agent-ken-pm-social-impact-copilot

Agent Ken — PM + Social Impact Copilot (Azure AI Foundry)

🇮🇩 Ringkasan

Agent Ken adalah AI agent berbasis Azure AI Foundry Agent Service yang berperan sebagai Product Manager Copilot dengan empati ala social worker dan prinsip harm-reduction. Repo ini berisi Model Card + artifacts (konfigurasi agent, guardrails, prompt tests) untuk portfolio dan reproducibility. 12

Penting: Repo ini tidak menyertakan model weights (mis. .pt/.safetensors). Inference berjalan melalui Azure-hosted endpoint (Azure AI Foundry). 268

🇬🇧 Summary

Agent Ken is an AI agent built on Azure AI Foundry Agent Service, acting as a Product Manager Copilot with a social worker–informed empathy and harm-reduction mindset. This repository contains the Model Card + artifacts (agent configuration, guardrails, test prompts) for portfolio and reproducibility. 12

Important: This repo does not ship downloadable model weights (e.g., .pt/.safetensors). Inference runs via an Azure-hosted endpoint (Azure AI Foundry). 268

---

✅ How you can use it | Bagaimana cara memakainya

Option A — Public Web Demo (recommended) | Demo Publik (disarankan)

🇮🇩 Kamu bisa mencoba Agent Ken lewat Hugging Face Space (UI web). Space akan menjalankan aplikasi demo dan memanggil agent Azure di belakang layar. Hugging Face Spaces adalah repo git yang auto-build dan bisa menyimpan secrets/variables di Settings. 345
🇬🇧 You can try Agent Ken via a Hugging Face Space (web UI). The Space runs a demo app and calls the Azure agent behind the scenes. Hugging Face Spaces are git repos with auto-build, and secrets/variables can be configured in Space Settings. 345

• Demo link: (put your Space URL here)
  Example: https://huggingface.co/spaces/<owner>/<space-name> 3

⚠️ Cost note (public demo): If the Space uses your Azure credentials, usage consumes your Azure quota/cost. Consider basic rate limits/cooldowns. 45

---

Option B — Bring Your Own Azure (BYO) | Jalankan sendiri pakai Azure kamu

🇮🇩 Developer bisa menjalankan agent dari kode (Python) jika punya Azure AI Foundry Project sendiri dan akses RBAC yang sesuai. SDK Projects menggunakan project endpoint format https://<resource>.services.ai.azure.com/api/projects/<project> dan menjalankan agent via flow thread → message → run → list messages. 678
🇬🇧 Developers can run the agent via code (Python) if they have their own Azure AI Foundry Project and proper RBAC. The Projects SDK uses a project endpoint of the form https://<resource>.services.ai.azure.com/api/projects/<project> and runs an agent using the thread → message → run → list messages flow. 678

Prerequisites | Prasyarat

• Python 3.9+ 6
• Azure AI Foundry Project endpoint:
  https://<resource>.services.ai.azure.com/api/projects/<project> 68
• Authentication via Microsoft Entra ID (recommended). The SDK expects a TokenCredential (e.g., DefaultAzureCredential). 611

Environment variables (recommended) | Variabel environment (disarankan)

Set these in your environment (do not commit secrets):

# Azure Foundry project + agent
PROJECT_ENDPOINT="https://<resource>.services.ai.azure.com/api/projects/<project>"
AGENT_ID="asst_..."

# Service principal auth (recommended for servers/CI)
AZURE_TENANT_ID="<tenant-id>"
AZURE_CLIENT_ID="<client-id>"
AZURE_CLIENT_SECRET="<client-secret>"