Removing old KServe implementation

docs/python_gaas_tools/gaas_model.md

@@ -27,8 +27,6 @@ The `py/gaas_gpt_model.py` module provides classes for interacting with various
         *   Constructs a Pixel script: `LLM(engine="<engine_id>", command="<question>", useHistory=<use_history>, context=["<context>"], paramValues=[<param_dict>]);`.
         *   Executes this Pixel script via `super().callReactor()`.
         *   Returns the response from the LLM, typically including the generated text and token counts.
-    *   `ner(text: str, entities: List[str], mask_entities: List[str] = [], param_dict: Optional[Dict] = None, insight_id: Optional[str] = None)`:
-        *   Executes an `NER(...)` Pixel command for Named Entity Recognition.
     *   `get_conversation_history(insight_id: Optional[str] = None) -> List[Dict]`:
         *   Executes a `GetRoomMessages(roomId="<insight_id>");` Pixel script to retrieve chat history from the `ModelInferenceLogsDatabase`.
     *   `embeddings(strings_to_embed: List[str], param_dict: Optional[Dict] = None, insight_id: Optional[str] = None) -> List[Dict]`:
@@ -74,7 +72,7 @@ The `py/gaas_gpt_model.py` module provides classes for interacting with various
 *   **Initialization `__init__(self, model_engine_class: Optional[str] = "TOMCAT", **kwargs)`**:
     *   `model_engine_class` (str, default: "TOMCAT"): Determines the type of engine to create. Can be "TOMCAT" or "LOCAL". "HF_PIPELINE" is mentioned but not fully implemented in the constructor logic shown.
     *   `**kwargs`: Passed to the constructor of the chosen model engine class (e.g., `engine_id` for `TomcatModelEngine`).
-*   **Key Methods**: All methods (`get_model_type`, `ask`, `instruct`, `embeddings`, `keyword_extraction`, `ner`, `do_call`, `get_model_engine_id`, `get_conversation_history`) are wrappers that delegate the call to the underlying `self.model_engine` instance.
+*   **Key Methods**: All methods (`get_model_type`, `ask`, `instruct`, `embeddings`, `keyword_extraction`, `do_call`, `get_model_engine_id`, `get_conversation_history`) are wrappers that delegate the call to the underlying `self.model_engine` instance.
 *   **Langchain Integration**:
     *   `to_langchain_embedder()`: Wraps the `ModelEngine` to make its `embeddings` method compatible with Langchain's `Embeddings` interface.
     *   `to_langchain_chat_model()`: Wraps the `ModelEngine` to make its `ask` and `get_conversation_history` methods compatible with Langchain's `BaseChatModel` interface. This includes converting message formats.

project/platform__model/app_root/version/assets/public/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: model
-description: Use when writing code in an app that calls an LLM, embedding model, or other model engine, OR when listing/selecting models the user has access to. Covers the LLM() and MyEngines() pixel commands via @semoss/sdk's runPixel, including prompt/completion calls, conversational history, structured outputs, attaching media of any type (image, PDF, Word, spreadsheet, audio, video) with the media argument, and parsing model responses. Also covers which top-level arguments full_prompt drops, media included, plus the other model pixels: Embeddings(), MultiModalEmbeddings(), Vision(), NER(), GetContextWindow(), GetUserModelUsage(), and the provider batch API (BatchLLM, GetModelBatchStatus, GetModelBatchResults). Do not use for vector database queries (see vector).
+description: Use when writing code in an app that calls an LLM, embedding model, or other model engine, OR when listing/selecting models the user has access to. Covers the LLM() and MyEngines() pixel commands via @semoss/sdk's runPixel, including prompt/completion calls, conversational history, structured outputs, attaching media of any type (image, PDF, Word, spreadsheet, audio, video) with the media argument, and parsing model responses. Also covers which top-level arguments full_prompt drops, media included, plus the other model pixels: Embeddings(), MultiModalEmbeddings(), Vision(), GetContextWindow(), GetUserModelUsage(), and the provider batch API (BatchLLM, GetModelBatchStatus, GetModelBatchResults). Do not use for vector database queries (see vector).
 ---
 
 # Model Engine
@@ -212,14 +212,13 @@ Embeddings(engine="${EMBEDDING_MODEL_ID}", values=["first text", "second text"])
 
 Returns a map with the vectors in `pixelReturn[0].output`. If the engine is not an embedding model, the output is the literal string "This model does not support embeddings." - check for it. `MultiModalEmbeddings(engine=, text=, image=, video=)` embeds any combination of modalities (image/video accept base64, data URLs, or remote URLs), with results broken out by modality.
 
-### Vision and NER
+### Vision
 
 ```
 Vision(engine="${MODEL_ID}", command=["Describe this diagram"], image=["${urlOrBase64}"]);
-NER(engine="${NER_MODEL_ID}", prompt=["<encode>${text}</encode>"], entities=["PERSON","EMAIL"], maskEntities=[true]);
 ```
 
-`Vision` is the image-text-to-text task for models exposing it. `NER` extracts (and with `maskEntities` redacts) named entities - the engine must actually be an NER model or the call fails.
+`Vision` is the image-text-to-text task for models exposing it.
 
 ### Token budget and usage
 

py/gaas_gpt_model.py

@@ -206,35 +206,6 @@ def ask(
 
         return pixelReturn
 
-    def ner(
-        self,
-        text: str,
-        entities: List[str],
-        mask_entities: List[str] = [],
-        param_dict: Optional[Dict] = None,
-        insight_id: Optional[str] = None,
-    ):
-        if insight_id is None:
-            insight_id = self.insight_id
-
-        epoc = super().get_next_epoc()
-
-        pixel = f'NER(engine="{self.engine_id}", prompt="{text}", entities={entities}, maskEntities={mask_entities});'
-
-        pixelReturn = super().callReactor(
-            epoc=epoc,
-            pixel=pixel,
-            insight_id=insight_id,
-        )
-
-        if pixelReturn is not None and len(pixelReturn) > 0:
-            output = pixelReturn[0]["pixelReturn"][0]
-            if "ERROR" in output["operationType"]:
-                raise Exception(output["output"])
-            return output["output"]
-
-        return pixelReturn
-
     def get_conversation_history(self, insight_id: Optional[str] = None) -> List[Dict]:
         """This method is responsible to get message history back from the model logs database.
 
@@ -608,15 +579,6 @@ def keyword_extraction(
     ):
         return self.model_engine.keyword_extraction(**kwargs)
 
-    def ner(
-        self,
-        insight_id: Optional[
-            str
-        ] = None,  # TODO remove once users stop using it. No longer needs to be set.
-        **kwargs,
-    ):
-        return self.model_engine.ner(**kwargs)
-
     def do_call(self, **kwargs: Any) -> Any:
         return self.model_engine.embeddings(**kwargs)