anminhhung/RAG_tool

Base RAG tool

Table of Contents

1. Installation

2. Ingest Data

3. Continuous Ingestion

4. Run Demo

5. Example Usage

6. Start Application

Installation

To install this application, follow these steps:

1. Clone the repository:

git clone https://github.com/anminhhung/RAG_tool
cd RAG_tool

clone-repository.mp4

2. (Optional) Create and activate a virtual environment:

• For Unix/macOS:

python3 -m venv venv
source venv/bin/activate

• For Windows:

python -m venv venv
.\venv\Scripts\activate

Note: Please downgrade to python3.11 if any conflicts occur.

venv-setup.mp4

3. Install the required dependencies:

pip install -r requirements.txt

install-requirements.mp4

4. After activating your environment, run:

bash scripts/contextual_rag_additional_installation.sh

(Optional) Verify installation: You should run this to ensure all packages are installed successfully !

pip install pytest

pytest tests/

additional-install.mp4

5. Run database:

docker compose up -d

run-db.mp4

6. Config URL for database: In config/config.yaml, please modify urls of QdrantVectorDB and ElasticSearch:

...
CONTEXTUAL_RAG:
    ...
    QDRANT_URL: <fill here>

    ELASTIC_SEARCH_URL: <fill here>

config-db-url.mp4

7. Setup Agent: In config/config.yaml, please select agent type:

    ...
AGENT:
    TYPE: <fill here> # [openai, react]

Currently, we support:

TYPE	Agent
openai	OpenAIAgent
react	ReActAgent

setup-agent-type.mp4

8. Setup API Keys: Please create .env file and provide these API keys:

NAME	Where to get ?
OPENAI_API_KEY	OpenAI Platform
LLAMA_PARSE_API_KEY	LlamaCloud
COHERE_API_KEY	Cohere

setup-api-keys.mp4

---

Ingest data (Examples)

bash scripts/contextual_rag_ingest.sh both sample/

Note: Please refer to scripts/contextual_rag_ingest.sh to change the files dir.

ingest-example.mp4

Continuous Ingestion

• You can add more file paths or even folder paths:

python src/ingest/add_files.py --type both --files a.pdf b.docx docs/ ...

---

File Readers

• You can refer to: here to see how to use each of them.

File extension	Reader
.pdf	LlamaParse
.docx	DocxReader
.html	UnstructuredReader
.json	JSONReader
.csv	PandasCSVReader
.xlsx	PandasExcelReader
.txt	TxtReader

• Example usage of LlamaParse:

import os
from pathlib import Path
from dotenv import load_dotenv
from llama_index.readers.llama_parse import LlamaParse

load_dotenv()

loader = LlamaParse(result_type="markdown", api_key=os.getenv("LLAMA_PARSE_API_KEY"))

documents = loader.load_data(Path("sample/2409.13588v1.pdf"))

...

---

Run demo

• 1. Contextual RAG

python demo/demo_contextual_rag.py --q "Cái gì thất bại đề cử di sản thế giới ?" --compare --debug

• 2. ChatbotAssistant

python demo/demo_chatbot_assistant.py --q "ChainBuddy là gì ?"

---

Example Usage

• 1. Contextual RAG

from src.embedding import RAG
from src.settings import Settings

setting = Settings()

rag = RAG(setting)

q = "Cái gì thất bại đề cử di sản thế giới ?"

print(rag.contextual_rag_search(q))

• 2. ChatbotAssistant

from api.service import ChatbotAssistant

bot = ChatbotAssistant()

q = "ChainBuddy là gì ?"

print(bot.complete(q))

Start application

# backend
uvicorn app:app --reload

# UI
streamlit run streamlit_ui.py