ommlds 0.0.0.dev441__py3-none-any.whl → 0.0.0.dev442__py3-none-any.whl

ommlds/minichain/backends/impls/google/chat.py

@@ -4,10 +4,13 @@ https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models
 import typing as ta
 
 from omlish import check
+from omlish import marshal as msh
 from omlish import typedvalues as tv
 from omlish.formats import json
 from omlish.http import all as http
 
+from .....backends.google.protocol.types import GenerateContentRequest
+from .....backends.google.protocol.types import GenerateContentResponse
 from ....chat.choices.services import ChatChoicesRequest
 from ....chat.choices.services import ChatChoicesResponse
 from ....chat.choices.services import static_check_is_chat_choices_service
@@ -57,25 +60,25 @@ class GoogleChatChoicesService:
         AiMessage: 'assistant',
     }
 
-    def invoke(
+    async def invoke(
             self,
             request: ChatChoicesRequest,
     ) -> ChatChoicesResponse:
         key = check.not_none(self._api_key).reveal()
 
-        req_dct = {
-            'contents': [
-                {
-                    'role': self.ROLES_MAP[type(m)],
-                    'parts': [
-                        {
-                            'text': self._get_msg_content(m),
-                        },
-                    ],
-                }
+        g_req = GenerateContentRequest(
+            contents=[
+                GenerateContentRequest.Content(
+                    parts=[GenerateContentRequest.Content.Part(
+                        text=check.not_none(self._get_msg_content(m)),
+                    )],
+                    role=self.ROLES_MAP[type(m)],  # type: ignore[arg-type]
+                )
                 for m in request.v
             ],
-        }
+        )
+
+        req_dct = msh.marshal(g_req)
 
         model_name = MODEL_NAMES.resolve(self._model_name.v)
 
@@ -88,7 +91,9 @@ class GoogleChatChoicesService:
 
         resp_dct = json.loads(check.not_none(resp.data).decode('utf-8'))
 
+        g_resp = msh.unmarshal(resp_dct, GenerateContentResponse)
+
         return ChatChoicesResponse([
-            AiChoice(AiMessage(c['content']['parts'][0]['text']))
-            for c in resp_dct['candidates']
+            AiChoice(AiMessage(c.content.parts[0].text))
+            for c in g_resp.candidates
         ])

ommlds/minichain/backends/impls/google/search.py

@@ -88,7 +88,7 @@ class CseSearchService:
         self._cse_id = cse_id
         self._cse_api_key = cse_api_key
 
-    def invoke(
+    async def invoke(
             self,
             request: SearchRequest,
     ) -> SearchResponse:

ommlds/minichain/backends/impls/llamacpp/chat.py

@@ -54,7 +54,7 @@ class LlamacppChatChoicesService:
         temperatur=Temperature,
     )
 
-    def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
+    async def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
         kwargs: dict = dict(
             # temperature=0,
             max_tokens=1024,

ommlds/minichain/backends/impls/llamacpp/completion.py

@@ -51,7 +51,7 @@ class LlamacppCompletionService:
         temperatur=Temperature,
     )
 
-    def invoke(self, request: CompletionRequest) -> CompletionResponse:
+    async def invoke(self, request: CompletionRequest) -> CompletionResponse:
         kwargs: dict = dict(
             # temperature=0,
             max_tokens=1024,

ommlds/minichain/backends/impls/llamacpp/stream.py

@@ -49,7 +49,7 @@ class LlamacppChatChoicesStreamService(lang.ExitStacked):
             verbose=False,
         )))
 
-    def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
+    async def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
         lcu.install_logging_hook()
 
         with UseResources.or_new(request.options) as rs:

ommlds/minichain/backends/impls/mistral.py

@@ -50,7 +50,7 @@ class MistralChatChoicesService:
         else:
             raise TypeError(m)
 
-    def invoke(
+    async def invoke(
             self,
             request: ChatChoicesRequest,
     ) -> ChatChoicesResponse:

ommlds/minichain/backends/impls/mlx/chat.py

@@ -96,7 +96,7 @@ class MlxChatChoicesService(lang.ExitStacked):
         max_tokens=MaxTokens,
     )
 
-    def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
+    async def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
         loaded_model = self._load_model()
 
         tokenizer = loaded_model.tokenization.tokenizer

ommlds/minichain/backends/impls/openai/chat.py

@@ -47,7 +47,7 @@ class OpenaiChatChoicesService:
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
             self._default_options: tv.TypedValues = DefaultOptions.pop(cc)
 
-    def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
+    async def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
         # check.isinstance(request, ChatRequest)
 
         rh = OpenaiChatRequestHandler(

ommlds/minichain/backends/impls/openai/completion.py

@@ -29,7 +29,7 @@ class OpenaiCompletionService:
         with tv.consume(*configs) as cc:
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
 
-    def invoke(self, t: CompletionRequest) -> CompletionResponse:
+    async def invoke(self, t: CompletionRequest) -> CompletionResponse:
         raw_request = dict(
             model=self.DEFAULT_MODEL_NAME,
             prompt=t.v,

ommlds/minichain/backends/impls/openai/embedding.py

@@ -28,7 +28,7 @@ class OpenaiEmbeddingService:
         with tv.consume(*configs) as cc:
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
 
-    def invoke(self, request: EmbeddingRequest) -> EmbeddingResponse:
+    async def invoke(self, request: EmbeddingRequest) -> EmbeddingResponse:
         raw_request = dict(
             model=self.model,
             input=check.isinstance(request.v, str),

ommlds/minichain/backends/impls/openai/stream.py

@@ -1,12 +1,17 @@
+"""
+https://platform.openai.com/docs/api-reference/responses-streaming
+"""
 import typing as ta
 
 from omlish import check
+from omlish import marshal as msh
 from omlish import typedvalues as tv
 from omlish.formats import json
 from omlish.http import all as http
 from omlish.http import sse
 from omlish.io.buffers import DelimitingBuffer
 
+from .....backends.openai.protocol.chatcompletion.chunk import ChatCompletionChunk
 from ....chat.choices.services import ChatChoicesOutputs
 from ....chat.stream.services import ChatChoicesStreamRequest
 from ....chat.stream.services import ChatChoicesStreamResponse
@@ -43,7 +48,7 @@ class OpenaiChatChoicesStreamService:
 
     READ_CHUNK_SIZE = 64 * 1024
 
-    def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
+    async def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
         # check.isinstance(request, ChatRequest)
 
         rh = OpenaiChatRequestHandler(
@@ -99,6 +104,9 @@ class OpenaiChatChoicesStreamService:
 
                                 check.state(sj['object'] == 'chat.completion.chunk')
 
+                                ccc = msh.unmarshal(sj, ChatCompletionChunk)  # noqa
+                                # print(ccc)
+
                                 # FIXME: stop reason
                                 if not sj['choices']:
                                     continue

ommlds/minichain/backends/impls/tinygrad/chat.py

@@ -116,7 +116,7 @@ class BaseTinygradLlama3ChatService(lang.ExitStacked, lang.Abstract):
 # )
 @static_check_is_chat_choices_service
 class TinygradLlama3ChatChoicesService(BaseTinygradLlama3ChatService):
-    def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
+    async def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
         llm = self._load_model()
         toks = _prepare_toks(llm, request.v, request.options)
 
@@ -136,7 +136,7 @@ class TinygradLlama3ChatChoicesService(BaseTinygradLlama3ChatService):
 # )
 @static_check_is_chat_choices_stream_service
 class TinygradLlama3ChatChoicesStreamService(BaseTinygradLlama3ChatService):
-    def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
+    async def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
         with UseResources.or_new(request.options) as rs:
             llm = self._load_model()
             toks = _prepare_toks(

ommlds/minichain/backends/impls/transformers/sentence.py

@@ -33,7 +33,7 @@ class SentenceTransformersEmbeddingService:
         with tv.consume(*configs) as cc:
             self._model_path = cc.pop(ModelPath(self.DEFAULT_MODEL))
 
-    def invoke(self, request: EmbeddingRequest) -> EmbeddingResponse:
+    async def invoke(self, request: EmbeddingRequest) -> EmbeddingResponse:
         mdl = stfm.SentenceTransformer(
             self._model_path.v,
         )

ommlds/minichain/backends/impls/transformers/transformers.py

@@ -59,7 +59,7 @@ class TransformersCompletionService(lang.ExitStacked):
             self._pipeline_kwargs = cc.pop(TransformersPipelineKwargs, [])
             self._huggingface_hub_token = HuggingfaceHubToken.pop_secret(cc, env='HUGGINGFACE_HUB_TOKEN')
 
-    def invoke(self, request: CompletionRequest) -> CompletionResponse:
+    async def invoke(self, request: CompletionRequest) -> CompletionResponse:
         pkw: dict[str, ta.Any] = dict(
             model=self._model_path.v,
             device='mps' if sys.platform == 'darwin' else 'cuda',
@@ -162,7 +162,7 @@ class TransformersChatChoicesService(lang.ExitStacked):
             **pkw,
         )
 
-    def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
+    async def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
         check.empty(request.options)
 
         pipeline = self._load_pipeline()

ommlds/minichain/chat/choices/adapters.py

@@ -17,6 +17,6 @@ from .types import ChatChoicesOutputs
 class ChatChoicesServiceChatService:
     service: ChatChoicesService
 
-    def invoke(self, request: ChatRequest) -> Response[AiMessage, ChatChoicesOutputs]:
-        resp = self.service.invoke(request)
+    async def invoke(self, request: ChatRequest) -> Response[AiMessage, ChatChoicesOutputs]:
+        resp = await self.service.invoke(request)
         return Response(check.single(resp.v).m, resp.outputs)

ommlds/minichain/chat/choices/services.py

@@ -36,5 +36,5 @@ def static_check_is_chat_choices_service[T: ChatChoicesService](t: type[T]) -> t
 @static_check_is_chat_choices_service
 class AbstractChatChoicesService(lang.Abstract):
     @abc.abstractmethod
-    def invoke(self, request: ChatChoicesRequest) -> ChatChoicesResponse:
+    def invoke(self, request: ChatChoicesRequest) -> ta.Awaitable[ChatChoicesResponse]:
         raise NotImplementedError

ommlds/minichain/chat/history.py

@@ -63,9 +63,9 @@ class HistoryAddingChatService:
         self._inner = inner
         self._history = history
 
-    def invoke(self, request: ChatRequest) -> ChatResponse:
+    async def invoke(self, request: ChatRequest) -> ChatResponse:
         new_req = dc.replace(request, v=[*self._history.get(), *request.v])
-        response = self._inner.invoke(new_req)
+        response = await self._inner.invoke(new_req)
         self._history.add(
             *request.v,
             response.v,

ommlds/minichain/chat/services.py

@@ -36,5 +36,5 @@ def static_check_is_chat_service[T: ChatService](t: type[T]) -> type[T]:
 @static_check_is_chat_service
 class AbstractChatService(lang.Abstract):
     @abc.abstractmethod
-    def invoke(self, request: ChatRequest) -> ChatResponse:
+    def invoke(self, request: ChatRequest) -> ta.Awaitable[ChatResponse]:
         raise NotImplementedError

ommlds/minichain/chat/stream/adapters.py

@@ -27,13 +27,13 @@ class ChatChoicesStreamServiceChatChoicesService:
         parts: list[str]
         trs: list[ToolExecRequest]
 
-    def invoke(self, request: ChatChoicesRequest) -> Response[
+    async def invoke(self, request: ChatChoicesRequest) -> Response[
         AiChoices,
         ChatChoicesOutputs | ChatChoicesStreamOutputs,
     ]:
         lst: list[ChatChoicesStreamServiceChatChoicesService._Choice] = []
 
-        resp = self.service.invoke(request)
+        resp = await self.service.invoke(request)
         with resp.v as resp_v:
             i = -1  # noqa
             for i, cs in enumerate(resp_v):

ommlds/minichain/chat/stream/services.py

@@ -41,7 +41,7 @@ def static_check_is_chat_choices_stream_service[T: ChatChoicesStreamService](t:
 @static_check_is_chat_choices_stream_service
 class AbstractChatChoicesStreamService(lang.Abstract):
     @abc.abstractmethod
-    def invoke(self, request: ChatChoicesStreamRequest) -> ChatChoicesStreamResponse:
+    def invoke(self, request: ChatChoicesStreamRequest) -> ta.Awaitable[ChatChoicesStreamResponse]:
         raise NotImplementedError
 
 

ommlds/minichain/chat/transforms/services.py

@@ -19,10 +19,10 @@ class RequestChatTransformingChatService:
     ct: ChatTransform
     svc: ChatService
 
-    def invoke(self, request: ChatRequest) -> ChatResponse:
+    async def invoke(self, request: ChatRequest) -> ChatResponse:
         new_chat = self.ct.transform_chat(request.v)
         new_req = dc.replace(request, v=new_chat)
-        return self.svc.invoke(new_req)
+        return await self.svc.invoke(new_req)
 
 
 #
@@ -34,7 +34,7 @@ class ResponseMessageTransformingChatService:
     mt: MessageTransform[AiMessage]
     svc: ChatService
 
-    def invoke(self, request: ChatRequest) -> ChatResponse:
-        orig_resp = self.svc.invoke(request)
+    async def invoke(self, request: ChatRequest) -> ChatResponse:
+        orig_resp = await self.svc.invoke(request)
         new_msg = self.mt.transform_message(orig_resp.v)
         return dc.replace(orig_resp, v=check.isinstance(new_msg, AiMessage))

ommlds/minichain/services/facades.py

@@ -44,15 +44,15 @@ class ServiceFacade(
         ],
     ]
 
-    def invoke(self, request: Request[RequestV, OptionT]) -> Response[ResponseV, OutputT]:
+    def invoke(self, request: Request[RequestV, OptionT]) -> ta.Awaitable[Response[ResponseV, OutputT]]:
         return self.service.invoke(request)
 
     @ta.overload
-    def __call__(self, request: Request[RequestV, OptionT]) -> Response[ResponseV, OutputT]:
+    def __call__(self, request: Request[RequestV, OptionT]) -> ta.Awaitable[Response[ResponseV, OutputT]]:
         ...
 
     @ta.overload
-    def __call__(self, v: RequestV, *options: OptionT) -> Response[ResponseV, OutputT]:
+    def __call__(self, v: RequestV, *options: OptionT) -> ta.Awaitable[Response[ResponseV, OutputT]]:
         ...
 
     def __call__(self, o, *args):

ommlds/minichain/services/services.py

@@ -11,4 +11,4 @@ from .responses import ResponseT_co
 
 @ta.runtime_checkable
 class Service(lang.ProtocolForbiddenAsBaseClass, ta.Protocol[RequestT_contra, ResponseT_co]):
-    def invoke(self, request: RequestT_contra) -> ResponseT_co: ...
+    def invoke(self, request: RequestT_contra) -> ta.Awaitable[ResponseT_co]: ...

ommlds/minichain/stream/wrap.py

@@ -43,9 +43,9 @@ class WrappedStreamService(ta.Generic[StreamRequestT, V, OutputT, StreamOutputT]
 
     #
 
-    def invoke(self, request: StreamRequestT) -> StreamResponse[V, OutputT, StreamOutputT]:
+    async def invoke(self, request: StreamRequestT) -> StreamResponse[V, OutputT, StreamOutputT]:
         with Resources.new() as rs:
-            in_response = self._inner.invoke(self._process_request(request))
+            in_response = await self._inner.invoke(self._process_request(request))
             in_vs: ResponseGenerator[V, OutputT] = rs.enter_context(in_response.v)
             out_vs = self._process_vs(in_vs)
 

ommlds/server/server.py

@@ -47,10 +47,10 @@ class McServerHandler(HttpHandler_):
 
         log.info('Server got prompt: %s', prompt)
 
-        resp = self.llm.invoke(mc.ChatChoicesRequest(
+        resp = lang.sync_await(self.llm.invoke(mc.ChatChoicesRequest(
             [mc.UserMessage(prompt)],
             # Temperature(.1),
-        ))
+        )))
         resp_txt = check.isinstance(resp.v[0].m.c, str)
 
         log.info('Server got response: %s', resp_txt)

ommlds/tools/git.py

@@ -64,10 +64,10 @@ class OpenaiGitAiBackend(GitAiBackend['OpenaiGitAiBackend.Config']):
 
         llm = OpenaiChatChoicesService()
 
-        resp = llm.invoke(mc.ChatChoicesRequest(
+        resp = lang.sync_await(llm.invoke(mc.ChatChoicesRequest(
             [mc.UserMessage(prompt)],
             # FIXME:  *((MaxTokens(self._config.max_tokens),) if self._config.max_tokens is not None else ()),
-        ))
+        )))
         return check.not_empty(check.isinstance(resp.v[0].m.c, str))
 
 
@@ -93,10 +93,10 @@ class MlxGitAiBackend(GitAiBackend['MlxGitAiBackend.Config']):
 
     def _run_prompt(self, prompt: str) -> str:
         with mc_mlx_chat.MlxChatChoicesService(mc.ModelRepo.parse(self._config.model)) as llm:
-            resp = llm.invoke(mc.ChatChoicesRequest(
+            resp = lang.sync_await(llm.invoke(mc.ChatChoicesRequest(
                 [mc.UserMessage(prompt)],
                 # FIXME: *((MaxTokens(self._config.max_tokens),) if self._config.max_tokens is not None else ()),
-            ))
+            )))
             text = check.not_empty(check.isinstance(resp.v[0].m.c, str))
 
             text = _strip_markdown_code_block(text)

{ommlds-0.0.0.dev441.dist-info → ommlds-0.0.0.dev442.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ommlds
-Version: 0.0.0.dev441
+Version: 0.0.0.dev442
 Summary: ommlds
 Author: wrmsr
 License-Expression: BSD-3-Clause
@@ -14,8 +14,8 @@ Classifier: Programming Language :: Python :: 3.13
 Requires-Python: >=3.13
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: omdev==0.0.0.dev441
-Requires-Dist: omlish==0.0.0.dev441
+Requires-Dist: omdev==0.0.0.dev442
+Requires-Dist: omlish==0.0.0.dev442
 Provides-Extra: all
 Requires-Dist: llama-cpp-python~=0.3; extra == "all"
 Requires-Dist: mlx~=0.29; extra == "all"
@@ -32,7 +32,7 @@ Requires-Dist: numpy>=1.26; extra == "all"
 Requires-Dist: pytesseract~=0.3; extra == "all"
 Requires-Dist: rapidocr-onnxruntime~=1.4; extra == "all"
 Requires-Dist: pillow~=11.3; extra == "all"
-Requires-Dist: ddgs~=9.5; extra == "all"
+Requires-Dist: ddgs~=9.6; extra == "all"
 Requires-Dist: mwparserfromhell~=0.7; extra == "all"
 Requires-Dist: wikitextparser~=0.56; extra == "all"
 Requires-Dist: lxml>=5.3; python_version < "3.13" and extra == "all"
@@ -57,7 +57,7 @@ Requires-Dist: rapidocr-onnxruntime~=1.4; extra == "ocr"
 Provides-Extra: pillow
 Requires-Dist: pillow~=11.3; extra == "pillow"
 Provides-Extra: search
-Requires-Dist: ddgs~=9.5; extra == "search"
+Requires-Dist: ddgs~=9.6; extra == "search"
 Provides-Extra: wiki
 Requires-Dist: mwparserfromhell~=0.7; extra == "wiki"
 Requires-Dist: wikitextparser~=0.56; extra == "wiki"