This project is an AI-powered Sentiment Analysis Agent developed using SmolAgents and Hugging Face Transformers. It demonstrates how LLM-based AI agents can interact with external tools, automate decision-making, and efficiently process text data.
The project was built as part of my AI Agents Fundamentals Course by Hugging Face, where I explored AI Agents, Code Agents, and structured reasoning in AI workflows. The core functionality revolves around integrating LLMs with SmolAgents to classify sentiment in text.
✅ Try it out here: Test the Sentiment Analysis Agent
This project is live on Hugging Face Spaces, where you can interact with the AI-powered sentiment analysis tool. The agent analyzes text sentiment using SmolAgents and Hugging Face Transformers, making structured decisions based on language context.
Example Interaction:
User Input: It is not true that I was angry or sad.
Step 1
Thought: I will use the advanced_sentiment_tool to analyze the sentiment of the given text and determine if the sentiment is positive, negative, or neutral.
When given the input:
"It is not true that I was angry or sad."
🔹 Step 1: The agent initially detected a negative sentiment.
🔹 Step 2: It rephrased the input to clarify intent.
🔹 Step 3: After reanalyzing, it correctly determined the sentiment was positive.
Final Answer: ✅ "The sentiment of the text is positive."
(The AI agent autonomously reasoned about negation and corrected its initial analysis.)
User Input:
It is not true that I was angry or sad.
AI Agent Response:
Step 1
Thought: I will use the advanced_sentiment_tool to analyze the sentiment of the given text and determine if the sentiment is positive, negative, or neutral.
Code:
sentiment = advanced_sentiment_tool(text="It is not true that I was angry or sad.")
print(f"The sentiment of the text is: {sentiment}")
🛠️ Used tool python_interpreter
Step 1 | Input-tokens:3,226 | Output-tokens:76 | Duration: 5.29
AI Agent Response:
Step 2
Thought: Based on the sentiment analysis, the sentiment of the text is negative. However, this might be due to the negation in the sentence. The person is actually stating that they were not angry or sad, which implies a positive sentiment. I will rephrase the text to make the sentiment clearer and analyze it again.
Code:
sentiment = advanced_sentiment_tool(text="I was neither angry nor sad.")
print(f"The sentiment of the text is: {sentiment}")
🛠️ Used tool python_interpreter
Step 2 | Input-tokens:3,427 | Output-tokens:104 | Duration: 5.54
AI Agent Response:
Step 3
Thought: Based on the sentiment analysis, the sentiment of the text is now correctly identified as positive. The person is stating that they were neither angry nor sad, which implies a positive sentiment. I will now return the final answer using the final_answer tool.
Code:
final_answer("The sentiment of the text is positive.")
🛠️ Used tool python_interpreter
Step 3 | Input-tokens:3,652 | Output-tokens:74 | Duration: 4.24
AI Agent Response:
✅ Final Answer:
The sentiment of the text is positive.
- 🤖 AI Agents: Uses SmolAgents to create an interactive AI-powered agent.
- 💬 Sentiment Analysis: Classifies text sentiment (e.g., Positive, Negative, Neutral) using Hugging Face Transformers.
- 🧠 LLM Reasoning: Employs LLMs for structured decision-making in NLP tasks.
- 📡 API Integration: Can be expanded to connect with real-time applications and external APIs.
- 🌍 Deployable: Easily deployable on Hugging Face Spaces, Gradio, or cloud platforms.
├── app.py # Main application file
├── agent.json # Agent configuration file
├── Gradio_UI.py # Gradio-based UI for interactive testing
├── tools/ # Custom tools for AI interaction
│ ├── final_answer.py # Tool to provide final responses
├── prompts.yaml # Prompt engineering templates
├── README.md # Project documentation (this file)
├── requirements.txt # Python dependencies
├── .gitignore # Files to exclude from GitHub
└── .gitattributes # Git Large File Storage (LFS) configurations
| Category | Tools & Libraries |
|---|---|
| AI Models | Hugging Face Transformers, PyTorch |
| AI Agents | SmolAgents, Code Agents |
| NLP | Sentiment Analysis, LLM Reasoning |
| Backend | Python |
| Frontend | Gradio (for UI) |
| Deployment | Hugging Face Spaces |
1️⃣ Clone this repository:
git clone https://github.com/khaledjamal/sentiment-analysis-agent.git
cd sentiment-analysis-agent2️⃣ Install dependencies:
pip install -r requirements.txt3️⃣ Run the AI Agent:
python app.py4️⃣ (Optional) Launch Gradio UI:
python Gradio_UI.pyYou can deploy this project on Hugging Face Spaces.
🔹 Enhance model fine-tuning with domain-specific data
🔹 Expand AI agent capabilities for more advanced LLM-powered decision-making
Khaled Jamal
🔗 LinkedIn
📧 javaobjects@gmail.com
⭐ If you found this project useful, please star this repository on GitHub!
💡 Feel free to submit issues or pull requests for enhancements!