diff --git a/README.md b/README.md
index 478a1de..3024b24 100644
--- a/README.md
+++ b/README.md
@@ -436,6 +436,157 @@ Process repeats until successful or max attempts reached
 
 ---
 
+## 📊 Adding to the Vector Database
+
+The system uses vector embeddings to find similar projects and error examples, which helps improve code generation quality. Here's how to add your own examples:
+
+### 🔧 Creating Vector Collections
+
+First, you need to create the necessary collections in Qdrant using these curl commands:
+
+```bash
+# Create project_examples collection with 1536 dimensions (default)
+curl -X PUT "http://localhost:6333/collections/project_examples" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "vectors": {
+      "size": 1536,
+      "distance": "Cosine"
+    }
+  }'
+
+# Create error_examples collection with 1536 dimensions (default)
+curl -X PUT "http://localhost:6333/collections/error_examples" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "vectors": {
+      "size": 1536,
+      "distance": "Cosine"
+    }
+  }'
+```
+Note: If you've configured a different embedding size via ```LLM_EMBED_SIZE``` environment variable, replace 1536 with that value.
+
+### Method 1: Using Python API Directly
+
+#### For Project Examples
+```python
+from app.llm_client import LlamaEdgeClient
+from app.vector_store import QdrantStore
+
+# Initialize the components
+llm_client = LlamaEdgeClient()
+vector_store = QdrantStore()
+
+# Ensure collection exists
+vector_store.create_collection("project_examples")
+
+# 1. Prepare your data
+project_data = {
+    "query": "A command-line calculator in Rust",
+    "example": "Your full project example with code here...",
+    "project_files": {
+        "src/main.rs": "fn main() {\n    println!(\"Hello, calculator!\");\n}",
+        "Cargo.toml": "[package]\nname = \"calculator\"\nversion = \"0.1.0\"\nedition = \"2021\"\n\n[dependencies]"
+    }
+}
+
+# 2. Get embedding for the query text
+embedding = llm_client.get_embeddings([project_data["query"]])[0]
+
+# 3. Add to vector database
+vector_store.add_item(
+    collection_name="project_examples",
+    vector=embedding,
+    item=project_data
+)
+```
+
+For Error Examples:
+```python
+from app.llm_client import LlamaEdgeClient
+from app.vector_store import QdrantStore
+
+# Initialize the components
+llm_client = LlamaEdgeClient()
+vector_store = QdrantStore()
+
+# Ensure collection exists
+vector_store.create_collection("error_examples")
+
+# 1. Prepare your error data
+error_data = {
+    "error": "error[E0502]: cannot borrow `*self` as mutable because it is also borrowed as immutable",
+    "solution": "Ensure mutable and immutable borrows don't overlap by using separate scopes",
+    "context": "This error occurs when you try to borrow a value mutably while an immutable borrow exists",
+    "example": "// Before (error)\nfn main() {\n    let mut v = vec![1, 2, 3];\n    let first = &v[0];\n    v.push(4); // Error: cannot borrow `v` as mutable\n    println!(\"{}\", first);\n}\n\n// After (fixed)\nfn main() {\n    let mut v = vec![1, 2, 3];\n    {\n        let first = &v[0];\n        println!(\"{}\", first);\n    } // immutable borrow ends here\n    v.push(4); // Now it's safe to borrow mutably\n}"
+}
+
+# 2. Get embedding for the error message
+embedding = llm_client.get_embeddings([error_data["error"]])[0]
+
+# 3. Add to vector database
+vector_store.add_item(
+    collection_name="error_examples",
+    vector=embedding,
+    item=error_data
+)
+```
+
+### Method 2: Adding Multiple Examples from JSON Files
+Place JSON files in the appropriate directories:
+
+Project examples: ```project_examples```
+Error examples: ```error_examples```
+Format for project examples (with optional project_files field):
+```json
+{
+  "query": "Description of the project",
+  "example": "Full example code or description",
+  "project_files": {
+    "src/main.rs": "// File content here",
+    "Cargo.toml": "// File content here"
+  }
+}
+```
+Format for error examples:
+```
+{
+  "error": "Rust compiler error message",
+  "solution": "How to fix the error",
+  "context": "Additional explanation (optional)",
+  "example": "// Code example showing the fix (optional)"
+}
+```
+Then run the data loading script:
+```
+python -c "from app.load_data import load_project_examples, load_error_examples; load_project_examples(); load_error_examples()"
+```
+
+### Method 3: Using the ```parse_and_save_qna.py``` Script
+For bulk importing from a Q&A format text file:
+
+Place your Q&A pairs in a text file with format similar to ```QnA_pair.txt```
+Modify the ```parse_and_save_qna.py``` script to point to your file
+Run the script:
+```
+python parse_and_save_qna.py
+```
+
+## ⚙️ Environment Variables for Vector Search
+The SKIP_VECTOR_SEARCH environment variable controls whether the system uses vector search:
+
+```SKIP_VECTOR_SEARCH```=true - Disables vector search functionality
+```SKIP_VECTOR_SEARCH```=false (or not set) - Enables vector search
+In your current .env file, you have:
+```
+SKIP_VECTOR_SEARCH=true
+```
+This means vector search is currently disabled. To enable it:
+- Change this value to false or remove the line completely
+- Ensure you have a running Qdrant instance (via Docker Compose or standalone)
+- Create the collections as shown above
+
 ## 🤝 Contributing
 Contributions are welcome! This project uses the Developer Certificate of Origin (DCO) to certify that contributors have the right to submit their code. Follow these steps:
 
@@ -458,3 +609,6 @@ This certifies that you wrote or have the right to submit the code you're contri
 ## 📜 License
 Licensed under [GPLv3](https://www.gnu.org/licenses/gpl-3.0.en.html).
 
+
+
+