diff --git a/index-server/README.md b/index-server/README.md
new file mode 100644
index 0000000..cbc25e7
--- /dev/null
+++ b/index-server/README.md
@@ -0,0 +1,24 @@
+# How to build with your own data
+1. Prepare your corpus and split it into small jsonl files with the following structure (like Pile, but we only need the `text` key), and put them into a folder named `splits`.
+```
+{
+  'meta': {'pile_set_name': 'Pile-CC'},
+  'text': 'It is done, and submitted. You can play “Survival of the Tastiest” on Android, and on the web. Playing on...'
+}
+```
+2. Build metadata
+```bash
+python build_split_meta.py 
+```
+3. Build shard
+```bash
+python build_shard.py
+```
+4. Build database
+```bash
+build_db.py
+```
+5. Build index. The Faiss parameter is hardcoded now. Choosing an index is like a kind of compute resource and recall tradeoff. See [Faiss wiki](https://github.com/facebookresearch/faiss/wiki/The-index-factory) for more details.
+```bash
+build_index.py
+```