Clarify JEVision model identity in visual examples and API

README.md

@@ -134,6 +134,8 @@ The saved [request](examples/real-photo/input.json) and [full response](examples
 }
 ```
 
+In these examples, `model: "jevvision"` is the API label echoed in the response; it does not choose the inference weights. The `images` field routes the request through the JEVision visual adapter and pointer head in this repository. The optional KEV-0.8B checkpoint is used only when selected for text-only requests. The saved photo response retains its recorded answer, probabilities, token counts, and latency; only its echoed model label was normalized from the older API alias.
+
 This is one recorded example. The full response also records the latency of that CPU run; it is not a speed comparison. Run it yourself after starting the server:
 
 ```bash

examples/long-context/input.json

@@ -1,5 +1,5 @@
 {
-  "model": "kev-latest",
+  "model": "jevvision",
   "state_target_tokens": 76000,
   "minimum_request_tokens": 75000,
   "state_file": "state-75k.txt",

examples/real-photo/input.json

@@ -1,6 +1,6 @@
 {
   "state": "Use the attached photo as visual context.",
-  "model": "kev-latest",
+  "model": "jevvision",
   "image_file": "coffee-and-laptop.jpg",
   "questions": {
     "scene": {

examples/real-photo/noul-input.json

@@ -1,6 +1,6 @@
 {
   "state": "Use the attached photo as visual context.",
-  "model": "kev-latest",
+  "model": "jevvision",
   "image_file": "coffee-and-laptop.jpg",
   "questions": {
     "is_macbook": {

examples/real-photo/output.json

@@ -1,5 +1,5 @@
 {
-  "model": "kev-latest",
+  "model": "jevvision",
   "answers": {
     "scene": {
       "type": "choice",

runtime/kev/api.py

@@ -64,7 +64,7 @@ class SystemOneRequest(BaseModel):
     model_config = ConfigDict(hide_input_in_errors=True)
 
     state: JSONContent
-    model: str = "kev-latest"
+    model: str = "jevvision"
     questions: dict[str, Question] = Field(min_length=1)
     images: list[ImageInput] | None = Field(default=None, min_length=1, max_length=MAX_IMAGES)   # omitted or null = text-only
 

runtime/kev/jevvision.py

@@ -124,7 +124,7 @@ class JEVision:
         state,
         questions: dict,
         images: list[dict[str, str]] | None = None,
-        model: str = "jev-latest",
+        model: str = "jevvision",
     ) -> dict:
         """Return one KEV/System One response dict for the supplied request."""
         request = SystemOneRequest(

runtime/kev/serve.py

@@ -129,7 +129,7 @@ def prepare(req):
     return req.model_copy(update={"state": with_date_facts(req.state)}) if DATE_FACTS else req
 
 
-app = FastAPI(title="kev")
+app = FastAPI(title="JEVision")
 app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])
 
 
@@ -194,8 +194,9 @@ def systemone_separate(req: SystemOneRequest):
 @app.get("/v1/models")
 def models():
     s = server()
-    return {"models": [{"id": "kev-latest", "aliases": ["jev-latest"], "run": s.checkpoint.requested, "base": s.checkpoint.meta.base,
-                        "lora": s.checkpoint.meta.lora, "device": s.device, "backend": s.model.backend, "dtype": s.model.dtype, "temperature": s.model.head.temperature,
+    return {"models": [{"id": "jevvision", "aliases": [], "run": s.checkpoint.requested, "base": s.checkpoint.meta.base,
+                        "lora": s.checkpoint.meta.lora, "visual_route": "JEVision visual sidecar" if s.visual is not None else None,
+                        "device": s.device, "backend": s.model.backend, "dtype": s.model.dtype, "temperature": s.model.head.temperature,
                         "context_tokens": s.context_limit,
                         "prefix_cache": {"size": PREFIX_CACHE_SIZE, "min_state_tokens": s.prefix_min_tokens, "hits": s.prefix_hits,
                                          "max_state_tokens": PREFIX_CACHE_MAX_TOKENS, "misses": s.prefix_misses,