Cloud SRE: Autonomous Fleet Management (OpenEnv)

Motivation

Environment Interface

Observation Space

• services: List of objects with status, replicas, latency_ms, cpu_usage_percent, and last_logs.
• system_metrics: High-level cluster health (error rate, overall load).
• current_time: Timestamp of the observation.
• last_action_result: Feedback from the previous step.

Action Space

• action_type: restart, scale, update_config, rollback, or noop.
• service_name: Target service ID.
• params: Action-specific parameters (e.g., replicas for scale, or key:value for config).

Reward Function

1. Service Uptime (Primary): Balanced across all services.
2. Performance Metrics: Penalties for latency spikes and high error rates.
3. Efficiency: Optimal resource allocation (not scaling more than needed).
4. Action Validity: Penalties for invalid actions or noops in critical states.

Tasks

Setup & Usage

Local Development

1. Install dependencies:
   pip install -r requirements.txt
2. Run the environment:
   python app.py
3. Run the baseline agent:
   bash

   1# Set your API keys first 2export HF_TOKEN="your_huggingface_token" 3export MODEL_NAME="gpt-4o" 4python inference.py

Docker

Baseline Scores (Estimated)

Documentation & Compliance

• OpenEnv Specification: Fully compliant. openenv validate passes.
• HF Space: Deploy as a containerized Space.
• Task Graders: Included in tasks.py, providing normalized scores (0.0–1.0). #� �S�o�l�o�-�W�a�r�r�i�o�r�-�m�e�t�a�-� � �