Add how to update binary files into /data bucket

README.md

@@ -24,8 +24,9 @@ A complete **Retrieval-Augmented Generation (RAG)** system deployed as a Hugging
 3. [Step-by-Step Explanation](#step-by-step-explanation)
 4. [API Endpoints](#api-endpoints)
 5. [Setup & Deployment](#setup--deployment)
-6. [Configuration](#configuration)
-7. [How It Works (Detailed)](#how-it-works-detailed)
+6. [Adding Binary Files to the HF Space](#adding-binary-files-to-the-hf-space)
+7. [Configuration](#configuration)
+8. [How It Works (Detailed)](#how-it-works-detailed)
 
 ---
 
@@ -333,6 +334,94 @@ curl -X POST http://localhost:7860/ingest \
 
 ---
 
+## Adding Binary Files to the HF Space
+
+Large or binary files (e.g., pre-built ChromaDB databases, PDFs, images, model weights) cannot be pushed with regular `git push` because Hugging Face repositories enforce a **10 MB file-size limit** for standard Git. Use one of the methods below to make them available to your Space at runtime.
+
+### Method 1: Git LFS (recommended for files < 5 GB)
+
+Hugging Face repos support [Git Large File Storage (LFS)](https://git-lfs.com/). Files tracked by LFS are stored in a dedicated bucket and downloaded transparently when the Space starts.
+
+```bash
+# 1. Install Git LFS (once per machine)
+git lfs install
+
+# 2. Track the binary patterns you need
+git lfs track "*.pdf"
+git lfs track "*.sqlite3"
+git lfs track "chroma_db/**"
+git lfs track "DataSet/**"
+
+# This updates .gitattributes — commit it
+git add .gitattributes
+git commit -m "chore: track binary files with Git LFS"
+
+# 3. Add and push the binary files
+git add chroma_db/ DataSet/
+git commit -m "feat: add pre-built vector store and datasets"
+git push
+```
+
+> **Verify:** After pushing, open your Space repo on huggingface.co and click on an LFS file — it should show `Stored with Git LFS` instead of raw content.
+
+### Method 2: Upload via the `huggingface_hub` Python SDK
+
+Use this when you want to upload files programmatically (e.g., from a CI pipeline or a notebook) without cloning the full repo.
+
+```python
+from huggingface_hub import HfApi
+
+api = HfApi()
+
+# Upload a single file
+api.upload_file(
+    path_or_fileobj="./chroma_db/chroma.sqlite3",
+    path_in_repo="chroma_db/chroma.sqlite3",
+    repo_id="YOUR_USERNAME/Example-App-Hackathon-Gustave-Eiffel-2026",
+    repo_type="space",
+)
+
+# Upload an entire folder
+api.upload_folder(
+    folder_path="./DataSet",
+    path_in_repo="DataSet",
+    repo_id="YOUR_USERNAME/Example-App-Hackathon-Gustave-Eiffel-2026",
+    repo_type="space",
+)
+```
+
+> **Authentication:** Set `HF_TOKEN` in your environment or call `huggingface_hub.login()` first. The token needs **write** access to the Space.
+
+### Method 3: Hugging Face Hub UI (quick one-off uploads)
+
+1. Open your Space repository on huggingface.co
+2. Navigate to **Files and versions**
+3. Click **Add file → Upload files**
+4. Drag-and-drop your binary files (up to 50 GB per file via the UI)
+5. Commit directly to `main`
+
+### Accessing Uploaded Files in Your Application
+
+Once binary files are in the repo (via any method above), they are available at their relative path inside the Space container:
+
+```python
+import chromadb
+
+# Files land at the repo root, same relative path you committed them at
+client = chromadb.PersistentClient(path="./chroma_db")
+```
+
+No extra download logic is needed — Hugging Face clones the full repo (including LFS objects) into the Space container on startup.
+
+### Tips
+
+- **`.gitattributes`** is the source of truth for LFS tracking. Verify your patterns are listed there before pushing.
+- For files **> 5 GB**, prefer uploading via the Hub UI or the SDK (Git LFS CLI may time out on slow connections).
+- If you update a binary file, push the new version the same way — LFS handles versioning automatically.
+- To exclude files from the Space build (e.g., test data not needed at runtime), add them to `.gitignore` instead.
+
+---
+
 ## Configuration
 
 | Variable | Default | Description |