ommlds 0.0.0.dev469__tar.gz → 0.0.0.dev471__tar.gz

{ommlds-0.0.0.dev469/ommlds.egg-info → ommlds-0.0.0.dev471}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ommlds
-Version: 0.0.0.dev469
+Version: 0.0.0.dev471
 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.dev469
-Requires-Dist: omlish==0.0.0.dev469
+Requires-Dist: omdev==0.0.0.dev471
+Requires-Dist: omlish==0.0.0.dev471
 Provides-Extra: all
 Requires-Dist: llama-cpp-python~=0.3; extra == "all"
 Requires-Dist: mlx~=0.29; extra == "all"

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/.omlish-manifests.json

@@ -714,7 +714,7 @@
     "module": ".tools.git",
     "attr": null,
     "file": "ommlds/tools/git.py",
-    "line": 186,
+    "line": 189,
     "value": {
       "!omdev.tools.git.messages.GitMessageGeneratorManifest": {
         "module": "ommlds.tools.git",

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/cli/sessions/chat/backends/catalog.py

@@ -15,33 +15,38 @@ from .types import ServiceT
 ##
 
 
-class _CatalogBackendProvider(BackendProvider[ServiceT], lang.Abstract):
+class CatalogBackendProvider(BackendProvider[ServiceT], lang.Abstract):
+    class Instantiator(lang.Func2['mc.BackendCatalog.Backend', BackendConfigs | None, ta.Awaitable[ta.Any]]):
+        pass
+
     def __init__(
             self,
             *,
             name: BackendName,
             catalog: 'mc.BackendCatalog',
             configs: BackendConfigs | None = None,
+            instantiator: Instantiator | None = None,
     ) -> None:
         super().__init__()
 
         self._name = name
         self._catalog = catalog
         self._configs = configs
+        if instantiator is None:
+            instantiator = CatalogBackendProvider.Instantiator(lang.as_async(lambda be, cfgs: be.factory(*cfgs or [])))
+        self._instantiator = instantiator
 
     @contextlib.asynccontextmanager
     async def _provide_backend(self, cls: type[ServiceT]) -> ta.AsyncIterator[ServiceT]:
+        be = self._catalog.get_backend(cls, self._name)
+
         service: ServiceT
-        async with lang.async_or_sync_maybe_managing(self._catalog.new_backend(
-                cls,
-                self._name,
-                *(self._configs or []),
-        )) as service:
+        async with lang.async_or_sync_maybe_managing(await self._instantiator(be, self._configs)) as service:
             yield service
 
 
 class CatalogChatChoicesServiceBackendProvider(
-    _CatalogBackendProvider['mc.ChatChoicesService'],
+    CatalogBackendProvider['mc.ChatChoicesService'],
     ChatChoicesServiceBackendProvider,
 ):
     def provide_backend(self) -> ta.AsyncContextManager['mc.ChatChoicesService']:
@@ -49,7 +54,7 @@ class CatalogChatChoicesServiceBackendProvider(
 
 
 class CatalogChatChoicesStreamServiceBackendProvider(
-    _CatalogBackendProvider['mc.ChatChoicesStreamService'],
+    CatalogBackendProvider['mc.ChatChoicesStreamService'],
     ChatChoicesStreamServiceBackendProvider,
 ):
     def provide_backend(self) -> ta.AsyncContextManager['mc.ChatChoicesStreamService']:

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/cli/sessions/chat/backends/inject.py

@@ -1,6 +1,9 @@
+import typing as ta
+
 from omlish import inject as inj
 from omlish import lang
 
+from ..... import minichain as mc
 from .injection import backend_configs
 
 
@@ -34,4 +37,16 @@ def bind_backends(
 
     #
 
+    async def catalog_backend_instantiator_provider(injector: inj.AsyncInjector) -> _catalog.CatalogBackendProvider.Instantiator:  # noqa
+        async def inner(be: 'mc.BackendCatalog.Backend', cfgs: _types.BackendConfigs | None) -> ta.Any:
+            kwt = inj.build_kwargs_target(be.factory, non_strict=True)
+            kw = await injector.provide_kwargs(kwt)
+            return be.factory(*cfgs or [], **kw)
+
+        return _catalog.CatalogBackendProvider.Instantiator(inner)
+
+    els.append(inj.bind(_catalog.CatalogBackendProvider.Instantiator, to_async_fn=catalog_backend_instantiator_provider))  # noqa
+
+    #
+
     return inj.as_elements(*els)

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/catalogs/base.py

@@ -21,7 +21,13 @@ class BackendCatalog(lang.Abstract):
     def get_backend(self, service_cls: type[T], name: str) -> Backend:
         raise NotImplementedError
 
-    def new_backend(self, service_cls: ta.Any, name: str, *args: ta.Any, **kwargs: ta.Any) -> ta.Any:
+    def new_backend(
+            self,
+            service_cls: ta.Any,
+            name: str,
+            *args: ta.Any,
+            **kwargs: ta.Any,
+    ) -> ta.Any:
         be = self.get_backend(service_cls, name)
         return be.factory(*be.configs or [], *args, **kwargs)
 

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/anthropic/chat.py

@@ -46,9 +46,12 @@ class AnthropicChatChoicesService:
     def __init__(
             self,
             *configs: ApiKey | ModelName,
+            http_client: http.AsyncHttpClient | None = None,
     ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._api_key = check.not_none(ApiKey.pop_secret(cc, env='ANTHROPIC_API_KEY'))
             self._model_name = cc.pop(self.DEFAULT_MODEL_NAME)
@@ -88,7 +91,7 @@ class AnthropicChatChoicesService:
 
         raw_request = msh.marshal(a_req)
 
-        raw_response = http.request(
+        raw_response = await http.async_request(
             'https://api.anthropic.com/v1/messages',
             headers={
                 http.consts.HEADER_CONTENT_TYPE: http.consts.CONTENT_TYPE_JSON,
@@ -96,6 +99,7 @@ class AnthropicChatChoicesService:
                 b'anthropic-version': b'2023-06-01',
             },
             data=json.dumps(raw_request).encode('utf-8'),
+            client=self._http_client,
         )
 
         response = json.loads(check.not_none(raw_response.data).decode('utf-8'))

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/anthropic/stream.py

@@ -39,9 +39,15 @@ from .protocol import build_protocol_tool
 # )
 @static_check_is_chat_choices_stream_service
 class AnthropicChatChoicesStreamService:
-    def __init__(self, *configs: Config) -> None:
+    def __init__(
+            self,
+            *configs: Config,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._model_name = cc.pop(AnthropicChatChoicesService.DEFAULT_MODEL_NAME)
             self._api_key = check.not_none(ApiKey.pop_secret(cc, env='ANTHROPIC_API_KEY'))
@@ -84,8 +90,8 @@ class AnthropicChatChoicesStreamService:
         )
 
         async with UseResources.or_new(request.options) as rs:
-            http_client = rs.enter_context(http.client())
-            http_response = rs.enter_context(http_client.stream_request(http_request))
+            http_client = await rs.enter_async_context(http.manage_async_client(self._http_client))
+            http_response = await rs.enter_async_context(await http_client.stream_request(http_request))
 
             async def inner(sink: StreamResponseSink[AiChoicesDeltas]) -> ta.Sequence[ChatChoicesOutputs] | None:
                 msg_start: AnthropicSseDecoderEvents.MessageStart | None = None
@@ -95,7 +101,7 @@ class AnthropicChatChoicesStreamService:
                 db = DelimitingBuffer([b'\r', b'\n', b'\r\n'])
                 sd = sse.SseDecoder()
                 while True:
-                    b = http_response.stream.read1(self.READ_CHUNK_SIZE)
+                    b = await http_response.stream.read1(self.READ_CHUNK_SIZE)
                     for l in db.feed(b):
                         if isinstance(l, DelimitingBuffer.Incomplete):
                             # FIXME: handle

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/google/chat.py

@@ -40,9 +40,15 @@ from .tools import build_tool_spec_schema
 class GoogleChatChoicesService:
     DEFAULT_MODEL_NAME: ta.ClassVar[ModelName] = ModelName(check.not_none(MODEL_NAMES.default))
 
-    def __init__(self, *configs: ApiKey | ModelName) -> None:
+    def __init__(
+            self,
+            *configs: ApiKey | ModelName,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._model_name = cc.pop(self.DEFAULT_MODEL_NAME)
             self._api_key = ApiKey.pop_secret(cc, env='GEMINI_API_KEY')
@@ -149,11 +155,12 @@ class GoogleChatChoicesService:
 
         model_name = MODEL_NAMES.resolve(self._model_name.v)
 
-        resp = http.request(
+        resp = await http.async_request(
             f'{self.BASE_URL.rstrip("/")}/{model_name}:generateContent?key={key}',
             headers={'Content-Type': 'application/json'},
             data=json.dumps_compact(req_dct).encode('utf-8'),
             method='POST',
+            client=self._http_client,
         )
 
         resp_dct = json.loads(check.not_none(resp.data).decode('utf-8'))

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/google/search.py

@@ -82,12 +82,16 @@ class CseSearchService:
             self,
             cse_id: str | None = None,
             cse_api_key: str | None = None,
+            *,
+            http_client: http.AsyncHttpClient | None = None,
     ) -> None:
         super().__init__()
 
         self._cse_id = cse_id
         self._cse_api_key = cse_api_key
 
+        self._http_client = http_client
+
     async def invoke(
             self,
             request: SearchRequest,
@@ -97,8 +101,9 @@ class CseSearchService:
             cx=check.non_empty_str(self._cse_id),
             q=request.v,
         ))
-        resp = http.request(
+        resp = await http.async_request(
             f'https://www.googleapis.com/customsearch/v1?{qs}',
+            client=self._http_client,
         )
         out = check.not_none(resp.data)
 

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/google/stream.py

@@ -46,9 +46,15 @@ from .tools import build_tool_spec_schema
 class GoogleChatChoicesStreamService:
     DEFAULT_MODEL_NAME: ta.ClassVar[ModelName] = ModelName(check.not_none(MODEL_NAMES.default))
 
-    def __init__(self, *configs: ApiKey | ModelName) -> None:
+    def __init__(
+            self,
+            *configs: ApiKey | ModelName,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._model_name = cc.pop(self.DEFAULT_MODEL_NAME)
             self._api_key = ApiKey.pop_secret(cc, env='GEMINI_API_KEY')
@@ -163,13 +169,13 @@ class GoogleChatChoicesStreamService:
         )
 
         async with UseResources.or_new(request.options) as rs:
-            http_client = rs.enter_context(http.client())
-            http_response = rs.enter_context(http_client.stream_request(http_request))
+            http_client = await rs.enter_async_context(http.manage_async_client(self._http_client))
+            http_response = await rs.enter_async_context(await http_client.stream_request(http_request))
 
             async def inner(sink: StreamResponseSink[AiChoicesDeltas]) -> ta.Sequence[ChatChoicesOutputs] | None:
                 db = DelimitingBuffer([b'\r', b'\n', b'\r\n'])
                 while True:
-                    b = http_response.stream.read1(self.READ_CHUNK_SIZE)
+                    b = await http_response.stream.read1(self.READ_CHUNK_SIZE)
                     for bl in db.feed(b):
                         if isinstance(bl, DelimitingBuffer.Incomplete):
                             # FIXME: handle

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/mistral.py

@@ -40,10 +40,16 @@ class MistralChatChoicesService:
         AiMessage: 'assistant',
     }
 
-    def __init__(self, *, api_key: str | None = None) -> None:
+    def __init__(
+            self,
+            *,
+            api_key: str | None = None,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
         self._api_key = api_key
+        self._http_client = http_client
 
     def _get_msg_content(self, m: Message) -> str | None:
         if isinstance(m, AiMessage):
@@ -73,7 +79,7 @@ class MistralChatChoicesService:
             ],
         }
 
-        resp = http.request(
+        resp = await http.async_request(
             'https://api.mistral.ai/v1/chat/completions',
             method='POST',
             data=json.dumps_compact(req_dct).encode('utf-8'),
@@ -82,6 +88,7 @@ class MistralChatChoicesService:
                 'Accept': 'application/json',
                 'Authorization': f'Bearer {key}',
             },
+            client=self._http_client,
         )
 
         if resp.status == 429:

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/ollama/chat.py

@@ -66,7 +66,7 @@ class BaseOllamaChatChoicesService(lang.Abstract):
 
     #
 
-    ROLE_MAP: ta.ClassVar[ta.Mapping[type[Message], pt.Role]] = {
+    ROLE_MAP: ta.ClassVar[ta.Mapping[type[Message], pt.Role]] = {  # noqa
         SystemMessage: 'system',
         UserMessage: 'user',
         AiMessage: 'assistant',

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/openai/chat.py

@@ -42,9 +42,15 @@ from .names import MODEL_NAMES
 class OpenaiChatChoicesService:
     DEFAULT_MODEL_NAME: ta.ClassVar[ModelName] = ModelName(check.not_none(MODEL_NAMES.default))
 
-    def __init__(self, *configs: ApiKey | ModelName | DefaultOptions) -> None:
+    def __init__(
+            self,
+            *configs: ApiKey | ModelName | DefaultOptions,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._model_name = cc.pop(self.DEFAULT_MODEL_NAME)
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
@@ -68,13 +74,14 @@ class OpenaiChatChoicesService:
 
         raw_request = msh.marshal(rh.oai_request())
 
-        http_response = http.request(
+        http_response = await http.async_request(
             'https://api.openai.com/v1/chat/completions',
             headers={
                 http.consts.HEADER_CONTENT_TYPE: http.consts.CONTENT_TYPE_JSON,
                 http.consts.HEADER_AUTH: http.consts.format_bearer_auth_header(check.not_none(self._api_key).reveal()),
             },
             data=json.dumps(raw_request).encode('utf-8'),
+            client=self._http_client,
         )
 
         raw_response = json.loads(check.not_none(http_response.data).decode('utf-8'))

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/openai/completion.py

@@ -23,9 +23,15 @@ from ....standard import ApiKey
 class OpenaiCompletionService:
     DEFAULT_MODEL_NAME: ta.ClassVar[str] = 'gpt-3.5-turbo-instruct'
 
-    def __init__(self, *configs: Config) -> None:
+    def __init__(
+            self,
+            *configs: Config,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
 
@@ -41,13 +47,14 @@ class OpenaiCompletionService:
             stream=False,
         )
 
-        raw_response = http.request(
+        raw_response = await http.async_request(
             'https://api.openai.com/v1/completions',
             headers={
                 http.consts.HEADER_CONTENT_TYPE: http.consts.CONTENT_TYPE_JSON,
                 http.consts.HEADER_AUTH: http.consts.format_bearer_auth_header(check.not_none(self._api_key).reveal()),
             },
             data=json.dumps(raw_request).encode('utf-8'),
+            client=self._http_client,
         )
 
         response = json.loads(check.not_none(raw_response.data).decode('utf-8'))

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/openai/embedding.py

@@ -22,9 +22,15 @@ from ....vectors.types import Vector
 class OpenaiEmbeddingService:
     model = 'text-embedding-3-small'
 
-    def __init__(self, *configs: Config) -> None:
+    def __init__(
+            self,
+            *configs: Config,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
 
@@ -34,13 +40,14 @@ class OpenaiEmbeddingService:
             input=check.isinstance(request.v, str),
         )
 
-        raw_response = http.request(
+        raw_response = await http.async_request(
             'https://api.openai.com/v1/embeddings',
             headers={
                 http.consts.HEADER_CONTENT_TYPE: http.consts.CONTENT_TYPE_JSON,
                 http.consts.HEADER_AUTH: http.consts.format_bearer_auth_header(check.not_none(self._api_key).reveal()),
             },
             data=json.dumps(raw_request).encode('utf-8'),
+            client=self._http_client,
         )
 
         response = json.loads(check.not_none(raw_response.data).decode('utf-8'))

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/minichain/backends/impls/openai/stream.py

@@ -41,9 +41,15 @@ from .names import MODEL_NAMES
 # )
 @static_check_is_chat_choices_stream_service
 class OpenaiChatChoicesStreamService:
-    def __init__(self, *configs: Config) -> None:
+    def __init__(
+            self,
+            *configs: Config,
+            http_client: http.AsyncHttpClient | None = None,
+    ) -> None:
         super().__init__()
 
+        self._http_client = http_client
+
         with tv.consume(*configs) as cc:
             self._model_name = cc.pop(OpenaiChatChoicesService.DEFAULT_MODEL_NAME)
             self._api_key = ApiKey.pop_secret(cc, env='OPENAI_API_KEY')
@@ -81,14 +87,14 @@ class OpenaiChatChoicesStreamService:
         )
 
         async with UseResources.or_new(request.options) as rs:
-            http_client = rs.enter_context(http.client())
-            http_response = rs.enter_context(http_client.stream_request(http_request))
+            http_client = await rs.enter_async_context(http.manage_async_client(self._http_client))
+            http_response = await rs.enter_async_context(await http_client.stream_request(http_request))
 
             async def inner(sink: StreamResponseSink[AiChoicesDeltas]) -> ta.Sequence[ChatChoicesOutputs]:
                 db = DelimitingBuffer([b'\r', b'\n', b'\r\n'])
                 sd = sse.SseDecoder()
                 while True:
-                    b = http_response.stream.read1(self.READ_CHUNK_SIZE)
+                    b = await http_response.stream.read1(self.READ_CHUNK_SIZE)
                     for l in db.feed(b):
                         if isinstance(l, DelimitingBuffer.Incomplete):
                             # FIXME: handle

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds/tools/git.py

@@ -16,6 +16,7 @@ from omdev.tools.git.messages import GitMessageGenerator
 from omlish import check
 from omlish import lang
 from omlish.configs.classes import Configurable
+from omlish.http import all as http
 from omlish.subprocesses.sync import subprocesses
 
 from .. import minichain as mc
@@ -76,7 +77,9 @@ class OpenaiGitAiBackend(GitAiBackend['OpenaiGitAiBackend.Config']):
             if (sec := load_secrets().try_get(key.lower())) is not None:
                 os.environ[key] = sec.reveal()
 
-        llm = OpenaiChatChoicesService()
+        llm = OpenaiChatChoicesService(
+            http_client=http.SyncAsyncHttpClient(http.client()),
+        )
 
         resp = lang.sync_await(llm.invoke(mc.ChatChoicesRequest(
             [mc.UserMessage(prompt)],

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471/ommlds.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ommlds
-Version: 0.0.0.dev469
+Version: 0.0.0.dev471
 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.dev469
-Requires-Dist: omlish==0.0.0.dev469
+Requires-Dist: omdev==0.0.0.dev471
+Requires-Dist: omlish==0.0.0.dev471
 Provides-Extra: all
 Requires-Dist: llama-cpp-python~=0.3; extra == "all"
 Requires-Dist: mlx~=0.29; extra == "all"

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds.egg-info/SOURCES.txt

@@ -5,7 +5,6 @@ pyproject.toml
 ommlds/.omlish-manifests.json
 ommlds/__about__.py
 ommlds/__init__.py
-ommlds/huggingface.py
 ommlds.egg-info/PKG-INFO
 ommlds.egg-info/SOURCES.txt
 ommlds.egg-info/dependency_links.txt
@@ -18,6 +17,7 @@ ommlds/_hacks/names.py
 ommlds/_hacks/params.py
 ommlds/_hacks/patches.py
 ommlds/backends/__init__.py
+ommlds/backends/huggingface.py
 ommlds/backends/anthropic/__init__.py
 ommlds/backends/anthropic/protocol/__init__.py
 ommlds/backends/anthropic/protocol/_marshal.py

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/ommlds.egg-info/requires.txt

@@ -1,5 +1,5 @@
-omdev==0.0.0.dev469
-omlish==0.0.0.dev469
+omdev==0.0.0.dev471
+omlish==0.0.0.dev471
 
 [all]
 llama-cpp-python~=0.3

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev471}/pyproject.toml

@@ -13,7 +13,7 @@ urls = {source = 'https://github.com/wrmsr/omlish'}
 license = 'BSD-3-Clause'
 readme = 'README.md'
 requires-python = '>=3.13'
-version = '0.0.0.dev469'
+version = '0.0.0.dev471'
 classifiers = [
     'Development Status :: 2 - Pre-Alpha',
     'Intended Audience :: Developers',
@@ -24,8 +24,8 @@ classifiers = [
 ]
 description = 'ommlds'
 dependencies = [
-    'omdev == 0.0.0.dev469',
-    'omlish == 0.0.0.dev469',
+    'omdev == 0.0.0.dev471',
+    'omlish == 0.0.0.dev471',
 ]
 
 [project.optional-dependencies]