| --- |
| title: Silent Frontier |
| emoji: 🛰️ |
| colorFrom: green |
| colorTo: gray |
| sdk: gradio |
| sdk_version: 6.17.3 |
| python_version: "3.11" |
| app_file: app.py |
| pinned: false |
| license: mit |
| short_description: A reasoning agent that role-plays seven NPCs. |
| tags: |
| - track:wood |
| - sponsor:nvidia |
| - achievement:offbrand |
| --- |
| |
| # 🛰️ Silent Frontier |
|
|
| An NPC dialogue reasoning agent. Interrogate a chain of seven characters across a sci-fi mystery, earn each one's clue, and follow the trail from Commander Orion to the Warden. |
|
|
| ## 🧠 Built small |
|
|
| The default text model is Groq Llama 3.1 8B, with NVIDIA Nemotron Nano 9B, a small reasoning model, selectable as an alternate and easily toggleable. Both models are under the 32B parameter cap. |
|
|
| ## ⚙️ How it's built |
|
|
| DSPy reasoning agent with best-of-N sampling, FAISS retrieval over a hand-authored lore graph, a neurosymbolic constraint checker, per-NPC SQLite memory, Orpheus voice, and a custom Gradio dashboard with a node map, case file, and live reasoning trace. |
| --- |
|
|
|
|
| Built by **mkhi238** for the Hugging Face × Gradio Build Small hackathon. |
|
|
| Demo video: https://youtu.be/UyvjY-OsLWA |
|
|
| Social Post: https://x.com/MukundHari0/status/2066322648947146780 |
