Update README.md
Browse files
README.md
CHANGED
|
@@ -1,18 +1,62 @@
|
|
| 1 |
-
---
|
| 2 |
title: Email Gatekeeper RL
|
| 3 |
-
emoji: π§
|
| 4 |
colorFrom: blue
|
| 5 |
colorTo: indigo
|
| 6 |
sdk: docker
|
| 7 |
pinned: false
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
---
|
| 9 |
|
| 10 |
-
#
|
| 11 |
-
**Meta x PyTorch Hackathon** | Gymnasium-based email triage agent
|
| 12 |
|
| 13 |
-
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
-
|
| 16 |
-
1. Select the task difficulty (Easy, Medium, Hard).
|
| 17 |
-
2. Click on **Run Episode**.
|
| 18 |
-
3. View the agent's performance and cumulative reward.
|
|
|
|
|
|
|
| 1 |
title: Email Gatekeeper RL
|
| 2 |
+
emoji: "π§"
|
| 3 |
colorFrom: blue
|
| 4 |
colorTo: indigo
|
| 5 |
sdk: docker
|
| 6 |
pinned: false
|
| 7 |
+
-------------
|
| 8 |
+
|
| 9 |
+
# π§ Email Gatekeeper RL
|
| 10 |
+
|
| 11 |
+
A **Reinforcement Learning environment** for intelligently triaging emails based on:
|
| 12 |
+
|
| 13 |
+
* β‘ Urgency
|
| 14 |
+
* π Routing
|
| 15 |
+
* π οΈ Resolution
|
| 16 |
+
|
| 17 |
+
---
|
| 18 |
+
|
| 19 |
+
## βοΈ Environment
|
| 20 |
+
|
| 21 |
+
* `reset()` β start new episode
|
| 22 |
+
* `step(action)` β apply agent decision
|
| 23 |
+
* `state()` β get current email
|
| 24 |
+
|
| 25 |
+
---
|
| 26 |
+
|
| 27 |
+
## π― Action Space
|
| 28 |
+
|
| 29 |
+
```python
|
| 30 |
+
[action_urgency, action_routing, action_resolution]
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
* Urgency: 0 (low), 1 (medium), 2 (high)
|
| 34 |
+
* Routing: 0 (general), 1 (support), 2 (security)
|
| 35 |
+
* Resolution: 0 (ignore), 1 (respond), 2 (escalate)
|
| 36 |
+
|
| 37 |
+
---
|
| 38 |
+
|
| 39 |
+
## π Reward
|
| 40 |
+
|
| 41 |
+
* Partial reward per correct label
|
| 42 |
+
* Score β **[0, 1]**
|
| 43 |
+
|
| 44 |
+
---
|
| 45 |
+
|
| 46 |
+
## π₯οΈ Features
|
| 47 |
+
|
| 48 |
+
* π€ Rule-based baseline agent
|
| 49 |
+
* π Reward + score visualization
|
| 50 |
+
* π API endpoints (`/reset`, `/status`)
|
| 51 |
+
* π³ Docker deployment
|
| 52 |
+
|
| 53 |
---
|
| 54 |
|
| 55 |
+
## π Run
|
|
|
|
| 56 |
|
| 57 |
+
```bash
|
| 58 |
+
docker build -t email-gatekeeper .
|
| 59 |
+
docker run -p 7860:7860 email-gatekeeper
|
| 60 |
+
```
|
| 61 |
|
| 62 |
+
Open: http://localhost:7860
|
|
|
|
|
|
|
|
|