| title | SupportEnv |
|---|---|
| emoji | ๐ |
| colorFrom | purple |
| colorTo | gray |
| sdk | docker |
| app_port | 7860 |
๐ Live Demo: https://munshiarpan19-support-env.hf.space/ui
SupportEnv is a real-world simulation environment designed for training and evaluating AI agents on customer support tasks.
Customer support is a complex, real-world task involving:
- Understanding user intent
- Handling emotions (neutral / angry users)
- Taking correct actions (reply, escalate)
- Providing accurate solutions
This environment simulates those challenges in a structured, testable way.
- โ Real-world task simulation (customer support tickets)
- โ OpenEnv-style step/reset/state API
- โ
Multi-difficulty tasks:
- Easy โ Password reset
- Medium โ Billing/refund issue
- Hard โ Technical crash issue
- โ Reward shaping (partial progress signals)
- โ Deterministic grading system (0.0โ1.0 score)
- โ Baseline AI agent with realistic performance variance
- โ Interactive UI dashboard for manual testing
| Endpoint | Description |
|---|---|
/reset |
Start new task |
/step |
Perform action |
/state |
Get current state |
/tasks |
List all tasks |
/grader |
Evaluate performance |
/baseline |
Run baseline agent |
/ui |
Interactive dashboard |
{
"response": "string",
"action_type": "reply | escalate | close"
}
## ๐๏ธ Observation Space
{
"message": "string",
"sentiment": "neutral | angry",
"difficulty": "easy | medium | hard"
}
## ๐ฅ Demo
Try it live:
- /ui โ Interactive dashboard
- /baseline โ Auto agent evaluation
- /tasks โ Task definitions