ommlds 0.0.0.dev466__py3-none-any.whl → 0.0.0.dev512__py3-none-any.whl

ommlds/backends/ollama/cli.py

@@ -0,0 +1,36 @@
+from omdev.cli import CliModule
+from omlish import check
+from omlish.argparse import all as ap
+from omlish.formats import json
+from omlish.http import all as http
+from omlish.logs import all as logs
+
+
+log = logs.get_module_logger(globals())
+
+
+##
+
+
+class Cli(ap.Cli):
+    @ap.cmd()
+    def list(self) -> None:
+        resp = http.request('http://localhost:11434/api/tags', check=True)
+        data = json.loads(check.not_none(resp.data).decode('utf-8'))
+        print(json.dumps_pretty(data))
+
+
+##
+
+
+def _main() -> None:
+    logs.configure_standard_logging('INFO')
+    Cli()()
+
+
+# @omlish-manifest
+_CLI_MODULE = CliModule('ollama', __name__)
+
+
+if __name__ == '__main__':
+    _main()

ommlds/backends/ollama/protocol.py

@@ -0,0 +1,201 @@
+"""
+https://docs.ollama.com/api
+"""
+import datetime
+import typing as ta
+
+from omlish import dataclasses as dc
+from omlish import lang
+
+
+##
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class Options:
+    # loading
+    numa: bool | None = None
+    num_ctx: int | None = None
+    num_batch: int | None = None
+    num_gpu: int | None = None
+    main_gpu: int | None = None
+    low_vram: bool | None = None
+    f16_kv: bool | None = None
+    logits_all: bool | None = None
+    vocab_only: bool | None = None
+    use_mmap: bool | None = None
+    use_mlock: bool | None = None
+    embedding_only: bool | None = None
+    num_thread: int | None = None
+
+    # querying
+    num_keep: int | None = None
+    seed: int | None = None
+    num_predict: int | None = None
+    top_k: int | None = None
+    top_p: float | None = None
+    tfs_z: float | None = None
+    typical_p: float | None = None
+    repeat_last_n: int | None = None
+    temperature: float | None = None
+    repeat_penalty: float | None = None
+    presence_penalty: float | None = None
+    frequency_penalty: float | None = None
+    mirostat: int | None = None
+    mirostat_tau: float | None = None
+    mirostat_eta: float | None = None
+    penalize_newline: bool | None = None
+    stop: ta.Sequence[str] | None = None
+
+
+##
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+class BaseRequest(lang.Abstract):
+    model: str
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+class BaseStreamableRequest(BaseRequest, lang.Abstract):
+    stream: bool | None = None
+
+
+##
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+class BaseGenerateRequest(BaseStreamableRequest, lang.Abstract):
+    options: Options | None = None
+    format: ta.Literal['', 'json'] | None = None  # TODO: jsonschema
+    keep_alive: float | str | None = None
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class GenerateRequest(BaseGenerateRequest):
+    prompt: str | None = None
+    suffix: str | None = None
+    system: str | None = None
+    template: str | None = None
+    context: ta.Sequence[int] | None = None
+    raw: bool | None = None
+    images: ta.Sequence[bytes] | None = None
+    think: bool | ta.Literal['low', 'medium', 'high'] | None = None
+
+
+#
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+class BaseGenerateResponse(lang.Abstract):
+    model: str | None = None
+    created_at: str | None = None
+    done: bool | None = None
+    done_reason: str | None = None
+    total_duration: int | None = None
+    load_duration: int | None = None
+    prompt_eval_count: int | None = None
+    prompt_eval_duration: int | None = None
+    eval_count: int | None = None
+    eval_duration: int | None = None
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class GenerateResponse(BaseGenerateResponse):
+    response: str
+    thinking: str | None = None
+    context: ta.Sequence[int] | None = None
+
+
+##
+
+
+Role: ta.TypeAlias = ta.Literal[
+    'system',
+    'user',
+    'assistant',
+    'tool',
+]
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class Message:
+    role: Role
+    content: str | None = None
+    thinking: str | None = None
+    images: ta.Sequence[bytes] | None = None
+    tool_name: str | None = None
+
+    @dc.dataclass(frozen=True, kw_only=True)
+    class ToolCall:
+        id: str | None = None
+
+        @dc.dataclass(frozen=True, kw_only=True)
+        class Function:
+            name: str
+            arguments: ta.Mapping[str, ta.Any]
+            index: int | None = None
+
+        function: Function
+
+    tool_calls: ta.Sequence[ToolCall] | None = None
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class Tool:
+    type: str | None = 'function'
+
+    @dc.dataclass(frozen=True, kw_only=True)
+    @dc.extra_class_params(default_repr_fn=lang.opt_repr)
+    class Function:
+        name: str | None = None
+        description: str | None = None
+        parameters: ta.Any | None = None
+
+    function: Function | None = None
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class ChatRequest(BaseGenerateRequest):
+    messages: ta.Sequence[Message] | None = None
+    tools: ta.Sequence[Tool] | None = None
+    think: bool | ta.Literal['low', 'medium', 'high'] | None = None
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class ChatResponse(BaseGenerateResponse):
+    message: Message
+
+
+##
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class ModelDetails:
+    parent_model: str
+    format: str
+    family: str
+    families: ta.Sequence[str]
+    parameter_size: str
+    quantization_level: str
+
+
+@dc.dataclass(frozen=True, kw_only=True)
+@dc.extra_class_params(default_repr_fn=lang.opt_repr)
+class ListModelResponse:
+    name: str
+    model: str
+    remote_model: str | None = None
+    remote_host: str | None = None
+    modified_at: datetime.datetime
+    size: int
+    digest: str
+    details: ModelDetails | None = None

ommlds/backends/openai/protocol/__init__.py

@@ -1,8 +1,22 @@
+# ruff: noqa: I001
 """
 https://platform.openai.com/docs/api-reference/introduction
 https://github.com/openai/openai-openapi/blob/master/openapi.yaml
 """
-# ruff: noqa: I001
+##
+
+
+from omlish import dataclasses as _dc  # noqa
+
+
+_dc.init_package(
+    globals(),
+    codegen=True,
+)
+
+
+##
+
 
 from .chatcompletion.chunk import (  # noqa
     ChatCompletionChunkChoiceDelta,

ommlds/backends/openai/protocol/_common.py

@@ -6,12 +6,10 @@ from omlish import marshal as msh
 
 
 def _set_class_marshal_options(cls):
-    msh.update_object_metadata(
+    msh.update_object_options(
         cls,
-        field_defaults=msh.FieldMetadata(
-            options=msh.FieldOptions(
-                omit_if=lang.is_none,
-            ),
+        field_defaults=msh.FieldOptions(
+            omit_if=lang.is_none,
         ),
     )