diff --git a/PROGRAM_FLOW.md b/PROGRAM_FLOW.md
new file mode 100644
index 0000000..81cadd6
--- /dev/null
+++ b/PROGRAM_FLOW.md
@@ -0,0 +1,45 @@
+# Program Flow Overview
+
+This document outlines the high-level workflow of the project and shows how the main modules interact. The diagram summarizes the preprocessing, training and generation stages.
+
+```mermaid
+flowchart TD
+    subgraph Preprocessing
+        A1([Raw text files]) --> A2(train_bpe.py)
+        A2 --> A3[Trained BytePairTokenizer]
+        A1 --> A4(tokenize_dataset.py)
+        A4 --> A5[Tokenized dataset]
+    end
+
+    subgraph Training
+        B1[TokenIDDataset / TokenIDSubset] --> B2[DataLoader]
+        B2 --> B3[Trainer]
+        B3 --> B4[GPT Model]
+        B3 --> B5[Checkpoint]
+    end
+
+    subgraph Generation
+        C1[Checkpoint] --> C2[GPT Model]
+        C2 --> C3[Sequencer]
+        C3 --> C4(Output text)
+    end
+
+    A3 -->|used by| B1
+    A5 -->|used by| B1
+    B5 -->|loads| C2
+```
+
+## Stage descriptions
+
+- **Preprocessing**
+  - `train_bpe.py` trains a `BytePairTokenizer` from raw text files.
+  - `tokenize_dataset.py` uses the trained tokenizer to convert the corpus into sequences of token IDs.
+
+- **Training**
+  - `TokenIDDataset`/`TokenIDSubset` create iterable datasets from the tokenized files.
+  - `DataLoader` feeds batches to the `Trainer`.
+  - `Trainer` runs the training loop on the `GPT` model, saving checkpoints for later use.
+
+- **Generation**
+  - The saved checkpoint is loaded into the `GPT` model.
+  - The `Sequencer` performs autoregressive decoding to produce the final text output.
diff --git a/README.md b/README.md
index d75bc51..d26fc81 100644
--- a/README.md
+++ b/README.md
@@ -21,7 +21,7 @@ The model's performance was particularly notable in tasks requiring contextual u
 ## Implementation 🛠️
 This repository provides implementation details and resources for GPT-1. Users can utilize this model for various NLP tasks, adapting it to specific requirements and datasets.
 
-Additional documentation is available in [MODEL.md](model/MODEL.md) and [DATASET.md](model/DATASET.md).
+Additional documentation is available in [MODEL.md](model/MODEL.md), [DATASET.md](model/DATASET.md), and [PROGRAM_FLOW.md](PROGRAM_FLOW.md).
 
 ### Getting Started ⚡
 See [SETUP.md](SETUP.md) for environment setup and training instructions. Once dependencies are installed you can pretrain the model with: