diff --git a/benchmarks/benches/utils.rs b/benchmarks/benches/utils.rs
index 24f5d5343..dbe8fffad 100644
--- a/benchmarks/benches/utils.rs
+++ b/benchmarks/benches/utils.rs
@@ -1,7 +1,7 @@
 #![allow(dead_code)]
 
 use std::fs::{create_dir_all, remove_dir_all, File};
-use std::io::{self, Cursor, Read, Seek};
+use std::io::{self, BufRead, BufReader, Cursor, Read, Seek};
 use std::num::ParseFloatError;
 use std::path::Path;
 
@@ -146,44 +146,34 @@ pub fn documents_from(filename: &str, filetype: &str) -> DocumentBatchReader<imp
     DocumentBatchReader::from_reader(Cursor::new(documents)).unwrap()
 }
 
-fn documents_from_jsonl(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
+fn documents_from_jsonl(reader: impl Read) -> anyhow::Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
     let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
 
-    let values = serde_json::Deserializer::from_reader(reader)
-        .into_iter::<serde_json::Map<String, serde_json::Value>>();
-    for document in values {
-        let document = document?;
-        documents.add_documents(document)?;
+    let mut buf = String::new();
+    let mut reader = BufReader::new(reader);
+
+    while reader.read_line(&mut buf)? > 0 {
+        documents.extend_from_json(&mut buf.as_bytes())?;
     }
     documents.finish()?;
 
     Ok(writer.into_inner())
 }
 
-fn documents_from_json(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
+fn documents_from_json(reader: impl Read) -> anyhow::Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
     let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
 
-    let json: serde_json::Value = serde_json::from_reader(reader)?;
-    documents.add_documents(json)?;
+    documents.extend_from_json(reader)?;
     documents.finish()?;
 
     Ok(writer.into_inner())
 }
 
-fn documents_from_csv(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
+fn documents_from_csv(reader: impl Read) -> anyhow::Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
-    let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
-
-    let iter = CSVDocumentDeserializer::from_reader(reader)?;
-
-    for doc in iter {
-        let doc = doc?;
-        documents.add_documents(doc)?;
-    }
-
-    documents.finish()?;
+    milli::documents::DocumentBatchBuilder::from_csv(reader, &mut writer)?.finish()?;
 
     Ok(writer.into_inner())
 }
diff --git a/cli/src/main.rs b/cli/src/main.rs
index b84ff3243..8e28d4a25 100644
--- a/cli/src/main.rs
+++ b/cli/src/main.rs
@@ -1,5 +1,5 @@
 use std::fs::File;
-use std::io::{stdin, Cursor, Read};
+use std::io::{stdin, BufRead, BufReader, Cursor, Read};
 use std::path::PathBuf;
 use std::str::FromStr;
 
@@ -9,7 +9,6 @@ use indicatif::{MultiProgress, ProgressBar, ProgressStyle};
 use milli::update::UpdateIndexingStep::{
     ComputeIdsAndMergeDocuments, IndexDocuments, MergeDataIntoFinalDatabase, RemapDocumentAddition,
 };
-use serde_json::{Map, Value};
 use structopt::StructOpt;
 
 #[cfg(target_os = "linux")]
@@ -202,11 +201,11 @@ fn documents_from_jsonl(reader: impl Read) -> Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
     let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
 
-    let values = serde_json::Deserializer::from_reader(reader)
-        .into_iter::<serde_json::Map<String, serde_json::Value>>();
-    for document in values {
-        let document = document?;
-        documents.add_documents(document)?;
+    let mut buf = String::new();
+    let mut reader = BufReader::new(reader);
+
+    while reader.read_line(&mut buf)? > 0 {
+        documents.extend_from_json(&mut buf.as_bytes())?;
     }
     documents.finish()?;
 
@@ -217,8 +216,7 @@ fn documents_from_json(reader: impl Read) -> Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
     let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
 
-    let json: serde_json::Value = serde_json::from_reader(reader)?;
-    documents.add_documents(json)?;
+    documents.extend_from_json(reader)?;
     documents.finish()?;
 
     Ok(writer.into_inner())
@@ -226,17 +224,7 @@ fn documents_from_json(reader: impl Read) -> Result<Vec<u8>> {
 
 fn documents_from_csv(reader: impl Read) -> Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
-    let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
-
-    let mut records = csv::Reader::from_reader(reader);
-    let iter = records.deserialize::<Map<String, Value>>();
-
-    for doc in iter {
-        let doc = doc?;
-        documents.add_documents(doc)?;
-    }
-
-    documents.finish()?;
+    milli::documents::DocumentBatchBuilder::from_csv(reader, &mut writer)?.finish()?;
 
     Ok(writer.into_inner())
 }
diff --git a/http-ui/src/documents_from_csv.rs b/http-ui/src/documents_from_csv.rs
deleted file mode 100644
index 2b62f23c2..000000000
--- a/http-ui/src/documents_from_csv.rs
+++ /dev/null
@@ -1,285 +0,0 @@
-use std::io::{Read, Result as IoResult};
-use std::num::ParseFloatError;
-
-use serde_json::{Map, Value};
-
-enum AllowedType {
-    String,
-    Number,
-}
-
-fn parse_csv_header(header: &str) -> (String, AllowedType) {
-    // if there are several separators we only split on the last one.
-    match header.rsplit_once(':') {
-        Some((field_name, field_type)) => match field_type {
-            "string" => (field_name.to_string(), AllowedType::String),
-            "number" => (field_name.to_string(), AllowedType::Number),
-            // we may return an error in this case.
-            _otherwise => (header.to_string(), AllowedType::String),
-        },
-        None => (header.to_string(), AllowedType::String),
-    }
-}
-
-pub struct CSVDocumentDeserializer<R>
-where
-    R: Read,
-{
-    documents: csv::StringRecordsIntoIter<R>,
-    headers: Vec<(String, AllowedType)>,
-}
-
-impl<R: Read> CSVDocumentDeserializer<R> {
-    pub fn from_reader(reader: R) -> IoResult<Self> {
-        let mut records = csv::Reader::from_reader(reader);
-
-        let headers = records.headers()?.into_iter().map(parse_csv_header).collect();
-
-        Ok(Self { documents: records.into_records(), headers })
-    }
-}
-
-impl<R: Read> Iterator for CSVDocumentDeserializer<R> {
-    type Item = anyhow::Result<Map<String, Value>>;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        let csv_document = self.documents.next()?;
-
-        match csv_document {
-            Ok(csv_document) => {
-                let mut document = Map::new();
-
-                for ((field_name, field_type), value) in
-                    self.headers.iter().zip(csv_document.into_iter())
-                {
-                    let parsed_value: Result<Value, ParseFloatError> = match field_type {
-                        AllowedType::Number => {
-                            value.parse::<f64>().map(Value::from).map_err(Into::into)
-                        }
-                        AllowedType::String => Ok(Value::String(value.to_string())),
-                    };
-
-                    match parsed_value {
-                        Ok(value) => drop(document.insert(field_name.to_string(), value)),
-                        Err(_e) => {
-                            return Some(Err(anyhow::anyhow!(
-                                "Value '{}' is not a valid number",
-                                value
-                            )))
-                        }
-                    }
-                }
-
-                Some(Ok(document))
-            }
-            Err(e) => Some(Err(anyhow::anyhow!("Error parsing csv document: {}", e))),
-        }
-    }
-}
-
-#[cfg(test)]
-mod test {
-    use serde_json::json;
-
-    use super::*;
-
-    #[test]
-    fn simple_csv_document() {
-        let documents = r#"city,country,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city": "Boston",
-                "country": "United States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn coma_in_field() {
-        let documents = r#"city,country,pop
-"Boston","United, States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city": "Boston",
-                "country": "United, States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn quote_in_field() {
-        let documents = r#"city,country,pop
-"Boston","United"" States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city": "Boston",
-                "country": "United\" States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn integer_in_field() {
-        let documents = r#"city,country,pop:number
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city": "Boston",
-                "country": "United States",
-                "pop": 4628910.0,
-            })
-        );
-    }
-
-    #[test]
-    fn float_in_field() {
-        let documents = r#"city,country,pop:number
-"Boston","United States","4628910.01""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city": "Boston",
-                "country": "United States",
-                "pop": 4628910.01,
-            })
-        );
-    }
-
-    #[test]
-    fn several_double_dot_in_header() {
-        let documents = r#"city:love:string,country:state,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city:love": "Boston",
-                "country:state": "United States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn ending_by_double_dot_in_header() {
-        let documents = r#"city:,country,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city:": "Boston",
-                "country": "United States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn starting_by_double_dot_in_header() {
-        let documents = r#":city,country,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                ":city": "Boston",
-                "country": "United States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn starting_by_double_dot_in_header2() {
-        let documents = r#":string,country,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "": "Boston",
-                "country": "United States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn double_double_dot_in_header() {
-        let documents = r#"city::string,country,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert_eq!(
-            Value::Object(csv_iter.next().unwrap().unwrap()),
-            json!({
-                "city:": "Boston",
-                "country": "United States",
-                "pop": "4628910",
-            })
-        );
-    }
-
-    #[test]
-    fn bad_type_in_header() {
-        let documents = r#"city,country:number,pop
-"Boston","United States","4628910""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert!(csv_iter.next().unwrap().is_err());
-    }
-
-    #[test]
-    fn bad_column_count1() {
-        let documents = r#"city,country,pop
-"Boston","United States","4628910", "too much""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert!(csv_iter.next().unwrap().is_err());
-    }
-
-    #[test]
-    fn bad_column_count2() {
-        let documents = r#"city,country,pop
-"Boston","United States""#;
-
-        let mut csv_iter = CSVDocumentDeserializer::from_reader(documents.as_bytes()).unwrap();
-
-        assert!(csv_iter.next().unwrap().is_err());
-    }
-}
diff --git a/http-ui/src/main.rs b/http-ui/src/main.rs
index d27c6d5bb..9e9fe4a2b 100644
--- a/http-ui/src/main.rs
+++ b/http-ui/src/main.rs
@@ -1,10 +1,9 @@
-mod documents_from_csv;
 mod update_store;
 
 use std::collections::{BTreeMap, BTreeSet, HashMap, HashSet};
 use std::fmt::Display;
 use std::fs::{create_dir_all, File};
-use std::io::Cursor;
+use std::io::{BufRead, BufReader, Cursor};
 use std::net::SocketAddr;
 use std::num::{NonZeroU32, NonZeroUsize};
 use std::path::PathBuf;
@@ -39,7 +38,6 @@ use warp::http::Response;
 use warp::Filter;
 
 use self::update_store::UpdateStore;
-use crate::documents_from_csv::CSVDocumentDeserializer;
 
 #[cfg(target_os = "linux")]
 #[global_allocator]
@@ -1041,11 +1039,11 @@ fn documents_from_jsonl(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
     let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
 
-    let values = serde_json::Deserializer::from_reader(reader)
-        .into_iter::<serde_json::Map<String, serde_json::Value>>();
-    for document in values {
-        let document = document?;
-        documents.add_documents(document)?;
+    let mut buf = String::new();
+    let mut reader = BufReader::new(reader);
+
+    while reader.read_line(&mut buf)? > 0 {
+        documents.extend_from_json(&mut buf.as_bytes())?;
     }
     documents.finish()?;
 
@@ -1056,8 +1054,7 @@ fn documents_from_json(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
     let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
 
-    let json: serde_json::Value = serde_json::from_reader(reader)?;
-    documents.add_documents(json)?;
+    documents.extend_from_json(reader)?;
     documents.finish()?;
 
     Ok(writer.into_inner())
@@ -1065,16 +1062,7 @@ fn documents_from_json(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
 
 fn documents_from_csv(reader: impl io::Read) -> anyhow::Result<Vec<u8>> {
     let mut writer = Cursor::new(Vec::new());
-    let mut documents = milli::documents::DocumentBatchBuilder::new(&mut writer)?;
-
-    let iter = CSVDocumentDeserializer::from_reader(reader)?;
-
-    for doc in iter {
-        let doc = doc?;
-        documents.add_documents(doc)?;
-    }
-
-    documents.finish()?;
+    milli::documents::DocumentBatchBuilder::from_csv(reader, &mut writer)?.finish()?;
 
     Ok(writer.into_inner())
 }
diff --git a/milli/Cargo.toml b/milli/Cargo.toml
index bda83c2df..209c8b1f7 100644
--- a/milli/Cargo.toml
+++ b/milli/Cargo.toml
@@ -48,6 +48,7 @@ itertools = "0.10.0"
 # logging
 log = "0.4.14"
 logging_timer = "1.0.0"
+csv = "1.1.6"
 
 [dev-dependencies]
 big_s = "1.0.2"
diff --git a/milli/src/documents/builder.rs b/milli/src/documents/builder.rs
index ba1319eff..6ba890b79 100644
--- a/milli/src/documents/builder.rs
+++ b/milli/src/documents/builder.rs
@@ -1,16 +1,20 @@
+use std::collections::BTreeMap;
 use std::io;
+use std::io::{Cursor, Write};
 
 use byteorder::{BigEndian, WriteBytesExt};
-use serde::ser::Serialize;
+use serde::Deserializer;
+use serde_json::Value;
 
-use super::serde::DocumentSerializer;
+use super::serde::DocumentVisitor;
 use super::{ByteCounter, DocumentsBatchIndex, DocumentsMetadata, Error};
+use crate::FieldId;
 
 /// The `DocumentsBatchBuilder` provides a way to build a documents batch in the intermediary
 /// format used by milli.
 ///
 /// The writer used by the DocumentBatchBuilder can be read using a `DocumentBatchReader` to
-/// iterate other the documents.
+/// iterate over the documents.
 ///
 /// ## example:
 /// ```
@@ -18,43 +22,48 @@ use super::{ByteCounter, DocumentsBatchIndex, DocumentsMetadata, Error};
 /// use serde_json::json;
 /// use std::io::Cursor;
 ///
+/// let json = r##"{"id": 1, "name": "foo"}"##;
 /// let mut writer = Cursor::new(Vec::new());
 /// let mut builder = DocumentBatchBuilder::new(&mut writer).unwrap();
-/// builder.add_documents(json!({"id": 1, "name": "foo"})).unwrap();
+/// builder.extend_from_json(&mut json.as_bytes()).unwrap();
 /// builder.finish().unwrap();
 /// ```
 pub struct DocumentBatchBuilder<W> {
-    serializer: DocumentSerializer<W>,
+    inner: ByteCounter<W>,
+    index: DocumentsBatchIndex,
+    obkv_buffer: Vec<u8>,
+    value_buffer: Vec<u8>,
+    values: BTreeMap<FieldId, Value>,
+    count: usize,
 }
 
 impl<W: io::Write + io::Seek> DocumentBatchBuilder<W> {
     pub fn new(writer: W) -> Result<Self, Error> {
-        let index = DocumentsBatchIndex::new();
+        let index = DocumentsBatchIndex::default();
         let mut writer = ByteCounter::new(writer);
         // add space to write the offset of the metadata at the end of the writer
         writer.write_u64::<BigEndian>(0)?;
 
-        let serializer =
-            DocumentSerializer { writer, buffer: Vec::new(), index, count: 0, allow_seq: true };
-
-        Ok(Self { serializer })
+        Ok(Self {
+            inner: writer,
+            index,
+            obkv_buffer: Vec::new(),
+            value_buffer: Vec::new(),
+            values: BTreeMap::new(),
+            count: 0,
+        })
     }
 
     /// Returns the number of documents that have been written to the builder.
     pub fn len(&self) -> usize {
-        self.serializer.count
+        self.count
     }
 
     /// This method must be called after the document addition is terminated. It will put the
     /// metadata at the end of the file, and write the metadata offset at the beginning on the
     /// file.
     pub fn finish(self) -> Result<(), Error> {
-        let DocumentSerializer {
-            writer: ByteCounter { mut writer, count: offset },
-            index,
-            count,
-            ..
-        } = self.serializer;
+        let Self { inner: ByteCounter { mut writer, count: offset }, index, count, .. } = self;
 
         let meta = DocumentsMetadata { count, index };
 
@@ -68,13 +77,478 @@ impl<W: io::Write + io::Seek> DocumentBatchBuilder<W> {
         Ok(())
     }
 
-    /// Adds documents to the builder.
+    /// Extends the builder with json documents from a reader.
+    pub fn extend_from_json<R: io::Read>(&mut self, reader: R) -> Result<(), Error> {
+        let mut de = serde_json::Deserializer::from_reader(reader);
+
+        let mut visitor = DocumentVisitor {
+            inner: &mut self.inner,
+            index: &mut self.index,
+            obkv_buffer: &mut self.obkv_buffer,
+            value_buffer: &mut self.value_buffer,
+            values: &mut self.values,
+            count: &mut self.count,
+        };
+
+        de.deserialize_any(&mut visitor).map_err(Error::JsonError)?
+    }
+
+    /// Creates a builder from a reader of CSV documents.
     ///
-    /// The internal index is updated with the fields found
-    /// in the documents. Document must either be a map or a sequences of map, anything else will
-    /// fail.
-    pub fn add_documents<T: Serialize>(&mut self, document: T) -> Result<(), Error> {
-        document.serialize(&mut self.serializer)?;
-        Ok(())
+    /// Since all fields in a csv documents are guaranteed to be ordered, we are able to perform
+    /// optimisations, and extending from another CSV is not allowed.
+    pub fn from_csv<R: io::Read>(reader: R, writer: W) -> Result<Self, Error> {
+        let mut this = Self::new(writer)?;
+        // Ensure that this is the first and only addition made with this builder
+        debug_assert!(this.index.is_empty());
+
+        let mut records = csv::Reader::from_reader(reader);
+
+        let headers = records
+            .headers()?
+            .into_iter()
+            .map(parse_csv_header)
+            .map(|(k, t)| (this.index.insert(&k), t))
+            .collect::<BTreeMap<_, _>>();
+
+        for (i, record) in records.into_records().enumerate() {
+            let record = record?;
+            this.obkv_buffer.clear();
+            let mut writer = obkv::KvWriter::new(&mut this.obkv_buffer);
+            for (value, (fid, ty)) in record.into_iter().zip(headers.iter()) {
+                let value = match ty {
+                    AllowedType::Number => {
+                        value.parse::<f64>().map(Value::from).map_err(|error| {
+                            Error::ParseFloat {
+                                error,
+                                // +1 for the header offset.
+                                line: i + 1,
+                                value: value.to_string(),
+                            }
+                        })?
+                    }
+                    AllowedType::String => Value::String(value.to_string()),
+                };
+
+                this.value_buffer.clear();
+                serde_json::to_writer(Cursor::new(&mut this.value_buffer), &value)?;
+                writer.insert(*fid, &this.value_buffer)?;
+            }
+
+            this.inner.write_u32::<BigEndian>(this.obkv_buffer.len() as u32)?;
+            this.inner.write_all(&this.obkv_buffer)?;
+
+            this.count += 1;
+        }
+
+        Ok(this)
+    }
+}
+
+#[derive(Debug)]
+enum AllowedType {
+    String,
+    Number,
+}
+
+fn parse_csv_header(header: &str) -> (String, AllowedType) {
+    // if there are several separators we only split on the last one.
+    match header.rsplit_once(':') {
+        Some((field_name, field_type)) => match field_type {
+            "string" => (field_name.to_string(), AllowedType::String),
+            "number" => (field_name.to_string(), AllowedType::Number),
+            // if the pattern isn't reconized, we keep the whole field.
+            _otherwise => (header.to_string(), AllowedType::String),
+        },
+        None => (header.to_string(), AllowedType::String),
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use std::io::Cursor;
+
+    use serde_json::{json, Map};
+
+    use super::*;
+    use crate::documents::DocumentBatchReader;
+
+    fn obkv_to_value(obkv: &obkv::KvReader<FieldId>, index: &DocumentsBatchIndex) -> Value {
+        let mut map = Map::new();
+
+        for (fid, value) in obkv.iter() {
+            let field_name = index.name(fid).unwrap().clone();
+            let value: Value = serde_json::from_slice(value).unwrap();
+
+            map.insert(field_name, value);
+        }
+
+        Value::Object(map)
+    }
+
+    #[test]
+    fn add_single_documents_json() {
+        let mut cursor = Cursor::new(Vec::new());
+        let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
+
+        let json = serde_json::json!({
+            "id": 1,
+            "field": "hello!",
+        });
+
+        builder.extend_from_json(Cursor::new(serde_json::to_vec(&json).unwrap())).unwrap();
+
+        let json = serde_json::json!({
+            "blabla": false,
+            "field": "hello!",
+            "id": 1,
+        });
+
+        builder.extend_from_json(Cursor::new(serde_json::to_vec(&json).unwrap())).unwrap();
+
+        assert_eq!(builder.len(), 2);
+
+        builder.finish().unwrap();
+
+        cursor.set_position(0);
+
+        let mut reader = DocumentBatchReader::from_reader(cursor).unwrap();
+
+        let (index, document) = reader.next_document_with_index().unwrap().unwrap();
+        assert_eq!(index.len(), 3);
+        assert_eq!(document.iter().count(), 2);
+
+        let (index, document) = reader.next_document_with_index().unwrap().unwrap();
+        assert_eq!(index.len(), 3);
+        assert_eq!(document.iter().count(), 3);
+
+        assert!(reader.next_document_with_index().unwrap().is_none());
+    }
+
+    #[test]
+    fn add_documents_seq_json() {
+        let mut cursor = Cursor::new(Vec::new());
+        let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
+
+        let json = serde_json::json!([{
+            "id": 1,
+            "field": "hello!",
+        },{
+            "blabla": false,
+            "field": "hello!",
+            "id": 1,
+        }
+        ]);
+
+        builder.extend_from_json(Cursor::new(serde_json::to_vec(&json).unwrap())).unwrap();
+
+        assert_eq!(builder.len(), 2);
+
+        builder.finish().unwrap();
+
+        cursor.set_position(0);
+
+        let mut reader = DocumentBatchReader::from_reader(cursor).unwrap();
+
+        let (index, document) = reader.next_document_with_index().unwrap().unwrap();
+        assert_eq!(index.len(), 3);
+        assert_eq!(document.iter().count(), 2);
+
+        let (index, document) = reader.next_document_with_index().unwrap().unwrap();
+        assert_eq!(index.len(), 3);
+        assert_eq!(document.iter().count(), 3);
+
+        assert!(reader.next_document_with_index().unwrap().is_none());
+    }
+
+    #[test]
+    fn add_documents_csv() {
+        let mut cursor = Cursor::new(Vec::new());
+
+        let csv = "id:number,field:string\n1,hello!\n2,blabla";
+
+        let builder =
+            DocumentBatchBuilder::from_csv(Cursor::new(csv.as_bytes()), &mut cursor).unwrap();
+        builder.finish().unwrap();
+
+        cursor.set_position(0);
+
+        let mut reader = DocumentBatchReader::from_reader(cursor).unwrap();
+
+        let (index, document) = reader.next_document_with_index().unwrap().unwrap();
+        assert_eq!(index.len(), 2);
+        assert_eq!(document.iter().count(), 2);
+
+        let (_index, document) = reader.next_document_with_index().unwrap().unwrap();
+        assert_eq!(document.iter().count(), 2);
+
+        assert!(reader.next_document_with_index().unwrap().is_none());
+    }
+
+    #[test]
+    fn simple_csv_document() {
+        let documents = r#"city,country,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city": "Boston",
+                "country": "United States",
+                "pop": "4628910",
+            })
+        );
+
+        assert!(reader.next_document_with_index().unwrap().is_none());
+    }
+
+    #[test]
+    fn coma_in_field() {
+        let documents = r#"city,country,pop
+"Boston","United, States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city": "Boston",
+                "country": "United, States",
+                "pop": "4628910",
+            })
+        );
+    }
+
+    #[test]
+    fn quote_in_field() {
+        let documents = r#"city,country,pop
+"Boston","United"" States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city": "Boston",
+                "country": "United\" States",
+                "pop": "4628910",
+            })
+        );
+    }
+
+    #[test]
+    fn integer_in_field() {
+        let documents = r#"city,country,pop:number
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city": "Boston",
+                "country": "United States",
+                "pop": 4628910.0,
+            })
+        );
+    }
+
+    #[test]
+    fn float_in_field() {
+        let documents = r#"city,country,pop:number
+"Boston","United States","4628910.01""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city": "Boston",
+                "country": "United States",
+                "pop": 4628910.01,
+            })
+        );
+    }
+
+    #[test]
+    fn several_colon_in_header() {
+        let documents = r#"city:love:string,country:state,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city:love": "Boston",
+                "country:state": "United States",
+                "pop": "4628910",
+            })
+        );
+    }
+
+    #[test]
+    fn ending_by_colon_in_header() {
+        let documents = r#"city:,country,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city:": "Boston",
+                "country": "United States",
+                "pop": "4628910",
+            })
+        );
+    }
+
+    #[test]
+    fn starting_by_colon_in_header() {
+        let documents = r#":city,country,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                ":city": "Boston",
+                "country": "United States",
+                "pop": "4628910",
+            })
+        );
+    }
+
+    #[ignore]
+    #[test]
+    fn starting_by_colon_in_header2() {
+        let documents = r#":string,country,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+
+        assert!(reader.next_document_with_index().is_err());
+    }
+
+    #[test]
+    fn double_colon_in_header() {
+        let documents = r#"city::string,country,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf))
+            .unwrap()
+            .finish()
+            .unwrap();
+        let mut reader = DocumentBatchReader::from_reader(Cursor::new(buf)).unwrap();
+        let (index, doc) = reader.next_document_with_index().unwrap().unwrap();
+        let val = obkv_to_value(&doc, index);
+
+        assert_eq!(
+            val,
+            json!({
+                "city:": "Boston",
+                "country": "United States",
+                "pop": "4628910",
+            })
+        );
+    }
+
+    #[test]
+    fn bad_type_in_header() {
+        let documents = r#"city,country:number,pop
+"Boston","United States","4628910""#;
+
+        let mut buf = Vec::new();
+        assert!(
+            DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf)).is_err()
+        );
+    }
+
+    #[test]
+    fn bad_column_count1() {
+        let documents = r#"city,country,pop
+"Boston","United States","4628910", "too much""#;
+
+        let mut buf = Vec::new();
+        assert!(
+            DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf)).is_err()
+        );
+    }
+
+    #[test]
+    fn bad_column_count2() {
+        let documents = r#"city,country,pop
+"Boston","United States""#;
+
+        let mut buf = Vec::new();
+        assert!(
+            DocumentBatchBuilder::from_csv(documents.as_bytes(), Cursor::new(&mut buf)).is_err()
+        );
     }
 }
diff --git a/milli/src/documents/mod.rs b/milli/src/documents/mod.rs
index f79c210fe..14d97ee7d 100644
--- a/milli/src/documents/mod.rs
+++ b/milli/src/documents/mod.rs
@@ -7,7 +7,8 @@ mod builder;
 mod reader;
 mod serde;
 
-use std::{fmt, io};
+use std::fmt::{self, Debug};
+use std::io;
 
 use ::serde::{Deserialize, Serialize};
 use bimap::BiHashMap;
@@ -17,7 +18,38 @@ pub use reader::DocumentBatchReader;
 use crate::FieldId;
 
 /// A bidirectional map that links field ids to their name in a document batch.
-pub type DocumentsBatchIndex = BiHashMap<FieldId, String>;
+#[derive(Default, Debug, Serialize, Deserialize)]
+pub struct DocumentsBatchIndex(pub BiHashMap<FieldId, String>);
+
+impl DocumentsBatchIndex {
+    /// Insert the field in the map, or return it's field id if it doesn't already exists.
+    pub fn insert(&mut self, field: &str) -> FieldId {
+        match self.0.get_by_right(field) {
+            Some(field_id) => *field_id,
+            None => {
+                let field_id = self.0.len() as FieldId;
+                self.0.insert(field_id, field.to_string());
+                field_id
+            }
+        }
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.0.is_empty()
+    }
+
+    pub fn len(&self) -> usize {
+        self.0.len()
+    }
+
+    pub fn iter(&self) -> bimap::hash::Iter<FieldId, String> {
+        self.0.iter()
+    }
+
+    pub fn name(&self, id: FieldId) -> Option<&String> {
+        self.0.get_by_left(&id)
+    }
+}
 
 #[derive(Debug, Serialize, Deserialize)]
 struct DocumentsMetadata {
@@ -50,14 +82,22 @@ impl<W: io::Write> io::Write for ByteCounter<W> {
 
 #[derive(Debug)]
 pub enum Error {
+    ParseFloat { error: std::num::ParseFloatError, line: usize, value: String },
     InvalidDocumentFormat,
     Custom(String),
     JsonError(serde_json::Error),
+    CsvError(csv::Error),
     Serialize(bincode::Error),
     Io(io::Error),
     DocumentTooLarge,
 }
 
+impl From<csv::Error> for Error {
+    fn from(e: csv::Error) -> Self {
+        Self::CsvError(e)
+    }
+}
+
 impl From<io::Error> for Error {
     fn from(other: io::Error) -> Self {
         Self::Io(other)
@@ -70,15 +110,25 @@ impl From<bincode::Error> for Error {
     }
 }
 
+impl From<serde_json::Error> for Error {
+    fn from(other: serde_json::Error) -> Self {
+        Self::JsonError(other)
+    }
+}
+
 impl fmt::Display for Error {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match self {
+            Error::ParseFloat { error, line, value } => {
+                write!(f, "Error parsing number {:?} at line {}: {}", value, line, error)
+            }
             Error::Custom(s) => write!(f, "Unexpected serialization error: {}", s),
             Error::InvalidDocumentFormat => f.write_str("Invalid document addition format."),
             Error::JsonError(err) => write!(f, "Couldn't serialize document value: {}", err),
-            Error::Io(e) => e.fmt(f),
+            Error::Io(e) => write!(f, "{}", e),
             Error::DocumentTooLarge => f.write_str("Provided document is too large (>2Gib)"),
-            Error::Serialize(e) => e.fmt(f),
+            Error::Serialize(e) => write!(f, "{}", e),
+            Error::CsvError(e) => write!(f, "{}", e),
         }
     }
 }
@@ -92,7 +142,8 @@ macro_rules! documents {
         let documents = serde_json::json!($data);
         let mut writer = std::io::Cursor::new(Vec::new());
         let mut builder = crate::documents::DocumentBatchBuilder::new(&mut writer).unwrap();
-        builder.add_documents(documents).unwrap();
+        let documents = serde_json::to_vec(&documents).unwrap();
+        builder.extend_from_json(std::io::Cursor::new(documents)).unwrap();
         builder.finish().unwrap();
 
         writer.set_position(0);
@@ -103,6 +154,8 @@ macro_rules! documents {
 
 #[cfg(test)]
 mod test {
+    use std::io::Cursor;
+
     use serde_json::{json, Value};
 
     use super::*;
@@ -119,12 +172,14 @@ mod test {
             "bool": true
         });
 
+        let json = serde_json::to_vec(&json).unwrap();
+
         let mut v = Vec::new();
         let mut cursor = io::Cursor::new(&mut v);
 
         let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
 
-        builder.add_documents(json).unwrap();
+        builder.extend_from_json(Cursor::new(json)).unwrap();
 
         builder.finish().unwrap();
 
@@ -148,13 +203,16 @@ mod test {
             "toto": false,
         });
 
+        let doc1 = serde_json::to_vec(&doc1).unwrap();
+        let doc2 = serde_json::to_vec(&doc2).unwrap();
+
         let mut v = Vec::new();
         let mut cursor = io::Cursor::new(&mut v);
 
         let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
 
-        builder.add_documents(doc1).unwrap();
-        builder.add_documents(doc2).unwrap();
+        builder.extend_from_json(Cursor::new(doc1)).unwrap();
+        builder.extend_from_json(Cursor::new(doc2)).unwrap();
 
         builder.finish().unwrap();
 
@@ -177,12 +235,14 @@ mod test {
             { "tata": "hello" },
         ]);
 
+        let docs = serde_json::to_vec(&docs).unwrap();
+
         let mut v = Vec::new();
         let mut cursor = io::Cursor::new(&mut v);
 
         let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
 
-        builder.add_documents(docs).unwrap();
+        builder.extend_from_json(Cursor::new(docs)).unwrap();
 
         builder.finish().unwrap();
 
@@ -210,11 +270,13 @@ mod test {
             { "tata": "hello" },
         ]]);
 
-        assert!(builder.add_documents(docs).is_err());
+        let docs = serde_json::to_vec(&docs).unwrap();
+        assert!(builder.extend_from_json(Cursor::new(docs)).is_err());
 
         let docs = json!("hello");
+        let docs = serde_json::to_vec(&docs).unwrap();
 
-        assert!(builder.add_documents(docs).is_err());
+        assert!(builder.extend_from_json(Cursor::new(docs)).is_err());
     }
 
     #[test]
diff --git a/milli/src/documents/serde.rs b/milli/src/documents/serde.rs
index 036ec246a..d57bf1ffb 100644
--- a/milli/src/documents/serde.rs
+++ b/milli/src/documents/serde.rs
@@ -1,474 +1,134 @@
 use std::collections::BTreeMap;
-use std::convert::TryInto;
-use std::io::Cursor;
-use std::{fmt, io};
+use std::fmt;
+use std::io::{Cursor, Write};
 
-use byteorder::{BigEndian, WriteBytesExt};
-use obkv::KvWriter;
-use serde::ser::{Impossible, Serialize, SerializeMap, SerializeSeq, Serializer};
+use byteorder::WriteBytesExt;
+use serde::de::{DeserializeSeed, MapAccess, SeqAccess, Visitor};
+use serde::Deserialize;
 use serde_json::Value;
 
 use super::{ByteCounter, DocumentsBatchIndex, Error};
 use crate::FieldId;
 
-pub struct DocumentSerializer<W> {
-    pub writer: ByteCounter<W>,
-    pub buffer: Vec<u8>,
-    pub index: DocumentsBatchIndex,
-    pub count: usize,
-    pub allow_seq: bool,
-}
-
-impl<'a, W: io::Write> Serializer for &'a mut DocumentSerializer<W> {
-    type Ok = ();
-
-    type Error = Error;
-
-    type SerializeSeq = SeqSerializer<'a, W>;
-    type SerializeTuple = Impossible<(), Self::Error>;
-    type SerializeTupleStruct = Impossible<(), Self::Error>;
-    type SerializeTupleVariant = Impossible<(), Self::Error>;
-    type SerializeMap = MapSerializer<'a, &'a mut ByteCounter<W>>;
-    type SerializeStruct = Impossible<(), Self::Error>;
-    type SerializeStructVariant = Impossible<(), Self::Error>;
-    fn serialize_map(self, _len: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
-        self.buffer.clear();
-        let cursor = io::Cursor::new(&mut self.buffer);
-        self.count += 1;
-        let map_serializer = MapSerializer {
-            map: KvWriter::new(cursor),
-            index: &mut self.index,
-            writer: &mut self.writer,
-            mapped_documents: BTreeMap::new(),
-        };
-
-        Ok(map_serializer)
-    }
-
-    fn serialize_seq(self, _len: Option<usize>) -> Result<Self::SerializeSeq, Self::Error> {
-        if self.allow_seq {
-            // Only allow sequence of documents of depth 1.
-            self.allow_seq = false;
-            Ok(SeqSerializer { serializer: self })
-        } else {
-            Err(Error::InvalidDocumentFormat)
+macro_rules! tri {
+    ($e:expr) => {
+        match $e {
+            Ok(r) => r,
+            Err(e) => return Ok(Err(e.into())),
         }
-    }
+    };
+}
 
-    fn serialize_bool(self, _v: bool) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
+struct FieldIdResolver<'a>(&'a mut DocumentsBatchIndex);
 
-    fn serialize_i8(self, _v: i8) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
+impl<'a, 'de> DeserializeSeed<'de> for FieldIdResolver<'a> {
+    type Value = FieldId;
 
-    fn serialize_i16(self, _v: i16) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_i32(self, _v: i32) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_i64(self, _v: i64) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u8(self, _v: u8) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u16(self, _v: u16) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u32(self, _v: u32) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u64(self, _v: u64) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_f32(self, _v: f32) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_f64(self, _v: f64) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_char(self, _v: char) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_str(self, _v: &str) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_bytes(self, _v: &[u8]) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_none(self) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_some<T: ?Sized>(self, _value: &T) -> Result<Self::Ok, Self::Error>
+    fn deserialize<D>(self, deserializer: D) -> Result<Self::Value, D::Error>
     where
-        T: Serialize,
+        D: serde::Deserializer<'de>,
     {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_unit(self) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_unit_struct(self, _name: &'static str) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_unit_variant(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-    ) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_newtype_struct<T: ?Sized>(
-        self,
-        _name: &'static str,
-        _value: &T,
-    ) -> Result<Self::Ok, Self::Error>
-    where
-        T: Serialize,
-    {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_newtype_variant<T: ?Sized>(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-        _value: &T,
-    ) -> Result<Self::Ok, Self::Error>
-    where
-        T: Serialize,
-    {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_tuple(self, _len: usize) -> Result<Self::SerializeTuple, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_tuple_struct(
-        self,
-        _name: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeTupleStruct, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_tuple_variant(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeTupleVariant, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_struct(
-        self,
-        _name: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeStruct, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_struct_variant(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeStructVariant, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
+        deserializer.deserialize_str(self)
     }
 }
 
-pub struct SeqSerializer<'a, W> {
-    serializer: &'a mut DocumentSerializer<W>,
-}
+impl<'a, 'de> Visitor<'de> for FieldIdResolver<'a> {
+    type Value = FieldId;
 
-impl<'a, W: io::Write> SerializeSeq for SeqSerializer<'a, W> {
-    type Ok = ();
-    type Error = Error;
-
-    fn serialize_element<T: ?Sized>(&mut self, value: &T) -> Result<(), Self::Error>
+    fn visit_str<E>(self, v: &str) -> Result<Self::Value, E>
     where
-        T: Serialize,
+        E: serde::de::Error,
     {
-        value.serialize(&mut *self.serializer)?;
-        Ok(())
+        Ok(self.0.insert(v))
     }
 
-    fn end(self) -> Result<Self::Ok, Self::Error> {
-        Ok(())
+    fn expecting(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "a string")
     }
 }
 
-pub struct MapSerializer<'a, W> {
-    map: KvWriter<io::Cursor<&'a mut Vec<u8>>, FieldId>,
-    index: &'a mut DocumentsBatchIndex,
-    writer: W,
-    mapped_documents: BTreeMap<FieldId, Value>,
+struct ValueDeserializer;
+
+impl<'de> DeserializeSeed<'de> for ValueDeserializer {
+    type Value = serde_json::Value;
+
+    fn deserialize<D>(self, deserializer: D) -> Result<Self::Value, D::Error>
+    where
+        D: serde::Deserializer<'de>,
+    {
+        serde_json::Value::deserialize(deserializer)
+    }
 }
 
-/// This implementation of SerializeMap uses serilialize_entry instead of seriliaze_key and
-/// serialize_value, therefore these to methods remain unimplemented.
-impl<'a, W: io::Write> SerializeMap for MapSerializer<'a, W> {
-    type Ok = ();
-    type Error = Error;
+pub struct DocumentVisitor<'a, W> {
+    pub inner: &'a mut ByteCounter<W>,
+    pub index: &'a mut DocumentsBatchIndex,
+    pub obkv_buffer: &'a mut Vec<u8>,
+    pub value_buffer: &'a mut Vec<u8>,
+    pub values: &'a mut BTreeMap<FieldId, Value>,
+    pub count: &'a mut usize,
+}
 
-    fn serialize_key<T: ?Sized + Serialize>(&mut self, _key: &T) -> Result<(), Self::Error> {
-        unreachable!()
-    }
+impl<'a, 'de, W: Write> Visitor<'de> for &mut DocumentVisitor<'a, W> {
+    /// This Visitor value is nothing, since it write the value to a file.
+    type Value = Result<(), Error>;
 
-    fn serialize_value<T: ?Sized>(&mut self, _value: &T) -> Result<(), Self::Error> {
-        unreachable!()
-    }
-
-    fn end(mut self) -> Result<Self::Ok, Self::Error> {
-        let mut buf = Vec::new();
-        for (key, value) in self.mapped_documents {
-            buf.clear();
-            let mut cursor = Cursor::new(&mut buf);
-            serde_json::to_writer(&mut cursor, &value).map_err(Error::JsonError)?;
-            self.map.insert(key, cursor.into_inner()).map_err(Error::Io)?;
+    fn visit_seq<A>(self, mut seq: A) -> Result<Self::Value, A::Error>
+    where
+        A: SeqAccess<'de>,
+    {
+        while let Some(v) = seq.next_element_seed(&mut *self)? {
+            tri!(v)
         }
 
-        let data = self.map.into_inner().map_err(Error::Io)?.into_inner();
-        let data_len: u32 = data.len().try_into().map_err(|_| Error::DocumentTooLarge)?;
-
-        self.writer.write_u32::<BigEndian>(data_len).map_err(Error::Io)?;
-        self.writer.write_all(&data).map_err(Error::Io)?;
-
-        Ok(())
+        Ok(Ok(()))
     }
 
-    fn serialize_entry<K: ?Sized, V: ?Sized>(
-        &mut self,
-        key: &K,
-        value: &V,
-    ) -> Result<(), Self::Error>
+    fn visit_map<A>(self, mut map: A) -> Result<Self::Value, A::Error>
     where
-        K: Serialize,
-        V: Serialize,
+        A: MapAccess<'de>,
     {
-        let field_serializer = FieldSerializer { index: &mut self.index };
-        let field_id: FieldId = key.serialize(field_serializer)?;
+        while let Some((key, value)) =
+            map.next_entry_seed(FieldIdResolver(&mut *self.index), ValueDeserializer)?
+        {
+            self.values.insert(key, value);
+        }
 
-        let value = serde_json::to_value(value).map_err(Error::JsonError)?;
+        self.obkv_buffer.clear();
+        let mut obkv = obkv::KvWriter::new(Cursor::new(&mut *self.obkv_buffer));
+        for (key, value) in self.values.iter() {
+            self.value_buffer.clear();
+            // This is guaranteed to work
+            tri!(serde_json::to_writer(Cursor::new(&mut *self.value_buffer), value));
+            tri!(obkv.insert(*key, &self.value_buffer));
+        }
 
-        self.mapped_documents.insert(field_id, value);
+        let reader = tri!(obkv.into_inner()).into_inner();
 
-        Ok(())
+        tri!(self.inner.write_u32::<byteorder::BigEndian>(reader.len() as u32));
+        tri!(self.inner.write_all(reader));
+
+        *self.count += 1;
+        self.values.clear();
+
+        Ok(Ok(()))
+    }
+
+    fn expecting(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "a documents, or a sequence of documents.")
     }
 }
 
-struct FieldSerializer<'a> {
-    index: &'a mut DocumentsBatchIndex,
-}
+impl<'a, 'de, W> DeserializeSeed<'de> for &mut DocumentVisitor<'a, W>
+where
+    W: Write,
+{
+    type Value = Result<(), Error>;
 
-impl<'a> serde::Serializer for FieldSerializer<'a> {
-    type Ok = FieldId;
-
-    type Error = Error;
-
-    type SerializeSeq = Impossible<FieldId, Self::Error>;
-    type SerializeTuple = Impossible<FieldId, Self::Error>;
-    type SerializeTupleStruct = Impossible<FieldId, Self::Error>;
-    type SerializeTupleVariant = Impossible<FieldId, Self::Error>;
-    type SerializeMap = Impossible<FieldId, Self::Error>;
-    type SerializeStruct = Impossible<FieldId, Self::Error>;
-    type SerializeStructVariant = Impossible<FieldId, Self::Error>;
-
-    fn serialize_str(self, ws: &str) -> Result<Self::Ok, Self::Error> {
-        let field_id = match self.index.get_by_right(ws) {
-            Some(field_id) => *field_id,
-            None => {
-                let field_id = self.index.len() as FieldId;
-                self.index.insert(field_id, ws.to_string());
-                field_id
-            }
-        };
-
-        Ok(field_id)
-    }
-
-    fn serialize_bool(self, _v: bool) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_i8(self, _v: i8) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_i16(self, _v: i16) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_i32(self, _v: i32) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_i64(self, _v: i64) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u8(self, _v: u8) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u16(self, _v: u16) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u32(self, _v: u32) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_u64(self, _v: u64) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_f32(self, _v: f32) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_f64(self, _v: f64) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_char(self, _v: char) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_bytes(self, _v: &[u8]) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_none(self) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_some<T: ?Sized>(self, _value: &T) -> Result<Self::Ok, Self::Error>
+    fn deserialize<D>(self, deserializer: D) -> Result<Self::Value, D::Error>
     where
-        T: Serialize,
+        D: serde::Deserializer<'de>,
     {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_unit(self) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_unit_struct(self, _name: &'static str) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_unit_variant(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-    ) -> Result<Self::Ok, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_newtype_struct<T: ?Sized>(
-        self,
-        _name: &'static str,
-        _value: &T,
-    ) -> Result<Self::Ok, Self::Error>
-    where
-        T: Serialize,
-    {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_newtype_variant<T: ?Sized>(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-        _value: &T,
-    ) -> Result<Self::Ok, Self::Error>
-    where
-        T: Serialize,
-    {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_seq(self, _len: Option<usize>) -> Result<Self::SerializeSeq, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_tuple(self, _len: usize) -> Result<Self::SerializeTuple, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_tuple_struct(
-        self,
-        _name: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeTupleStruct, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_tuple_variant(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeTupleVariant, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_map(self, _len: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_struct(
-        self,
-        _name: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeStruct, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-
-    fn serialize_struct_variant(
-        self,
-        _name: &'static str,
-        _variant_index: u32,
-        _variant: &'static str,
-        _len: usize,
-    ) -> Result<Self::SerializeStructVariant, Self::Error> {
-        Err(Error::InvalidDocumentFormat)
-    }
-}
-
-impl serde::ser::Error for Error {
-    fn custom<T: fmt::Display>(msg: T) -> Self {
-        Error::Custom(msg.to_string())
+        deserializer.deserialize_map(self)
     }
 }
diff --git a/milli/src/index.rs b/milli/src/index.rs
index 6ce693fbe..fe89fe734 100644
--- a/milli/src/index.rs
+++ b/milli/src/index.rs
@@ -953,6 +953,7 @@ pub(crate) mod tests {
             { "id": 1, "name": "kevin", "has_dog": true },
             { "id": 2, "name": "bob" }
         ]);
+
         let mut wtxn = index.write_txn().unwrap();
         let builder = IndexDocuments::new(&mut wtxn, &index, 0);
         builder.execute(content, |_, _| ()).unwrap();
diff --git a/milli/src/search/distinct/mod.rs b/milli/src/search/distinct/mod.rs
index deb51a053..11f6379e3 100644
--- a/milli/src/search/distinct/mod.rs
+++ b/milli/src/search/distinct/mod.rs
@@ -68,7 +68,9 @@ mod test {
                 "txts": sample_txts[..(rng.gen_range(0..3))],
                 "cat-ints": sample_ints[..(rng.gen_range(0..3))],
             });
-            builder.add_documents(doc).unwrap();
+
+            let doc = Cursor::new(serde_json::to_vec(&doc).unwrap());
+            builder.extend_from_json(doc).unwrap();
         }
 
         builder.finish().unwrap();
diff --git a/milli/src/update/index_documents/mod.rs b/milli/src/update/index_documents/mod.rs
index 92bcab0e9..440546b10 100644
--- a/milli/src/update/index_documents/mod.rs
+++ b/milli/src/update/index_documents/mod.rs
@@ -877,7 +877,8 @@ mod tests {
         let mut cursor = Cursor::new(Vec::new());
 
         let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
-        builder.add_documents(big_object).unwrap();
+        let big_object = Cursor::new(serde_json::to_vec(&big_object).unwrap());
+        builder.extend_from_json(big_object).unwrap();
         builder.finish().unwrap();
         cursor.set_position(0);
         let content = DocumentBatchReader::from_reader(cursor).unwrap();
@@ -905,8 +906,9 @@ mod tests {
 
         let mut cursor = Cursor::new(Vec::new());
 
+        let big_object = serde_json::to_string(&big_object).unwrap();
         let mut builder = DocumentBatchBuilder::new(&mut cursor).unwrap();
-        builder.add_documents(big_object).unwrap();
+        builder.extend_from_json(&mut big_object.as_bytes()).unwrap();
         builder.finish().unwrap();
         cursor.set_position(0);
         let content = DocumentBatchReader::from_reader(cursor).unwrap();
diff --git a/milli/src/update/index_documents/transform.rs b/milli/src/update/index_documents/transform.rs
index c0c88abed..08aa72d35 100644
--- a/milli/src/update/index_documents/transform.rs
+++ b/milli/src/update/index_documents/transform.rs
@@ -75,7 +75,7 @@ fn create_fields_mapping(
         .collect()
 }
 
-fn find_primary_key(index: &bimap::BiHashMap<u16, String>) -> Option<&str> {
+fn find_primary_key(index: &DocumentsBatchIndex) -> Option<&str> {
     index
         .iter()
         .sorted_by_key(|(k, _)| *k)
@@ -179,7 +179,7 @@ impl Transform<'_, '_> {
                         if !self.autogenerate_docids {
                             let mut json = Map::new();
                             for (key, value) in document.iter() {
-                                let key = addition_index.get_by_left(&key).cloned();
+                                let key = addition_index.name(key).cloned();
                                 let value = serde_json::from_slice::<Value>(&value).ok();
 
                                 if let Some((k, v)) = key.zip(value) {
@@ -544,6 +544,7 @@ mod test {
     mod primary_key_inference {
         use bimap::BiHashMap;
 
+        use crate::documents::DocumentsBatchIndex;
         use crate::update::index_documents::transform::find_primary_key;
 
         #[test]
@@ -557,7 +558,7 @@ mod test {
                 map.insert(4, "fakeId".to_string());
                 map.insert(0, "realId".to_string());
 
-                assert_eq!(find_primary_key(&map), Some("realId"));
+                assert_eq!(find_primary_key(&DocumentsBatchIndex(map)), Some("realId"));
             }
         }
     }
diff --git a/milli/tests/search/mod.rs b/milli/tests/search/mod.rs
index cda0da617..e8fb3fdfa 100644
--- a/milli/tests/search/mod.rs
+++ b/milli/tests/search/mod.rs
@@ -61,9 +61,12 @@ pub fn setup_search_index_with_criteria(criteria: &[Criterion]) -> Index {
     let mut cursor = Cursor::new(Vec::new());
     let mut documents_builder = DocumentBatchBuilder::new(&mut cursor).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();
+        let doc = Cursor::new(serde_json::to_vec(&doc.unwrap()).unwrap());
+        documents_builder.extend_from_json(doc).unwrap();
     }
+
     documents_builder.finish().unwrap();
 
     cursor.set_position(0);
diff --git a/milli/tests/search/query_criteria.rs b/milli/tests/search/query_criteria.rs
index f3b04c4fa..e5dde049c 100644
--- a/milli/tests/search/query_criteria.rs
+++ b/milli/tests/search/query_criteria.rs
@@ -409,7 +409,8 @@ fn criteria_ascdesc() {
             "age": age,
         });
 
-        batch_builder.add_documents(json).unwrap();
+        let json = Cursor::new(serde_json::to_vec(&json).unwrap());
+        batch_builder.extend_from_json(json).unwrap();
     });
 
     batch_builder.finish().unwrap();