Update README.md

README.md

@@ -1,12 +1,69 @@
 ---
-title: Mail Triage
-emoji: 🐢
-colorFrom: red
+title: Mail Triage Agent
+emoji: 👀
+colorFrom: blue
 colorTo: green
 sdk: docker
 pinned: false
 license: mit
-short_description: The project classifies mail based on useful, spam or junk.
+---
+# AI MAIL TRIAGE AGENT SYSTEM
+---
+
+## PROJECT OVERVIEW
+
+This project implements an **Automated Email Triage System** designed to handle high-volume incoming mail streams. Leveraging the `openenv-core` framework and **GPT-4o**, the agent serves as a primary security layer, classifying emails into distinct categories based on their security profile, metadata, and content analysis.
+
+---
+
+## SYSTEM ARCHITECTURE
+
+### 1. OBSERVATION SPACE
+
+The agent processes a structured `MyEnvV4Observation` for every email, evaluating:
+
+* **Sender Identification**: Validation of origin domains (e.g., distinguishing legitimate `.edu` domains from malicious alternatives).
+* **Content Analysis**: Scanning subjects and message bodies for high-urgency keywords, phishing links, and social engineering tactics.
+* **Authentication Headers**: Analyzing technical metadata, including **SPF**, **DKIM**, and **DMARC** status, to detect spoofing.
+
+### 2. ACTION SPACE
+
+For every observation, the agent generates a `MyEnvV4Action` containing a classification:
+
+1. **INBOX**: Legitimate academic or personal communications.
+2. **SPAM**: Unsolicited marketing or mass junk mail.
+3. **QUARANTINE**: High-risk phishing attempts or suspicious security alerts.
+
+---
+
+## EVALUATION AND REWARDS
+
+The environment assesses agent performance through a precision-based reward system:
+
+* **CORRECT CLASSIFICATION**: `+1.0 Reward`
+* **INCORRECT CLASSIFICATION**: `0.0 Reward`
+* **TERMINATION**: The environment concludes once the triage dataset has been fully processed.
+
+---
+
+## TECHNICAL STACK
+
+* **CORE FRAMEWORK**: `openenv-core==0.2.3`
+* **WEB SERVER**: FastAPI / Uvicorn
+* **LANGUAGE**: Python 3.10-slim
+* **LLM INFERENCE**: OpenAI GPT-4o
+* **INFRASTRUCTURE**: Docker-based deployment on Hugging Face Spaces
+
+---
+
+## INSTALLATION AND EXECUTION
+
+To deploy this environment:
+
+1. Configure the `HF_TOKEN` within the Environment Secrets.
+2. Build the container utilizing the provided `Dockerfile`.
+3. The server initializes on **Port 8000**, exposing endpoints for `/reset`, `/step`, and `/state`.
+
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+**Developed by B.Tech CSE Students | Manipal Institute of Technology**