Implement models for BrainForge service

neon_data_models/models/api/http/__init__.py

@@ -0,0 +1,31 @@
+# NEON AI (TM) SOFTWARE, Software Development Kit & Application Development System
+# All trademark and other rights reserved by their respective owners
+# Copyright 2008-2024 Neongecko.com Inc.
+# BSD-3
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+# 1. Redistributions of source code must retain the above copyright notice,
+#    this list of conditions and the following disclaimer.
+# 2. Redistributions in binary form must reproduce the above copyright notice,
+#    this list of conditions and the following disclaimer in the documentation
+#    and/or other materials provided with the distribution.
+# 3. Neither the name of the copyright holder nor the names of its
+#    contributors may be used to endorse or promote products derived from this
+#    software without specific prior written permission.
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+# THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
+# CONTRIBUTORS  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
+# OR PROFITS;  OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+# SOFTWARE,  EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+from neon_data_models.models.api.http.brainforge import *
+
+"""
+This module contains models for interacting via HANA (HTTP).
+"""

neon_data_models/models/api/http/brainforge.py

@@ -0,0 +1,57 @@
+# NEON AI (TM) SOFTWARE, Software Development Kit & Application Development System
+# All trademark and other rights reserved by their respective owners
+# Copyright 2008-2024 Neongecko.com Inc.
+# BSD-3
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+# 1. Redistributions of source code must retain the above copyright notice,
+#    this list of conditions and the following disclaimer.
+# 2. Redistributions in binary form must reproduce the above copyright notice,
+#    this list of conditions and the following disclaimer in the documentation
+#    and/or other materials provided with the distribution.
+# 3. Neither the name of the copyright holder nor the names of its
+#    contributors may be used to endorse or promote products derived from this
+#    software without specific prior written permission.
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+# THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
+# CONTRIBUTORS  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
+# OR PROFITS;  OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+# SOFTWARE,  EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+from typing import List, Optional
+from pydantic import Field
+
+from neon_data_models.models.base import BaseModel
+from neon_data_models.models.api.llm import BrainForgeLLM, LLMPersona, LLMRequest
+
+
+class LLMGetModelsHttpResponse(BaseModel):
+    models: List[BrainForgeLLM]
+
+
+class LLMGetPersonasHttpRequest(BaseModel):
+    model_id: str = Field(
+        description="Model ID (<name>@<version>) to get personas for")
+
+
+class LLMGetPersonasHttpResponse(BaseModel):
+    personas: List[LLMPersona] = Field(
+        description="List of personas associated with the requested model.")
+
+
+class LLMGetInferenceHttpRequest(LLMRequest):
+    llm_name: str = Field(description="Model name to request")
+    llm_revision: str = Field(description="Model revision to request")
+    model: Optional[str] = None
+
+
+__all__ = [LLMGetModelsHttpResponse.__name__,
+           LLMGetPersonasHttpRequest.__name__,
+           LLMGetPersonasHttpResponse.__name__,
+           LLMGetInferenceHttpRequest.__name__]

neon_data_models/models/api/llm.py

@@ -37,7 +37,7 @@ class LLMPersona(BaseModel):
     name: str = Field(description="Unique name for this persona")
     description: Optional[str] = Field(
         None, description="Human-readable description of this persona")
-    system_prompt: str = Field(
+    system_prompt: Optional[str] = Field(
         None, description="System prompt associated with this persona. "
                           "If None, `description` will be used.")
     enabled: bool = Field(
@@ -48,8 +48,7 @@ class LLMPersona(BaseModel):
 
     @model_validator(mode='after')
     def validate_request(self):
-        assert any((self.description, self.system_prompt))
-        if self.system_prompt is None:
+        if self.system_prompt is None and self.description:
             self.system_prompt = self.description
         return self
 
@@ -71,7 +70,7 @@ class LLMRequest(BaseModel):
                     "OpenAI-compatible requests.")
     persona: LLMPersona = Field(
         description="Requested persona to respond to this message")
-    model: str = Field(description="Model to request")
+    model: str = Field(description="Model to request (<name>@<revision>)")
     max_tokens: int = Field(
         default=512, ge=64, le=2048,
         description="Maximum number of tokens to include in the response")
@@ -95,7 +94,8 @@ class LLMRequest(BaseModel):
                                   "Mutually exclusive with `stream`.")
     max_history: int = Field(
         default=2, description="Maximum number of user/assistant "
-                               "message pairs to include in history context.")
+                               "message pairs to include in history context. "
+                               "Excludes system prompt and incoming query.")
 
     @model_validator(mode='before')
     @classmethod
@@ -165,8 +165,10 @@ def to_completion_kwargs(self, mq2role: dict = None) -> dict:
         history = self.messages[-2*self.max_history:]
         for msg in history:
             msg["role"] = mq2role.get(msg["role"]) or msg["role"]
-        history.insert(0, {"role": "system",
-                           "content": self.persona.system_prompt})
+        if self.persona.system_prompt is not None:
+            history.insert(0, {"role": "system",
+                               "content": self.persona.system_prompt})
+        history.append({"role": "user", "content": self.query})
         return {"model": self.model,
                 "messages": history,
                 "max_tokens": self.max_tokens,
@@ -197,4 +199,19 @@ def validate_inputs(cls, values):
         return values
 
 
-__all__ = [LLMPersona.__name__, LLMRequest.__name__, LLMResponse.__name__]
+class BrainForgeLLM(BaseModel):
+    name: str = Field(description="LLM Name")
+    version: str = Field(description="LLM Version")
+    personas: List[LLMPersona] = Field(
+        default=[], description="List of personas defined in this model")
+
+    @property
+    def vllm_spec(self):
+        """
+        Model identifier used by vllm (<name>@<version>)
+        """
+        return f"{self.name}@{self.version}"
+
+
+__all__ = [LLMPersona.__name__, LLMRequest.__name__, LLMResponse.__name__,
+           BrainForgeLLM.__name__]

neon_data_models/models/api/mq/__init__.py

@@ -26,3 +26,7 @@
 
 from neon_data_models.models.api.mq.llm import *
 from neon_data_models.models.api.mq.users import *
+
+"""
+This module contains models for interacting via the MQ bus.
+"""

neon_data_models/models/api/mq/brainforge.py

@@ -0,0 +1,112 @@
+# NEON AI (TM) SOFTWARE, Software Development Kit & Application Development System
+# All trademark and other rights reserved by their respective owners
+# Copyright 2008-2024 Neongecko.com Inc.
+# BSD-3
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+# 1. Redistributions of source code must retain the above copyright notice,
+#    this list of conditions and the following disclaimer.
+# 2. Redistributions in binary form must reproduce the above copyright notice,
+#    this list of conditions and the following disclaimer in the documentation
+#    and/or other materials provided with the distribution.
+# 3. Neither the name of the copyright holder nor the names of its
+#    contributors may be used to endorse or promote products derived from this
+#    software without specific prior written permission.
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+# THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
+# CONTRIBUTORS  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
+# OR PROFITS;  OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+# SOFTWARE,  EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+from typing import List, Optional, Any, Dict, Literal, Union
+from pydantic import Field
+
+from neon_data_models.models.base.contexts import MQContext
+from neon_data_models.models.api.llm import BrainForgeLLM, LLMRequest, LLMResponse, LLMPersona
+from neon_data_models.models.api.http.brainforge import LLMGetModelsHttpResponse, LLMGetPersonasHttpRequest
+
+
+class LLMGetModels(MQContext):
+    user_id: str = Field(
+        description="ID of user to get models for")
+
+
+class LLMGetModelsResponse(MQContext, LLMGetModelsHttpResponse):
+    pass
+
+
+class LLMGetPersonas(LLMGetModels, LLMGetPersonasHttpRequest):
+    @property
+    def model_name(self):
+        return self.model_id.split("@")[0]
+
+    @property
+    def model_version(self):
+        return self.model_id.split("@", 1)[1]
+
+
+class LLMGetPersonasResponse(MQContext):
+    model: Optional[BrainForgeLLM] = Field(
+        "Full configuration of requested model if model is loaded and access "
+        "is allowed, else None.")
+
+    @property
+    def personas(self) -> List[LLMPersona]:
+        """
+        Convenience property defined to easily reference the personas requested
+        """
+        return self.model.personas if self.model else []
+
+
+class LLMGetInference(LLMRequest, MQContext):
+    user_id: str = Field("ID of user making the request")
+
+    def as_llm_request(self):
+        """
+        Get a plain `LLMRequest` object from this `LLMGetInference` object.
+        """
+        return LLMRequest(**self.model_dump())
+
+
+class LLMGetCompletion(LLMGetModels):
+    model: str = Field(description="Model to request (<name>@<revision>)")
+    completion_kwargs: Dict[str, Any] = Field(
+        description="Dictionary of kwargs to pass to OpenAI completion request")
+
+
+class LLMGetCompletionResponse(MQContext):
+    openai_response: dict = Field(
+        description="OpenAI ChatCompletion model")
+
+
+class LLMGetTokenizerChatTemplatedString(LLMGetModels):
+    model: str = Field(description="Model to request (<name>@<revision>)")
+    messages: List[Dict[Literal["role", "content"], str]] = Field(
+        description="List of dict messages in OpenAI format")
+    add_generation_prompt: bool = Field(
+        description="If true, assistant start tokens will be appended to the "
+                    "formatted output.")
+    tokenize: bool = Field(
+        False,
+        description="If true, a list of token strings is returned, "
+                    "else a single string")
+
+
+class LLMGetTokenizerChatTemplatedStringResponse(MQContext):
+    prompt: Union[List[str], str] = Field(
+        description="Prompt generated by the tokenizer")
+
+
+class LLMGetInferenceResponse(LLMResponse, MQContext):
+    pass
+
+
+__all__ = [LLMGetModels.__name__, LLMGetModelsResponse.__name__,
+           LLMGetPersonas.__name__, LLMGetPersonasResponse.__name__,
+           LLMGetInference.__name__, LLMGetInferenceResponse.__name__]

neon_data_models/models/api/node_v1/__init__.py

@@ -33,6 +33,11 @@
 from neon_data_models.models.base.messagebus import BaseMessage, MessageContext
 
 
+"""
+This module contains models for interacting via the Node socket (WS).
+"""
+
+
 class AudioInputData(BaseModel):
     audio_data: str = Field(description="base64-encoded audio")
     lang: str = Field(description="BCP-47 language code")

tests/models/api/test_llm.py

@@ -51,7 +51,7 @@ def test_llm_persona(self):
                          "A customized bot for something")
 
         with self.assertRaises(ValidationError):
-            LLMPersona(name="underdefined persona")
+            LLMPersona(description="underdefined persona")
 
     def test_llm_request(self):
         from neon_data_models.models.api.llm import LLMRequest, LLMPersona
@@ -71,7 +71,9 @@ def test_llm_request(self):
         self.assertFalse(valid_request.beam_search)
         self.assertEqual(len(valid_request.history), len(test_history))
         self.assertEqual(len(valid_request.to_completion_kwargs()['messages']),
-                         2 * valid_request.max_history + 1)
+                         2 * valid_request.max_history + 2)
+        self.assertEqual(valid_request.to_completion_kwargs()['messages'][-1]['content'],
+                         test_query)
 
         # Valid explicit streaming
         streaming_request = LLMRequest(query=test_query, history=test_history,