOpenOptions::new();
-        options.map_size(10 * 1024 * 1024); // 10 MB
-        let index = Index::new(options, &path).unwrap();
-
-        // First we send 3 documents with an id for only one of them.
-        let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
-            { "name": "kevin" },
-            { "name": "kevina", "id": 21 },
-            { "name": "benoit" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Json);
-        builder.execute(content, |_, _| ()).unwrap();
-        wtxn.commit().unwrap();
-
-        // Check that there is 3 documents now.
-        let rtxn = index.read_txn().unwrap();
-        let count = index.number_of_documents(&rtxn).unwrap();
-        assert_eq!(count, 3);
-        drop(rtxn);
-    }
-
-    #[test]
-    fn empty_json_update() {
-        let path = tempfile::tempdir().unwrap();
-        let mut options = EnvOpenOptions::new();
-        options.map_size(10 * 1024 * 1024); // 10 MB
-        let index = Index::new(options, &path).unwrap();
-
-        // First we send 0 documents.
-        let mut wtxn = index.write_txn().unwrap();
-        let content = &b"[]"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Json);
-        builder.execute(content, |_, _| ()).unwrap();
-        wtxn.commit().unwrap();
-
-        // Check that there is no documents.
-        let rtxn = index.read_txn().unwrap();
-        let count = index.number_of_documents(&rtxn).unwrap();
-        assert_eq!(count, 0);
-        drop(rtxn);
-    }
-
-    #[test]
-    fn json_stream_documents() {
-        let path = tempfile::tempdir().unwrap();
-        let mut options = EnvOpenOptions::new();
-        options.map_size(10 * 1024 * 1024); // 10 MB
-        let index = Index::new(options, &path).unwrap();
-
-        // First we send 3 documents with an id for only one of them.
-        let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"
-        { "name": "kevin" }
-        { "name": "kevina", "id": 21 }
-        { "name": "benoit" }
-        "#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::JsonStream);
-        builder.execute(content, |_, _| ()).unwrap();
-        wtxn.commit().unwrap();
-
-        // Check that there is 3 documents now.
-        let rtxn = index.read_txn().unwrap();
-        let count = index.number_of_documents(&rtxn).unwrap();
-        assert_eq!(count, 3);
-        drop(rtxn);
-    }
-
     #[test]
     fn invalid_documents_ids() {
         let path = tempfile::tempdir().unwrap();
@@ -1188,18 +1102,16 @@ mod tests {
         // First we send 1 document with an invalid id.
         let mut wtxn = index.write_txn().unwrap();
         // There is a space in the document id.
-        let content = &b"id,name\nbrume bleue,kevin\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([ { "id": "brume bleue", "name": "kevin" } ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         assert!(builder.execute(content, |_, _| ()).is_err());
         wtxn.commit().unwrap();
 
         // First we send 1 document with a valid id.
         let mut wtxn = index.write_txn().unwrap();
         // There is a space in the document id.
-        let content = &b"id,name\n32,kevin\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Csv);
+        let content = documents!([ { "id": 32, "name": "kevin" } ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 1);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -1211,7 +1123,7 @@ mod tests {
     }
 
     #[test]
-    fn complex_json_documents() {
+    fn complex_documents() {
         let path = tempfile::tempdir().unwrap();
         let mut options = EnvOpenOptions::new();
         options.map_size(10 * 1024 * 1024); // 10 MB
@@ -1219,13 +1131,12 @@ mod tests {
 
         // First we send 3 documents with an id for only one of them.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
+        let content = documents!([
             { "id": 0, "name": "kevin", "object": { "key1": "value1", "key2": "value2" } },
             { "id": 1, "name": "kevina", "array": ["I", "am", "fine"] },
             { "id": 2, "name": "benoit", "array_of_object": [{ "wow": "amazing" }] }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -1256,33 +1167,31 @@ mod tests {
 
         // First we send 3 documents with an id for only one of them.
         let mut wtxn = index.write_txn().unwrap();
-        let documents = &r#"[
+        let documents = documents!([
           { "id": 2,    "title": "Pride and Prejudice",                    "author": "Jane Austin",              "genre": "romance",    "price": 3.5 },
           { "id": 456,  "title": "Le Petit Prince",                        "author": "Antoine de Saint-Exupéry", "genre": "adventure" , "price": 10.0 },
           { "id": 1,    "title": "Alice In Wonderland",                    "author": "Lewis Carroll",            "genre": "fantasy",    "price": 25.99 },
           { "id": 1344, "title": "The Hobbit",                             "author": "J. R. R. Tolkien",         "genre": "fantasy" },
           { "id": 4,    "title": "Harry Potter and the Half-Blood Prince", "author": "J. K. Rowling",            "genre": "fantasy" },
           { "id": 42,   "title": "The Hitchhiker's Guide to the Galaxy",   "author": "Douglas Adams" }
-        ]"#[..];
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
         builder.index_documents_method(IndexDocumentsMethod::ReplaceDocuments);
-        builder.execute(Cursor::new(documents), |_, _| ()).unwrap();
+        builder.execute(documents, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Json);
         builder.index_documents_method(IndexDocumentsMethod::UpdateDocuments);
-        let documents = &r#"[
+        let documents = documents!([
           {
             "id": 2,
             "author": "J. Austen",
             "date": "1813"
           }
-        ]"#[..];
+        ]);
 
-        builder.execute(Cursor::new(documents), |_, _| ()).unwrap();
+        builder.execute(documents, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
     }
 
@@ -1294,14 +1203,13 @@ mod tests {
         let index = Index::new(options, &path).unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
+        let content = documents!([
             { "objectId": 123, "title": "Pride and Prejudice", "comment": "A great book" },
             { "objectId": 456, "title": "Le Petit Prince",     "comment": "A french book" },
             { "objectId": 1,   "title": "Alice In Wonderland", "comment": "A weird book" },
             { "objectId": 30,  "title": "Hamlet" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         assert_eq!(index.primary_key(&wtxn).unwrap(), Some("objectId"));
@@ -1314,21 +1222,19 @@ mod tests {
         let external_documents_ids = index.external_documents_ids(&wtxn).unwrap();
         assert!(external_documents_ids.get("30").is_none());
 
-        let content = &br#"[
+        let content = documents!([
             { "objectId": 30, "title": "Hamlet" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         let external_documents_ids = index.external_documents_ids(&wtxn).unwrap();
         assert!(external_documents_ids.get("30").is_some());
 
-        let content = &br#"[
+        let content = documents!([
             { "objectId": 30, "title": "Hamlet" }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
 
         wtxn.commit().unwrap();
@@ -1350,12 +1256,16 @@ mod tests {
             big_object.insert(key, "I am a text!");
         }
 
-        let content = vec![big_object];
-        let content = serde_json::to_string(&content).unwrap();
+        let mut cursor = Cursor::new(Vec::new());
 
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Json);
-        builder.execute(Cursor::new(content), |_, _| ()).unwrap();
+        let mut builder = DocumentsBuilder::new(&mut cursor, BiHashMap::new()).unwrap();
+        builder.add_documents(big_object).unwrap();
+        builder.finish().unwrap();
+        cursor.set_position(0);
+        let content = DocumentsReader::from_reader(cursor).unwrap();
+
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
+        builder.execute(content, |_, _| ()).unwrap();
 
         wtxn.commit().unwrap();
     }
@@ -1368,16 +1278,38 @@ mod tests {
         let index = Index::new(options, &path).unwrap();
 
         let mut wtxn = index.write_txn().unwrap();
-        let content = r#"#id,title,au{hor,genre,price$
-2,"Prideand Prejudice","Jane Austin","romance",3.5$
-456,"Le Petit Prince","Antoine de Saint-Exupéry","adventure",10.0$
-1,Wonderland","Lewis Carroll","fantasy",25.99$
-4,"Harry Potter ing","fantasy\0lood Prince","J. K. Rowling","fantasy\0,
-"#;
-
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
-        builder.execute(content.as_bytes(), |_, _| ()).unwrap();
+        let content = documents!([
+            {
+                "id": 2,
+                "title": "Prideand Prejudice",
+                "au{hor": "Jane Austin",
+                "genre": "romance",
+                "price$": "3.5$",
+            },
+            {
+                "id": 456,
+                "title": "Le Petit Prince",
+                "au{hor": "Antoine de Saint-Exupéry",
+                "genre": "adventure",
+                "price$": "10.0$",
+            },
+            {
+                "id": 1,
+                "title": "Wonderland",
+                "au{hor": "Lewis Carroll",
+                "genre": "fantasy",
+                "price$": "25.99$",
+            },
+            {
+                "id": 4,
+                "title": "Harry Potter ing fantasy\0lood Prince",
+                "au{hor": "J. K. Rowling",
+                "genre": "fantasy\0",
+            },
+        ]);
+
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
+        builder.execute(content, |_, _| ()).unwrap();
 
         wtxn.commit().unwrap();
     }
diff --git a/milli/src/update/settings.rs b/milli/src/update/settings.rs
index c0b5e45498..24b29f7047 100644
--- a/milli/src/update/settings.rs
+++ b/milli/src/update/settings.rs
@@ -113,6 +113,10 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
         }
     }
 
+    pub fn log_every_n(&mut self, n: usize) {
+        self.log_every_n = Some(n);
+    }
+
     pub fn reset_searchable_fields(&mut self) {
         self.searchable_fields = Setting::Reset;
     }
@@ -501,7 +505,7 @@ mod tests {
 
     use super::*;
     use crate::error::Error;
-    use crate::update::{IndexDocuments, UpdateFormat};
+    use crate::update::IndexDocuments;
     use crate::{Criterion, FilterCondition, SearchResult};
 
     #[test]
@@ -513,9 +517,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"id,name,age\n0,kevin,23\n1,kevina,21\n2,benoit,34\n"[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-        builder.update_format(UpdateFormat::Csv);
+
+        let content = documents!([
+            { "id": 1, "name": "kevin", "age": 23 },
+            { "id": 2, "name": "kevina", "age": 21},
+            { "id": 3, "name": "benoit", "age": 34 }
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -567,10 +575,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name,age\nkevin,23\nkevina,21\nbenoit,34\n"[..];
+        let content = documents!([
+            { "name": "kevin", "age": 23},
+            { "name": "kevina", "age": 21 },
+            { "name": "benoit", "age": 34 }
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -611,10 +622,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name,age\nkevin,23\nkevina,21\nbenoit,34\n"[..];
+        let content = documents!([
+            { "name": "kevin", "age": 23},
+            { "name": "kevina", "age": 21 },
+            { "name": "benoit", "age": 34 }
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -633,10 +647,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name,age\nkevin,23\nkevina,21\nbenoit,34\n"[..];
+        let content = documents!([
+            { "name": "kevin", "age": 23},
+            { "name": "kevina", "age": 21 },
+            { "name": "benoit", "age": 34 }
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
 
         // In the same transaction we change the displayed fields to be only the age.
@@ -678,13 +695,12 @@ mod tests {
         builder.execute(|_, _| ()).unwrap();
 
         // Then index some documents.
-        let content = &br#"[
-            { "name": "kevin",  "age": 23 },
+        let content = documents!([
+            { "name": "kevin", "age": 23},
             { "name": "kevina", "age": 21 },
             { "name": "benoit", "age": 34 }
-        ]"#[..];
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Json);
         builder.enable_autogenerate_docids();
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
@@ -695,11 +711,19 @@ mod tests {
         assert_eq!(fields_ids, hashset! { S("age") });
         // Only count the field_id 0 and level 0 facet values.
         // TODO we must support typed CSVs for numbers to be understood.
+        let fidmap = index.fields_ids_map(&rtxn).unwrap();
+        println!("fidmap: {:?}", fidmap);
+        for document in index.all_documents(&rtxn).unwrap() {
+            let document = document.unwrap();
+            let json = crate::obkv_to_json(&fidmap.ids().collect::<Vec<_>>(), &fidmap, document.1).unwrap();
+            println!("json: {:?}", json);
+
+        }
         let count = index
             .facet_id_f64_docids
             .remap_key_type::<ByteSlice>()
             // The faceted field id is 2u16
-            .prefix_iter(&rtxn, &[0, 2, 0])
+            .prefix_iter(&rtxn, &[0, 1, 0])
             .unwrap()
             .count();
         assert_eq!(count, 3);
@@ -707,25 +731,23 @@ mod tests {
 
         // Index a little more documents with new and current facets values.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &br#"[
-            { "name": "kevin2",  "age": 23 },
+        let content = documents!([
+            { "name": "kevin2", "age": 23},
             { "name": "kevina2", "age": 21 },
-            { "name": "benoit",  "age": 35 }
-        ]"#[..];
+            { "name": "benoit", "age": 35 }
+        ]);
 
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 2);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Json);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
         let rtxn = index.read_txn().unwrap();
         // Only count the field_id 0 and level 0 facet values.
-        // TODO we must support typed CSVs for numbers to be understood.
         let count = index
             .facet_id_f64_docids
             .remap_key_type::<ByteSlice>()
-            .prefix_iter(&rtxn, &[0, 2, 0])
+            .prefix_iter(&rtxn, &[0, 1, 0])
             .unwrap()
             .count();
         assert_eq!(count, 4);
@@ -747,13 +769,12 @@ mod tests {
         builder.execute(|_, _| ()).unwrap();
 
         // Then index some documents.
-        let content = &br#"[
-            { "name": "kevin",  "age": 23 },
+        let content = documents!([
+            { "name": "kevin", "age": 23},
             { "name": "kevina", "age": 21 },
             { "name": "benoit", "age": 34 }
-        ]"#[..];
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Json);
         builder.enable_autogenerate_docids();
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
@@ -790,7 +811,7 @@ mod tests {
         builder.execute(|_, _| ()).unwrap();
 
         // Then index some documents.
-        let content = &br#"[
+        let content = documents!([
             { "name": "kevin",  "age": 23 },
             { "name": "kevina", "age": 21 },
             { "name": "benoit", "age": 34 },
@@ -798,9 +819,8 @@ mod tests {
             { "name": "bertrand", "age": 34 },
             { "name": "bernie", "age": 34 },
             { "name": "ben", "age": 34 }
-        ]"#[..];
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Json);
         builder.enable_autogenerate_docids();
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
@@ -822,10 +842,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name,age\nkevin,23\nkevina,21\nbenoit,34\n"[..];
+        let content = documents!([
+            { "name": "kevin", "age": 23},
+            { "name": "kevina", "age": 21 },
+            { "name": "benoit", "age": 34 }
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
@@ -844,10 +867,13 @@ mod tests {
 
         // First we send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name,age,maxim\nkevin,23,I love dogs\nkevina,21,Doggos are the best\nbenoit,34,The crepes are really good\n"[..];
+        let content = documents!([
+            { "name": "kevin", "age": 23, "maxim": "I love dogs" },
+            { "name": "kevina", "age": 21, "maxim": "Doggos are the best" },
+            { "name": "benoit", "age": 34, "maxim": "The crepes are really good" },
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
 
         // In the same transaction we provide some stop_words
@@ -915,10 +941,13 @@ mod tests {
 
         // Send 3 documents with ids from 1 to 3.
         let mut wtxn = index.write_txn().unwrap();
-        let content = &b"name,age,maxim\nkevin,23,I love dogs\nkevina,21,Doggos are the best\nbenoit,34,The crepes are really good\n"[..];
+        let content = documents!([
+            { "name": "kevin", "age": 23, "maxim": "I love dogs"},
+            { "name": "kevina", "age": 21, "maxim": "Doggos are the best"},
+            { "name": "benoit", "age": 34, "maxim": "The crepes are really good"},
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.enable_autogenerate_docids();
-        builder.update_format(UpdateFormat::Csv);
         builder.execute(content, |_, _| ()).unwrap();
 
         // In the same transaction provide some synonyms
@@ -1038,7 +1067,7 @@ mod tests {
         assert_eq!(index.primary_key(&wtxn).unwrap(), Some("mykey"));
 
         // Then index some documents with the "mykey" primary key.
-        let content = &br#"[
+        let content = documents!([
             { "mykey": 1, "name": "kevin",  "age": 23 },
             { "mykey": 2, "name": "kevina", "age": 21 },
             { "mykey": 3, "name": "benoit", "age": 34 },
@@ -1046,9 +1075,8 @@ mod tests {
             { "mykey": 5, "name": "bertrand", "age": 34 },
             { "mykey": 6, "name": "bernie", "age": 34 },
             { "mykey": 7, "name": "ben", "age": 34 }
-        ]"#[..];
+        ]);
         let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Json);
         builder.disable_autogenerate_docids();
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
@@ -1087,7 +1115,7 @@ mod tests {
         builder.set_filterable_fields(hashset! { S("genres") });
         builder.execute(|_, _| ()).unwrap();
 
-        let content = &br#"[
+        let content = documents!([
           {
             "id": 11,
             "title": "Star Wars",
@@ -1105,9 +1133,8 @@ mod tests {
             "poster": "https://image.tmdb.org/t/p/w500/gSuHDeWemA1menrwfMRChnSmMVN.jpg",
             "release_date": 819676800
           }
-        ]"#[..];
-        let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
-        builder.update_format(UpdateFormat::Json);
+        ]);
+        let builder = IndexDocuments::new(&mut wtxn, &index, 1);
         builder.execute(content, |_, _| ()).unwrap();
         wtxn.commit().unwrap();
 
diff --git a/milli/tests/search/mod.rs b/milli/tests/search/mod.rs
index 7d4043ff18..604ae04db5 100644
--- a/milli/tests/search/mod.rs
+++ b/milli/tests/search/mod.rs
@@ -1,12 +1,14 @@
 use std::cmp::Reverse;
 use std::collections::HashSet;
+use std::io::Cursor;
 
 use big_s::S;
 use either::{Either, Left, Right};
 use heed::EnvOpenOptions;
 use maplit::{hashmap, hashset};
-use milli::update::{IndexDocuments, Settings, UpdateFormat};
+use milli::update::{IndexDocuments, Settings};
 use milli::{AscDesc, Criterion, DocumentId, Index};
+use milli::documents::{DocumentsBuilder, DocumentsReader};
 use serde::Deserialize;
 use slice_group_by::GroupBy;
 
@@ -49,11 +51,21 @@ pub fn setup_search_index_with_criteria(criteria: &[Criterion]) -> Index {
     builder.set_searchable_fields(vec![S("title"), S("description")]);
     builder.execute(|_, _| ()).unwrap();
 
-    // index documents
     let mut builder = IndexDocuments::new(&mut wtxn, &index, 0);
-    builder.update_format(UpdateFormat::JsonStream);
     builder.enable_autogenerate_docids();
-    builder.execute(CONTENT.as_bytes(), |_, _| ()).unwrap();
+    let mut cursor = Cursor::new(Vec::new());
+    let mut documents_builder = DocumentsBuilder::new(&mut cursor, bimap::BiHashMap::new()).unwrap();
+    let reader = Cursor::new(CONTENT.as_bytes());
+    for doc in serde_json::Deserializer::from_reader(reader).into_iter::<serde_json::Value>() {
+        documents_builder.add_documents(doc.unwrap()).unwrap();
+    }
+    documents_builder.finish().unwrap();
+
+    cursor.set_position(0);
+
+    // index documents
+    let content = DocumentsReader::from_reader(cursor).unwrap();
+    builder.execute(content, |_, _| ()).unwrap();
 
     wtxn.commit().unwrap();