diff --git a/patchwork/steps/FilterBySimilarity/README.md b/patchwork/steps/FilterBySimilarity/README.md
new file mode 100644
index 000000000..308d2ba67
--- /dev/null
+++ b/patchwork/steps/FilterBySimilarity/README.md
@@ -0,0 +1,20 @@
+# Code Documentation
+
+## Inputs
+### File: patchwork/steps/FilterBySimilarity/typed.py
+- Two typed dictionary classes are defined: `FilterBySimilarityRequiredInputs` and `FilterBySimilarityInputs`.
+- `FilterBySimilarityRequiredInputs` contains keys `list` and `keywords`.
+- `FilterBySimilarityInputs` extends `FilterBySimilarityRequiredInputs` with additional keys `keys` and `top_k`.
+
+## Outputs
+### File: patchwork/steps/FilterBySimilarity/FilterBySimilarity.py
+- A class `FilterBySimilarity` is defined as a subclass of `Step` with specific input and output classes.
+- The class takes a list of items, keywords, optional keys, and top-k parameter as inputs.
+- The `run` method uses TF-IDF vectorization and cosine similarity to filter items based on similarity to keywords.
+- The output is a dictionary with a list of items sorted by their average similarity score to the keywords, limited by `top_k`.
+
+## Description
+- The code consists of a step in a system designed to filter items by their similarity to given keywords.
+- It utilizes TF-IDF vectorization and cosine similarity for this task.
+- The main logic is encapsulated in the `FilterBySimilarity` class.
+- Input parameters such as list of items, keywords, keys, and top-k are crucial for determining the filtering process.
\ No newline at end of file