title	Customer Support Agent Env
emoji	🎫
colorFrom	blue
colorTo	purple
sdk	docker
app_port	7860
tags	openenv
pinned	false

Customer Support Agent Environment

An OpenEnv-compatible RL environment where an AI agent handles customer support tickets end-to-end — classifying issues, deciding resolution actions, and drafting professional replies.

What the agent does

1. Reads a customer ticket: "I was charged twice for my order"
2. Classifies the issue: billing / shipping / product
3. Decides the action: refund / replace / escalate / resolve
4. Writes a professional reply to the customer
5. Gets a reward score (0.0 – 1.0) based on accuracy

Tasks

Task	Difficulty	What is graded
easy	Easy	Issue classification only (1.0 if correct)
medium	Medium	Issue + Action (0.5 each)
hard	Hard	Issue + Action + Reply length + Action in reply (0.25 each)

Reward function

Component	Points	Rule
Issue correct	+0.25	Exact match
Action correct	+0.25	Exact match
Reply length	+0.25	More than 30 characters
Action in reply	+0.25	Action word appears in reply
Total	1.0	Clamped to [0.0, 1.0]

Dataset

20 labelled tickets: 6 easy + 6 medium + 8 hard Issues covered: billing, shipping, product

Observation space

State:
  ticket: str        # The raw customer support ticket
  difficulty: str    # easy | medium | hard
  reward: float      # Score from last step (0.0–1.0)
  done: bool         # Is the episode complete?

Action space

Action:
  issue:  billing | shipping | product
  action: refund | replace | escalate | resolve
  reply:  str  (professional reply to the customer)

Run locally

git clone https://huggingface.co/spaces/xyzzzz21122/customer-support-agent-env
cd customer-support-agent-env
pip install -r requirements.txt
python app.py

Run with Docker

docker build -t customer-support-env .
docker run -p 7860:7860 customer-support-env

Run inference (baseline benchmark)

python inference.py

Baseline scores (rule-based keyword agent)

Difficulty	Avg Score
Easy	1.00
Medium	0.75
Hard	0.88
Overall	0.88

Files

app.py          Gradio UI
env.py          CustomerSupportEnv — reset() / step() / state()
evaluator.py    Difficulty-aware reward function
agent.py        Keyword-based baseline agent
dataset.py      20 labelled support tickets
models.py       State and Action dataclasses
inference.py    Benchmark script
openenv.yaml    OpenEnv manifest
Dockerfile      Container setup