# modelblob.py from fastapi import APIRouter, Request from fastapi.responses import HTMLResponse, JSONResponse import json, os, pathlib router = APIRouter() LOCAL_BLOB_PATH = os.getenv("MODEL_BLOB_PATH", "/tmp/model_blob.json") # ---- EMBEDDED MODEL BLOB (exactly as provided) ---- EMBEDDED_MODEL_BLOB = { "name": "publishers/hf-misri/models/realismenginesdxl_v30vae", "versionId": "001", "openSourceCategory": "THIRD_PARTY_OWNED_OSS", "supportedActions": { "deploy": { "modelDisplayName": "misri/realismEngineSDXL_v30VAE", "containerSpec": { "imageUri": "us-docker.pkg.dev/deeplearning-platform-release/vertex-model-garden/hf-inference-toolkit.cu125.0-1.ubuntu2204.py311:model-garden.hf-inference-toolkit-0-1-release_20250927.00_p0", "env": [ {"name": "HF_TASK","value": "text-to-image"}, {"name": "MODEL_ID","value": "misri/realismEngineSDXL_v30VAE"}, {"name": "HF_MODEL_ID","value": "misri/realismEngineSDXL_v30VAE"}, {"name": "HF_REVISION","value": "7d2f2de544b4aa26148b3a16b3469ed6dbb38a5c"}, {"name": "DEPLOY_SOURCE","value": "UI_HF_VERIFIED_MODEL"} ], "ports": [{"containerPort": 8080}] }, "dedicatedResources": { "machineSpec": { "machineType": "a3-highgpu-1g", "acceleratorType": "NVIDIA_H100_80GB", "acceleratorCount": 1 }, "maxReplicaCount": 1 }, "deployTaskName": "1 NVIDIA_H100_80GB a3-highgpu-1g", "deployMetadata": {} }, "multiDeployVertex": { "multiDeployVertex": [ { "modelDisplayName": "misri/realismEngineSDXL_v30VAE", "containerSpec": { "imageUri": "us-docker.pkg.dev/deeplearning-platform-release/vertex-model-garden/hf-inference-toolkit.cu125.0-1.ubuntu2204.py311:model-garden.hf-inference-toolkit-0-1-release_20250927.00_p0", "env": [ {"name": "HF_TASK","value": "text-to-image"}, {"name": "MODEL_ID","value": "misri/realismEngineSDXL_v30VAE"}, {"name": "HF_MODEL_ID","value": "misri/realismEngineSDXL_v30VAE"}, {"name": "HF_REVISION","value": "7d2f2de544b4aa26148b3a16b3469ed6dbb38a5c"}, {"name": "DEPLOY_SOURCE","value": "UI_HF_VERIFIED_MODEL"} ], "ports": [{"containerPort": 8080}] }, "dedicatedResources": { "machineSpec": { "machineType": "a3-highgpu-1g", "acceleratorType": "NVIDIA_H100_80GB", "acceleratorCount": 1 }, "maxReplicaCount": 1 }, "deployTaskName": "1 NVIDIA_H100_80GB a3-highgpu-1g", "deployMetadata": {} }, { "modelDisplayName": "misri/realismEngineSDXL_v30VAE", "containerSpec": { "imageUri": "us-docker.pkg.dev/deeplearning-platform-release/vertex-model-garden/hf-inference-toolkit.cu125.0-1.ubuntu2204.py311:model-garden.hf-inference-toolkit-0-1-release_20250927.00_p0", "env": [ {"name": "HF_TASK","value": "text-to-image"}, {"name": "MODEL_ID","value": "misri/realismEngineSDXL_v30VAE"}, {"name": "HF_MODEL_ID","value": "misri/realismEngineSDXL_v30VAE"}, {"name": "HF_REVISION","value": "7d2f2de544b4aa26148b3a16b3469ed6dbb38a5c"}, {"name": "DEPLOY_SOURCE","value": "UI_HF_VERIFIED_MODEL"} ], "ports": [{"containerPort": 8080}] }, "dedicatedResources": { "machineSpec": { "machineType": "g2-standard-12", "acceleratorType": "NVIDIA_L4", "acceleratorCount": 1 }, "maxReplicaCount": 1 }, "deployTaskName": "1 NVIDIA_L4 g2-standard-12", "deployMetadata": {} } ] } } } # --------------------------------------------------- def _ensure_dir(p: str): pathlib.Path(p).parent.mkdir(parents=True, exist_ok=True) def _pretty(obj) -> str: try: return json.dumps(obj, indent=2) except Exception: return str(obj) @router.get("/modelblob", response_class=HTMLResponse) def view_model_blob(): """ Render: if /tmp/model_blob.json exists, show it; otherwise show the embedded blob. """ try: if os.path.exists(LOCAL_BLOB_PATH): raw = pathlib.Path(LOCAL_BLOB_PATH).read_text(encoding="utf-8") try: disp = json.dumps(json.loads(raw), indent=2) except json.JSONDecodeError: disp = raw source = f"File • {LOCAL_BLOB_PATH}" else: disp = _pretty(EMBEDDED_MODEL_BLOB) source = "Embedded (not yet written to file)" except Exception as e: disp = _pretty({"error": str(e)}) source = "Error" html = f""" Model Blob

Current Model Blob ({source})

Write to /tmp/model_blob.json Back 
{disp}
""" return HTMLResponse(html) @router.get("/modelblob/write") def write_model_blob(): """ Write the embedded blob verbatim to LOCAL_BLOB_PATH. """ try: _ensure_dir(LOCAL_BLOB_PATH) pathlib.Path(LOCAL_BLOB_PATH).write_text(_pretty(EMBEDDED_MODEL_BLOB), encoding="utf-8") return JSONResponse({"ok": True, "path": LOCAL_BLOB_PATH}) except Exception as e: return JSONResponse({"error": str(e)}, 500) @router.post("/modelblob/overwrite") async def overwrite_blob(req: Request): """ Overwrite the file with a posted blob (JSON or form 'blob'). """ try: ctype = req.headers.get("content-type","") if "application/json" in ctype: data = await req.json() txt = data if isinstance(data, (dict, list)) else data.get("blob") else: form = await req.form() txt = form.get("blob") if isinstance(txt, (dict, list)): out = json.dumps(txt, indent=2) elif isinstance(txt, str) and txt.strip(): out = txt else: return JSONResponse({"error": "Missing valid blob"}, 400) _ensure_dir(LOCAL_BLOB_PATH) pathlib.Path(LOCAL_BLOB_PATH).write_text(out, encoding="utf-8") return JSONResponse({"ok": True, "path": LOCAL_BLOB_PATH}) except Exception as e: return JSONResponse({"error": str(e)}, 500) # --------------------------------------------------------------------- # JSON outlet for backend ingestion # --------------------------------------------------------------------- @router.get("/modelblob.json") def modelblob_json(): """Return the current model blob as JSON for backend ingestion.""" import os, json, pathlib path = os.getenv("MODEL_BLOB_PATH", "/tmp/model_blob.json") try: if os.path.exists(path): return json.loads(pathlib.Path(path).read_text(encoding="utf-8")) except Exception: pass # fallback to embedded blob if file missing or unreadable return EMBEDDED_MODEL_BLOB