Add command-line option for selecting embeddings backend (Oracle DB or ChromaDB). Oracle DB is now the default, with ChromaDB as fallback. Update documentation.

Author: jasperan

agentic_rag/README.md

@@ -358,7 +358,7 @@ The system consists of several key components:
 1. **PDF Processor**: we use `docling` to extract and chunk text from PDF documents
 2. **Web Processor**: we use `trafilatura` to extract and chunk text from websites
 3. **GitHub Repository Processor**: we use `gitingest` to extract and chunk text from repositories
-4. **Vector Store**: Manages document embeddings and similarity search using `ChromaDB` and `Oracle Database 23ai`
+4. **Vector Store**: Manages document embeddings and similarity search using `Oracle Database 23ai` (default) or `ChromaDB` (fallback)
 5. **RAG Agent**: Makes intelligent decisions about query routing and response generation
    - OpenAI Agent: Uses `gpt-4-turbo-preview` for high-quality responses, but requires an OpenAI API key
    - Local Agent: Uses `Mistral-7B` as an open-source alternative
@@ -373,6 +373,50 @@ The RAG Agent flow is the following:
 4. If no PDF context is found OR if it's a general knowledge query, use the pre-trained LLM directly
 5. Fall back to a "no information" response only in edge cases.
 
+## Annex: Command Line Usage
+
+You can run the system from the command line using:
+
+```bash
+python local_rag_agent.py --query "Your question here" [options]
+```
+
+### Command Line Arguments
+
+| Argument | Description | Default |
+| --- | --- | --- |
+| `--query` | The query to process | *Required* |
+| `--embeddings` | Select embeddings backend (`oracle` or `chromadb`) | `oracle` |
+| `--model` | Model to use for inference | `mistralai/Mistral-7B-Instruct-v0.2` |
+| `--collection` | Collection to query (PDF, Repository, Web, General) | Auto-determined |
+| `--use-cot` | Enable Chain of Thought reasoning | `False` |
+| `--store-path` | Path to ChromaDB store (if using ChromaDB) | `embeddings` |
+| `--skip-analysis` | Skip query analysis step | `False` |
+| `--verbose` | Show full content of sources | `False` |
+| `--quiet` | Disable verbose logging | `False` |
+
+### Examples
+
+Query using Oracle DB (default):
+```bash
+python local_rag_agent.py --query "How does vector search work?"
+```
+
+Force using ChromaDB:
+```bash
+python local_rag_agent.py --query "How does vector search work?" --embeddings chromadb
+```
+
+Query with Chain of Thought reasoning:
+```bash
+python local_rag_agent.py --query "Explain the difference between RAG and fine-tuning" --use-cot
+```
+
+Query a specific collection:
+```bash
+python local_rag_agent.py --query "How to implement a queue?" --collection "Repository Collection"
+```
+
 ## Contributing
 
 This project is open source. Please submit your contributions by forking this repository and submitting a pull request! Oracle appreciates any contributions that are made by the open source community.

agentic_rag/local_rag_agent.py

@@ -510,6 +510,8 @@ def main():
                         help="Specify which collection to query")
     parser.add_argument("--skip-analysis", action="store_true", help="Skip query analysis step")
     parser.add_argument("--verbose", action="store_true", help="Show full content of sources")
+    parser.add_argument("--embeddings", choices=["oracle", "chromadb"], default="oracle", 
+                        help="Select embeddings backend (default: oracle)")
 
     args = parser.parse_args()
 
@@ -523,15 +525,37 @@ def main():
     print("=" * 50)
 
     try:
-        logger.info(f"Initializing vector store from: {args.store_path}")
-        store = VectorStore(persist_directory=args.store_path)
+        # Determine which vector store to use based on args.embeddings
+        if args.embeddings == "oracle" and ORACLE_DB_AVAILABLE:
+            try:
+                logger.info("Initializing Oracle DB vector store")
+                store = OraDBVectorStore()
+                print("✓ Using Oracle DB for vector storage")
+            except Exception as e:
+                logger.warning(f"Failed to initialize Oracle DB: {str(e)}")
+                logger.info(f"Falling back to ChromaDB from: {args.store_path}")
+                store = VectorStore(persist_directory=args.store_path)
+                print("⚠ Oracle DB initialization failed, using ChromaDB instead")
+        else:
+            if args.embeddings == "oracle" and not ORACLE_DB_AVAILABLE:
+                logger.warning("Oracle DB support not available")
+                print("⚠ Oracle DB support not available (missing dependencies)")
+                
+            logger.info(f"Initializing ChromaDB vector store from: {args.store_path}")
+            store = VectorStore(persist_directory=args.store_path)
+            print("✓ Using ChromaDB for vector storage")
+        
         logger.info("Initializing local RAG agent...")
+        # Set use_oracle_db based on the actual store type
+        use_oracle_db = args.embeddings == "oracle" and isinstance(store, OraDBVectorStore)
+        
         agent = LocalRAGAgent(
             store, 
             model_name=args.model, 
             use_cot=args.use_cot, 
             collection=args.collection,
-            skip_analysis=args.skip_analysis
+            skip_analysis=args.skip_analysis,
+            use_oracle_db=use_oracle_db
         )
 
         print(f"\nProcessing query: {args.query}")