init mirror from ashirato shard

README.md

@@ -1,12 +1,41 @@
 ---
-title: Surrogate 1 Zero Gpu
-emoji: 🌍
-colorFrom: pink
-colorTo: blue
+title: Surrogate-1 ZeroGPU
+emoji: 🚀
+colorFrom: indigo
+colorTo: purple
 sdk: gradio
-sdk_version: 6.13.0
+sdk_version: 4.44.0
 app_file: app.py
-pinned: false
+pinned: true
+license: apache-2.0
+short_description: Surrogate-1 v1 LoRA on Qwen2.5-Coder-7B (ZeroGPU A10G)
+suggested_hardware: zero-a10g
+hf_oauth: false
+models:
+  - Qwen/Qwen2.5-Coder-7B-Instruct
+  - axentx/surrogate-1-coder-7b-lora-v1
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Surrogate-1 ZeroGPU
+
+DevSecOps + SRE + coding agent. Qwen2.5-Coder-7B-Instruct + Surrogate-1 v1
+LoRA, served via HF ZeroGPU (A10G, 60-120s per request).
+
+## Endpoints
+
+- Web UI: this Space (Gradio chat)
+- OpenAI-compatible: `/api/predict` (Gradio API auto-generated)
+- Use programmatically: `gradio_client.Client("ashirato/surrogate-1-zero-gpu")`
+
+## Why ZeroGPU
+
+PRO unlocks 25K minutes/mo of A10G time at $0/mo. Each request gets fresh
+GPU, so cold-start ~5-10s but no idle cost. Perfect for low-traffic
+agentic loops (self-improve, constitutional, validator-RLVR judge calls).
+
+## Connected to axentx/surrogate-1
+
+This Space serves inference. The orchestration Space at
+`axentx/surrogate-1` runs cron loops + bulk-mirror harvest + state DBs;
+those loops can call THIS endpoint for actual model output instead of
+free-tier API ladder.