ommlds 0.0.0.dev469__tar.gz → 0.0.0.dev475__tar.gz

{ommlds-0.0.0.dev469/ommlds.egg-info → ommlds-0.0.0.dev475}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ommlds
-Version: 0.0.0.dev469
+Version: 0.0.0.dev475
 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.dev475
+Requires-Dist: omlish==0.0.0.dev475
 Provides-Extra: all
 Requires-Dist: llama-cpp-python~=0.3; extra == "all"
 Requires-Dist: mlx~=0.29; extra == "all"
@@ -27,12 +27,13 @@ Requires-Dist: torch~=2.9; extra == "all"
 Requires-Dist: transformers~=4.57; extra == "all"
 Requires-Dist: sentence-transformers~=5.1; extra == "all"
 Requires-Dist: huggingface-hub~=0.36; extra == "all"
-Requires-Dist: datasets~=4.3; extra == "all"
+Requires-Dist: datasets~=4.4; extra == "all"
+Requires-Dist: regex>=2025.0; extra == "all"
 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~=12.0; extra == "all"
-Requires-Dist: ddgs~=9.6; extra == "all"
+Requires-Dist: ddgs~=9.8; 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"
@@ -48,7 +49,9 @@ Requires-Dist: transformers~=4.57; extra == "backends"
 Requires-Dist: sentence-transformers~=5.1; extra == "backends"
 Provides-Extra: huggingface
 Requires-Dist: huggingface-hub~=0.36; extra == "huggingface"
-Requires-Dist: datasets~=4.3; extra == "huggingface"
+Requires-Dist: datasets~=4.4; extra == "huggingface"
+Provides-Extra: nanochat
+Requires-Dist: regex>=2025.0; extra == "nanochat"
 Provides-Extra: numpy
 Requires-Dist: numpy>=1.26; extra == "numpy"
 Provides-Extra: ocr
@@ -57,7 +60,7 @@ Requires-Dist: rapidocr-onnxruntime~=1.4; extra == "ocr"
 Provides-Extra: pillow
 Requires-Dist: pillow~=12.0; extra == "pillow"
 Provides-Extra: search
-Requires-Dist: ddgs~=9.6; extra == "search"
+Requires-Dist: ddgs~=9.8; extra == "search"
 Provides-Extra: wiki
 Requires-Dist: mwparserfromhell~=0.7; extra == "wiki"
 Requires-Dist: wikitextparser~=0.56; extra == "wiki"

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev475}/ommlds/.omlish-manifests.json

@@ -91,6 +91,52 @@
       }
     }
   },
+  {
+    "module": ".minichain.backends.impls.dummy.chat",
+    "attr": null,
+    "file": "ommlds/minichain/backends/impls/dummy/chat.py",
+    "line": 26,
+    "value": {
+      "!.minichain.registries.manifests.RegistryManifest": {
+        "module": "ommlds.minichain.backends.impls.dummy.chat",
+        "attr": "DummyChatChoicesService",
+        "name": "dummy",
+        "aliases": null,
+        "type": "ChatChoicesService"
+      }
+    }
+  },
+  {
+    "module": ".minichain.backends.impls.dummy.chat",
+    "attr": null,
+    "file": "ommlds/minichain/backends/impls/dummy/chat.py",
+    "line": 39,
+    "value": {
+      "!.minichain.registries.manifests.RegistryManifest": {
+        "module": "ommlds.minichain.backends.impls.dummy.chat",
+        "attr": "DummyChatChoicesStreamService",
+        "name": "dummy",
+        "aliases": null,
+        "type": "ChatChoicesStreamService"
+      }
+    }
+  },
+  {
+    "module": ".minichain.backends.impls.dummy.chat",
+    "attr": null,
+    "file": "ommlds/minichain/backends/impls/dummy/chat.py",
+    "line": 63,
+    "value": {
+      "!.minichain.backends.strings.manifests.BackendStringsManifest": {
+        "service_cls_names": [
+          "ChatChoicesService",
+          "ChatChoicesStreamService"
+        ],
+        "backend_name": "dummy",
+        "model_names": null
+      }
+    }
+  },
   {
     "module": ".minichain.backends.impls.google.chat",
     "attr": null,
@@ -468,7 +514,7 @@
       "!.minichain.registries.manifests.RegistryManifest": {
         "module": "ommlds.minichain.backends.impls.tinygrad.chat",
         "attr": "TinygradLlama3ChatChoicesService",
-        "name": "tinygrad_llama3",
+        "name": "tinygrad-llama3",
         "aliases": null,
         "type": "ChatChoicesService"
       }
@@ -483,7 +529,7 @@
       "!.minichain.registries.manifests.RegistryManifest": {
         "module": "ommlds.minichain.backends.impls.tinygrad.chat",
         "attr": "TinygradLlama3ChatChoicesStreamService",
-        "name": "tinygrad_llama3",
+        "name": "tinygrad-llama3",
         "aliases": null,
         "type": "ChatChoicesStreamService"
       }
@@ -500,7 +546,7 @@
           "ChatChoicesService",
           "ChatChoicesStreamService"
         ],
-        "backend_name": "tinygrad_llama3",
+        "backend_name": "tinygrad-llama3",
         "model_names": null
       }
     }
@@ -714,7 +760,7 @@
     "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.dev475}/ommlds/__about__.py

@@ -38,7 +38,11 @@ class Project(ProjectBase):
 
         'huggingface': [
             'huggingface-hub ~= 0.36',
-            'datasets ~= 4.3',
+            'datasets ~= 4.4',
+        ],
+
+        'nanochat': [
+            'regex >= 2025.0',
         ],
 
         'numpy': [
@@ -56,7 +60,7 @@ class Project(ProjectBase):
         ],
 
         'search': [
-            'ddgs ~= 9.6',
+            'ddgs ~= 9.8',
         ],
 
         'wiki': [
@@ -76,6 +80,8 @@ class Project(ProjectBase):
 
 
 class Setuptools(SetuptoolsBase):
+    rs = True
+
     find_packages = {
         'include': [Project.name, f'{Project.name}.*'],
         'exclude': [*SetuptoolsBase.find_packages['exclude']],

ommlds-0.0.0.dev475/ommlds/backends/tinygrad/models/llama3/__init__.py

@@ -0,0 +1,28 @@
+# ruff: noqa: I001
+
+# https://github.com/tinygrad/tinygrad/blob/2d6d8b735506464367b0315f9a2f424e0d19f66e/examples/llama3.py
+# https://github.com/tinygrad/tinygrad/blob/ef17af85c6d3f84d1e1cc084d6dee8ced3d1a33e/extra/models/llama.py
+
+
+from omlish import lang as _lang
+
+
+with _lang.auto_proxy_init(globals()):
+    from .fetch import (  # noqa
+        fetch_model,
+    )
+
+    from .llm import (  # noqa
+        Llama3Llm,
+        run_llm,
+        run_llm_to_stop,
+        RunLlmToStopResult,
+    )
+
+    from .tokenization import (  # noqa
+        Tokenizer,
+    )
+
+    from .transformer import (  # noqa
+        Transformer,
+    )

ommlds-0.0.0.dev475/ommlds/cli/asyncs.py

@@ -0,0 +1,30 @@
+import abc
+import functools
+import typing as ta
+
+from omlish import lang
+
+
+with lang.auto_proxy_import(globals()):
+    import anyio
+
+
+T = ta.TypeVar('T')
+P = ta.ParamSpec('P')
+
+
+##
+
+
+class AsyncThreadRunner(lang.Abstract):
+    @abc.abstractmethod
+    def run_in_thread(self, fn: ta.Callable[P, T], *args: P.args, **kwargs: P.kwargs) -> ta.Awaitable[T]:
+        raise NotImplementedError
+
+
+##
+
+
+class AnyioAsyncThreadRunner(AsyncThreadRunner):
+    def run_in_thread(self, fn: ta.Callable[P, T], *args: P.args, **kwargs: P.kwargs) -> ta.Awaitable[T]:
+        return anyio.to_thread.run_sync(functools.partial(fn, *args, **kwargs))

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev475}/ommlds/cli/inject.py

@@ -5,6 +5,7 @@ from omlish import lang
 
 
 with lang.auto_proxy_import(globals()):
+    from . import asyncs
     from .backends import inject as _backends
     from .sessions import inject as _sessions
     from .state import inject as _state
@@ -34,4 +35,10 @@ def bind_main(
 
     #
 
+    els.extend([
+        inj.bind(asyncs.AsyncThreadRunner, to_ctor=asyncs.AnyioAsyncThreadRunner),
+    ])
+
+    #
+
     return inj.as_elements(*els)

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev475}/ommlds/cli/main.py

@@ -4,10 +4,8 @@ See:
  - https://github.com/TheR1D/shell_gpt
  - https://github.com/paul-gauthier/aider
 """
-import argparse
 import functools
 import os.path
-import sys
 import typing as ta
 
 import anyio
@@ -16,7 +14,7 @@ from omdev.home.secrets import load_secrets
 from omlish import check
 from omlish import inject as inj
 from omlish import lang
-from omlish.diag import pycharm
+from omlish.argparse import all as ap
 from omlish.logs import all as logs
 from omlish.subprocesses.editor import edit_text_with_user_editor
 from omlish.subprocesses.sync import subprocesses
@@ -39,7 +37,7 @@ else:
 
 
 async def _a_main(args: ta.Any = None) -> None:
-    parser = argparse.ArgumentParser()
+    parser = ap.ArgumentParser()
     parser.add_argument('prompt', nargs='*')
 
     parser.add_argument('-b', '--backend', default='openai')
@@ -105,16 +103,6 @@ async def _a_main(args: ta.Any = None) -> None:
     elif not args.prompt:
         raise ValueError('Must provide prompt')
 
-    else:
-        prompt = ' '.join(args.prompt)
-
-        # FIXME: ptk / maysync
-        if not sys.stdin.isatty() and not pycharm.is_pycharm_hosted():
-            stdin_data = sys.stdin.read()
-            prompt = '\n'.join([prompt, stdin_data])
-
-        content = prompt
-
     #
 
     # FIXME: lol garbage
@@ -191,7 +179,7 @@ def _main(args: ta.Any = None) -> None:
             _a_main,
             args,
         ),
-        backend='asyncio',
+        # backend='trio',
     )  # noqa
 
 

ommlds-0.0.0.dev475/ommlds/cli/main2.py

@@ -0,0 +1,220 @@
+"""
+TODO:
+ - bootstrap lol
+"""
+import abc
+import functools
+import typing as ta
+
+import anyio
+
+from omlish import check
+from omlish import dataclasses as dc
+from omlish import inject as inj
+from omlish import lang
+from omlish.argparse import all as ap
+from omlish.logs import all as logs
+
+from .inject import bind_main
+from .sessions.base import Session
+from .sessions.chat.configs import ChatConfig
+
+
+##
+
+
+MAIN_EXTRA_ARGS: ta.Sequence[ap.Arg] = [
+    ap.arg('-v', '--verbose', action='store_true'),
+]
+
+
+def _process_main_extra_args(args: ap.Namespace) -> None:
+    if args.verbose:
+        logs.configure_standard_logging('DEBUG')
+    else:
+        logs.configure_standard_logging('INFO')
+        logs.silence_noisy_loggers()
+
+
+##
+
+
+class Profile(lang.Abstract):
+    @abc.abstractmethod
+    def run(self, argv: ta.Sequence[str]) -> ta.Awaitable[None]:
+        raise NotImplementedError
+
+
+##
+
+
+# class ChatAspect(lang.Abstract):
+#     def get_parser_args(self) -> ta.Sequence[ap.Arg]: ...
+#     def set_args(self, args: ap.Namespace) -> None: ...
+#     def configure(self, cfg: ChatConfig) -> ChatConfig: ...
+
+
+class ChatProfile(Profile):
+    _args: ap.Namespace
+
+    #
+
+    BACKEND_ARGS: ta.ClassVar[ta.Sequence[ap.Arg]] = [
+        ap.arg('-b', '--backend', group='backend'),
+    ]
+
+    def configure_backend(self, cfg: ChatConfig) -> ChatConfig:
+        return dc.replace(
+            cfg,
+            backend=self._args.backend,
+        )
+
+    #
+
+    INPUT_ARGS: ta.ClassVar[ta.Sequence[ap.Arg]] = [
+        ap.arg('message', nargs='*', group='input'),
+        ap.arg('-i', '--interactive', action='store_true', group='input'),
+        ap.arg('-e', '--editor', action='store_true', group='input'),
+    ]
+
+    def configure_input(self, cfg: ChatConfig) -> ChatConfig:
+        if self._args.editor:
+            check.arg(not self._args.interactive)
+            check.arg(not self._args.message)
+            raise NotImplementedError
+
+        elif self._args.interactive:
+            check.arg(not self._args.message)
+            return dc.replace(
+                cfg,
+                interactive=True,
+            )
+
+        elif self._args.message:
+            # TODO: '-' -> stdin
+            return dc.replace(
+                cfg,
+                initial_user_content=' '.join(self._args.message),
+            )
+
+        else:
+            raise ValueError('Must specify input')
+
+    #
+
+    STATE_ARGS: ta.ClassVar[ta.Sequence[ap.Arg]] = [
+        ap.arg('-n', '--new', action='store_true', group='state'),
+        ap.arg('--ephemeral', action='store_true', group='state'),
+    ]
+
+    def configure_state(self, cfg: ChatConfig) -> ChatConfig:
+        return dc.replace(
+            cfg,
+            state='ephemeral' if self._args.ephemeral else 'new' if self._args.new else 'continue',
+        )
+
+    #
+
+    OUTPUT_ARGS: ta.ClassVar[ta.Sequence[ap.Arg]] = [
+        ap.arg('-s', '--stream', action='store_true', group='output'),
+        ap.arg('-M', '--markdown', action='store_true', group='output'),
+    ]
+
+    def configure_output(self, cfg: ChatConfig) -> ChatConfig:
+        return dc.replace(
+            cfg,
+            stream=bool(self._args.stream),
+            markdown=bool(self._args.markdown),
+        )
+
+    #
+
+    async def run(self, argv: ta.Sequence[str]) -> None:
+        parser = ap.ArgumentParser()
+
+        for grp_name, grp_args in [
+            ('backend', self.BACKEND_ARGS),
+            ('input', self.INPUT_ARGS),
+            ('state', self.STATE_ARGS),
+            ('output', self.OUTPUT_ARGS),
+        ]:
+            grp = parser.add_argument_group(grp_name)
+            for a in grp_args:
+                grp.add_argument(*a.args, **a.kwargs)
+
+        self._args = parser.parse_args(argv)
+
+        cfg = ChatConfig()
+        cfg = self.configure_backend(cfg)
+        cfg = self.configure_input(cfg)
+        cfg = self.configure_state(cfg)
+        cfg = self.configure_output(cfg)
+
+        # session_cfg = ChatConfig(
+        #     initial_system_content=system_content,
+        #     enable_tools=(
+        #             args.enable_fs_tools or
+        #             args.enable_todo_tools or
+        #             args.enable_unsafe_tools_do_not_use_lol or
+        #             args.enable_test_weather_tool or
+        #             args.code
+        #     ),
+        #     enabled_tools={  # noqa
+        #         *(['fs'] if args.enable_fs_tools else []),
+        #         *(['todo'] if args.enable_todo_tools else []),
+        #         *(['weather'] if args.enable_test_weather_tool else []),
+        #         # FIXME: enable_unsafe_tools_do_not_use_lol
+        #     },
+        #     dangerous_no_tool_confirmation=bool(args.dangerous_no_tool_confirmation),
+        # )
+
+        with inj.create_managed_injector(bind_main(
+                session_cfg=cfg,
+                enable_backend_strings=True,
+        )) as injector:
+            await injector[Session].run()
+
+
+##
+
+
+PROFILE_TYPES: ta.Mapping[str, type[Profile]] = {
+    'chat': ChatProfile,
+}
+
+
+##
+
+
+MAIN_PROFILE_ARGS: ta.Sequence[ap.Arg] = [
+    ap.arg('profile', nargs='?', default='chat'),
+    ap.arg('args', nargs=ap.REMAINDER),
+]
+
+
+async def _a_main(argv: ta.Any = None) -> None:
+    parser = ap.ArgumentParser()
+
+    for a in [*MAIN_EXTRA_ARGS, *MAIN_PROFILE_ARGS]:
+        parser.add_argument(*a.args, **a.kwargs)
+
+    args, unk_args = parser.parse_known_args(argv)
+
+    _process_main_extra_args(args)
+
+    profile_cls = PROFILE_TYPES[args.profile]
+    profile = profile_cls()
+    await profile.run([*unk_args, *args.args])
+
+
+def _main(args: ta.Any = None) -> None:
+    anyio.run(
+        functools.partial(
+            _a_main,
+            args,
+        ),
+    )  # noqa
+
+
+if __name__ == '__main__':
+    _main()

{ommlds-0.0.0.dev469 → ommlds-0.0.0.dev475}/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.dev475}/ommlds/cli/sessions/chat/backends/inject.py

@@ -1,6 +1,10 @@
+import typing as ta
+
 from omlish import inject as inj
 from omlish import lang
+from omlish import typedvalues as tv
 
+from ..... import minichain as mc
 from .injection import backend_configs
 
 
@@ -34,4 +38,16 @@ 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(*tv.collect(*(be.configs or []), *(cfgs or []), override=True), **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.dev475}/ommlds/cli/sessions/chat/chat/user/inject.py

@@ -11,6 +11,7 @@ from ...phases.types import ChatPhaseCallback
 
 with lang.auto_proxy_import(globals()):
     from ..state import types as _state
+    from . import inputs as _inputs
     from . import interactive as _interactive
     from . import oneshot as _oneshot
     from . import types as _types
@@ -24,6 +25,7 @@ def bind_user(
         initial_system_content: ta.Optional['mc.Content'] = None,
         initial_user_content: ta.Optional['mc.Content'] = None,
         interactive: bool = False,
+        use_readline: bool | ta.Literal['auto'] = False,
 ) -> inj.Elements:
     els: list[inj.Elemental] = []
 
@@ -49,6 +51,11 @@ def bind_user(
 
         els.append(inj.bind(_types.UserChatInput, to_ctor=_interactive.InteractiveUserChatInput, singleton=True))
 
+        els.extend([
+            inj.bind(_inputs.SyncStringInput, to_const=_inputs.InputSyncStringInput(use_readline=use_readline)),  # noqa
+            inj.bind(_inputs.AsyncStringInput, to_ctor=_inputs.ThreadAsyncStringInput, singleton=True),
+        ])
+
     else:
         if initial_user_content is None:
             raise ValueError('Initial user content is required for non-interactive chat')