kainguru-sdk 0.2.11.dev0__tar.gz → 0.2.13.dev0__tar.gz

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: kainguru-sdk
-Version: 0.2.11.dev0
+Version: 0.2.13.dev0
 Summary: Python SDK for the Kainguru ML platform — run models and fine-tune them.
 Project-URL: Homepage, https://kainguru.com
 Project-URL: Repository, https://github.com/kainguru/kainguru

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "kainguru-sdk"
-version = "0.2.11.dev0"
+version = "0.2.13.dev0"
 description = "Python SDK for the Kainguru ML platform — run models and fine-tune them."
 readme = "README.md"
 requires-python = ">=3.10"

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/__init__.py

@@ -22,6 +22,7 @@ from .models import (
     ModelFineTunedDetailDto,
     ModelStatus,
 )
+from .resources import AsyncRagResource, RagResource
 
 __version__ = "0.1.0"
 
@@ -39,6 +40,9 @@ __all__ = [
     "KainguruRateLimitError",
     "KainguruTimeoutError",
     "KainguruConnectionError",
+    # Resources
+    "RagResource",
+    "AsyncRagResource",
     # Models
     "ModelStatus",
     "ModelExecResponseDto",

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/async_client.py

@@ -13,7 +13,7 @@ from ._config import (
     KainguruConfig,
 )
 from ._transport import AsyncTransport, enable_debug_logging
-from .resources import AsyncExecutionsResource, AsyncFineTuningResource
+from .resources import AsyncExecutionsResource, AsyncFineTuningResource, AsyncRagResource
 
 
 class AsyncKainguruClient:
@@ -58,6 +58,7 @@ class AsyncKainguruClient:
         self._transport = AsyncTransport(client, self._config)
         self.executions = AsyncExecutionsResource(self._transport)
         self.fine_tuning = AsyncFineTuningResource(self._transport)
+        self.rag = AsyncRagResource(self._transport)
 
     @property
     def config(self) -> KainguruConfig:

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/client.py

@@ -13,7 +13,7 @@ from ._config import (
     KainguruConfig,
 )
 from ._transport import Transport, enable_debug_logging
-from .resources import ExecutionsResource, FineTuningResource
+from .resources import ExecutionsResource, FineTuningResource, RagResource
 
 
 class KainguruClient:
@@ -58,6 +58,7 @@ class KainguruClient:
         self._transport = Transport(client, self._config)
         self.executions = ExecutionsResource(self._transport)
         self.fine_tuning = FineTuningResource(self._transport)
+        self.rag = RagResource(self._transport)
 
     @property
     def config(self) -> KainguruConfig:

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/models.py

@@ -7,12 +7,13 @@ from enum import Enum
 from typing import Any
 from uuid import UUID
 
-from pydantic import AwareDatetime, BaseModel, constr
+from pydantic import AwareDatetime, BaseModel, Field, constr
 
 
-class ModelFineTuneRequestDto(BaseModel):
+class RagIngestRequestDto(BaseModel):
     name: constr(min_length=1)
-    input: dict[str, Any]
+    fileIds: list[UUID] = Field(..., min_length=1)
+    params: dict[str, Any] | None = None
 
 
 class CategoryShortDto(BaseModel):
@@ -24,39 +25,54 @@ class CategoryShortDto(BaseModel):
     modifiedDate: AwareDatetime | None = None
 
 
-class InputType(Enum):
-    STRING = 'STRING'
-    INTEGER = 'INTEGER'
-    FLOAT = 'FLOAT'
-    BOOLEAN = 'BOOLEAN'
-    DICT = 'DICT'
-    PATH = 'PATH'
-    PATH_ARRAY = 'PATH_ARRAY'
-    LABELED_PATH_ARRAY = 'LABELED_PATH_ARRAY'
-
-
-class Scope(Enum):
-    BASE_ONLY = 'BASE_ONLY'
-    FINE_TUNED_ONLY = 'FINE_TUNED_ONLY'
-    BOTH = 'BOTH'
+class FileType(Enum):
+    JPEG = 'JPEG'
+    PNG = 'PNG'
+    WEBP = 'WEBP'
+    GIF = 'GIF'
+    BMP = 'BMP'
+    TIFF = 'TIFF'
+    SVG = 'SVG'
+    MP3 = 'MP3'
+    WAV = 'WAV'
+    FLAC = 'FLAC'
+    OGG = 'OGG'
+    M4A = 'M4A'
+    AAC = 'AAC'
+    WEBM_AUDIO = 'WEBM_AUDIO'
+    MP4 = 'MP4'
+    WEBM_VIDEO = 'WEBM_VIDEO'
+    MOV = 'MOV'
+    AVI = 'AVI'
+    MKV = 'MKV'
+    JSON = 'JSON'
+    JSONL = 'JSONL'
+    TXT = 'TXT'
+    CSV = 'CSV'
+    PDF = 'PDF'
+    DOCX = 'DOCX'
+    ZIP = 'ZIP'
+    PY = 'PY'
+    YAML = 'YAML'
+    MD = 'MD'
+    OBJ = 'OBJ'
+    GLB = 'GLB'
+    GLTF = 'GLTF'
+    PLY = 'PLY'
+
+
+class AccessLevel(Enum):
+    PUBLIC = 'PUBLIC'
+    PRIVATE = 'PRIVATE'
 
 
-class ModelExecInputDto(BaseModel):
+class FileDto(BaseModel):
     id: UUID | None = None
-    createDate: AwareDatetime | None = None
-    modifiedDate: AwareDatetime | None = None
     name: str | None = None
-    description: str | None = None
-    inputType: InputType | None = None
-    validations: dict[str, Any] | None = None
-    required: bool | None = None
-    mediaTypes: list[str] | None = None
-    scope: Scope | None = None
-    label: str | None = None
-    defaultValue: Any | None = None
-    allowedValues: list[Any] | None = None
-    requirementGroup: list[str] | None = None
-    requiredWhen: dict[str, Any] | None = None
+    url: str | None = None
+    s3Key: str | None = None
+    fileType: FileType | None = None
+    accessLevel: AccessLevel | None = None
 
 
 class AccessType(Enum):
@@ -90,14 +106,28 @@ class ModelStatus(Enum):
     DELETED = 'DELETED'
 
 
-class UserDto(BaseModel):
+class RagOperation(Enum):
+    RAG_INGEST = 'RAG_INGEST'
+    RAG_APPEND = 'RAG_APPEND'
+    RAG_DOC_DELETE = 'RAG_DOC_DELETE'
+    RAG_DELETE = 'RAG_DELETE'
+
+
+class RagOperationResultDto(BaseModel):
     id: UUID | None = None
-    username: str | None = None
-    fullname: str | None = None
-    email: str | None = None
-    emailVerified: bool | None = None
-    createDate: AwareDatetime | None = None
-    modifiedDate: AwareDatetime | None = None
+    operation: RagOperation | None = None
+    status: ModelStatus | None = None
+    added: list[str] | None = None
+    replaced: list[str] | None = None
+    skipped: list[str] | None = None
+    deleted: list[dict[str, Any]] | None = None
+    notFound: list[str] | None = None
+    sourceFiles: list[str] | None = None
+    numChunks: int | None = None
+    verified: bool | None = None
+    startDate: AwareDatetime | None = None
+    endDate: AwareDatetime | None = None
+    duration: float | None = None
 
 
 class CommonApiResponse(BaseModel):
@@ -111,9 +141,80 @@ class FieldError(BaseModel):
     message: str | None = None
 
 
+class RagFileSelectionRequestDto(BaseModel):
+    fileIds: list[UUID] = Field(..., min_length=1)
+
+
+class ModelFineTuneRequestDto(BaseModel):
+    name: constr(min_length=1)
+    input: dict[str, Any]
+
+
+class FineTuneResultDto(BaseModel):
+    id: UUID | None = None
+    status: ModelStatus | None = None
+    input: dict[str, Any] | None = None
+    output: Any | None = None
+    step: str | None = None
+    percent: float | None = None
+    stepsHistory: list[dict[str, Any]] | None = None
+    startDate: AwareDatetime | None = None
+    endDate: AwareDatetime | None = None
+    duration: float | None = None
+
+
+class InputType(Enum):
+    STRING = 'STRING'
+    INTEGER = 'INTEGER'
+    FLOAT = 'FLOAT'
+    BOOLEAN = 'BOOLEAN'
+    DICT = 'DICT'
+    PATH = 'PATH'
+    PATH_ARRAY = 'PATH_ARRAY'
+    LABELED_PATH_ARRAY = 'LABELED_PATH_ARRAY'
+
+
+class Scope(Enum):
+    BASE_ONLY = 'BASE_ONLY'
+    FINE_TUNED_ONLY = 'FINE_TUNED_ONLY'
+    BOTH = 'BOTH'
+
+
+class ModelExecInputDto(BaseModel):
+    id: UUID | None = None
+    createDate: AwareDatetime | None = None
+    modifiedDate: AwareDatetime | None = None
+    name: str | None = None
+    description: str | None = None
+    inputType: InputType | None = None
+    validations: dict[str, Any] | None = None
+    required: bool | None = None
+    mediaTypes: list[str] | None = None
+    scope: Scope | None = None
+    label: str | None = None
+    defaultValue: Any | None = None
+    allowedValues: list[Any] | None = None
+    requirementGroup: list[str] | None = None
+    requiredWhen: dict[str, Any] | None = None
+
+
+class UserDto(BaseModel):
+    id: UUID | None = None
+    username: str | None = None
+    fullname: str | None = None
+    email: str | None = None
+    emailVerified: bool | None = None
+    createDate: AwareDatetime | None = None
+    modifiedDate: AwareDatetime | None = None
+
+
 class ModelExecRequestDto(BaseModel):
     input: dict[str, Any]
     outputFormat: str | None = None
+    ragHandle: str | None = None
+    ragTopK: int | None = None
+    ragMinScore: float | None = None
+    ragPromptMode: str | None = None
 
 
 class Status(Enum):
@@ -131,20 +232,6 @@ class ApplicationDto(BaseModel):
     archived: bool | None = None
 
 
-class ExecutionResultDto(BaseModel):
-    id: UUID | None = None
-    input: dict[str, Any] | None = None
-    output: Any | None = None
-    inputFormats: dict[str, str] | None = None
-    outputFormat: str | None = None
-    step: str | None = None
-    percent: float | None = None
-    stepsHistory: list[dict[str, Any]] | None = None
-    startDate: AwareDatetime | None = None
-    endDate: AwareDatetime | None = None
-    duration: float | None = None
-
-
 class ModelExecFineTunedRefDto(BaseModel):
     id: UUID | None = None
     mlflowModelId: str | None = None
@@ -156,28 +243,32 @@ class ModelExecModelRefDto(BaseModel):
     version: str | None = None
 
 
-class ModelExecResponseDto(BaseModel):
+class RagRequestDto(BaseModel):
+    handle: str | None = None
+    topK: int | None = None
+    minScore: float | None = None
+    promptMode: str | None = None
+
+
+class SourceDto(BaseModel):
+    sourceFile: str | None = None
+    chunkIndex: int | None = None
+
+
+class RagKnowledgeBaseDto(BaseModel):
     id: UUID | None = None
+    name: str | None = None
+    ragHandle: str | None = None
     status: ModelStatus | None = None
     model: ModelExecModelRefDto | None = None
-    fineTunedModel: ModelExecFineTunedRefDto | None = None
-    application: ApplicationDto | None = None
-    execution: ExecutionResultDto | None = None
     createDate: AwareDatetime | None = None
     modifiedDate: AwareDatetime | None = None
 
 
-class FineTuneResultDto(BaseModel):
-    id: UUID | None = None
-    status: ModelStatus | None = None
-    input: dict[str, Any] | None = None
-    output: Any | None = None
-    step: str | None = None
-    percent: float | None = None
-    stepsHistory: list[dict[str, Any]] | None = None
-    startDate: AwareDatetime | None = None
-    endDate: AwareDatetime | None = None
-    duration: float | None = None
+class RagKnowledgeBaseListApiResponse(BaseModel):
+    success: bool | None = None
+    code: int | None = None
+    data: list[RagKnowledgeBaseDto] | None = None
 
 
 class ModelListDto(BaseModel):
@@ -199,6 +290,18 @@ class ModelListDto(BaseModel):
     banner: str | None = None
 
 
+class RagKnowledgeBaseDetailDto(BaseModel):
+    id: UUID | None = None
+    name: str | None = None
+    ragHandle: str | None = None
+    status: ModelStatus | None = None
+    model: ModelListDto | None = None
+    files: list[FileDto] | None = None
+    latestOperation: RagOperationResultDto | None = None
+    createDate: AwareDatetime | None = None
+    modifiedDate: AwareDatetime | None = None
+
+
 class ErrorDto(BaseModel):
     success: bool | None = None
     code: int | None = None
@@ -209,12 +312,6 @@ class ErrorDto(BaseModel):
     errors: list[FieldError] | None = None
 
 
-class ModelExecApiResponse(BaseModel):
-    success: bool | None = None
-    code: int | None = None
-    data: ModelExecResponseDto | None = None
-
-
 class ModelFineTunedDetailDto(BaseModel):
     id: UUID | None = None
     name: str | None = None
@@ -230,7 +327,49 @@ class ModelFineTunedDetailDto(BaseModel):
     modifiedDate: AwareDatetime | None = None
 
 
+class ExecutionResultDto(BaseModel):
+    id: UUID | None = None
+    input: dict[str, Any] | None = None
+    output: Any | None = None
+    inputFormats: dict[str, str] | None = None
+    outputFormat: str | None = None
+    step: str | None = None
+    percent: float | None = None
+    stepsHistory: list[dict[str, Any]] | None = None
+    startDate: AwareDatetime | None = None
+    endDate: AwareDatetime | None = None
+    duration: float | None = None
+    mode: str | None = None
+    sources: list[SourceDto] | None = None
+    loraRunId: str | None = None
+    ragHandle: str | None = None
+    ragRequest: RagRequestDto | None = None
+
+
+class ModelExecResponseDto(BaseModel):
+    id: UUID | None = None
+    status: ModelStatus | None = None
+    model: ModelExecModelRefDto | None = None
+    fineTunedModel: ModelExecFineTunedRefDto | None = None
+    application: ApplicationDto | None = None
+    execution: ExecutionResultDto | None = None
+    createDate: AwareDatetime | None = None
+    modifiedDate: AwareDatetime | None = None
+
+
+class RagKnowledgeBaseApiResponse(BaseModel):
+    success: bool | None = None
+    code: int | None = None
+    data: RagKnowledgeBaseDetailDto | None = None
+
+
 class ModelFineTuneApiResponse(BaseModel):
     success: bool | None = None
     code: int | None = None
     data: ModelFineTunedDetailDto | None = None
+
+
+class ModelExecApiResponse(BaseModel):
+    success: bool | None = None
+    code: int | None = None
+    data: ModelExecResponseDto | None = None

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/resources/__init__.py

@@ -1,9 +1,12 @@
 from .executions import AsyncExecutionsResource, ExecutionsResource
 from .finetuning import AsyncFineTuningResource, FineTuningResource
+from .rag import AsyncRagResource, RagResource
 
 __all__ = [
     "ExecutionsResource",
     "AsyncExecutionsResource",
     "FineTuningResource",
     "AsyncFineTuningResource",
+    "RagResource",
+    "AsyncRagResource",
 ]

kainguru_sdk-0.2.13.dev0/src/kainguru/resources/_common.py

@@ -0,0 +1,88 @@
+"""Helpers shared across sync and async resources."""
+
+from __future__ import annotations
+
+from typing import Any
+
+#: A RAG knowledge base's *latest operation* is terminal at these statuses. Unlike
+#: executions/fine-tuning, RAG also treats CANCELLED/DELETED as terminal since
+#: `cancel`/`delete` are themselves operations callers may want to await.
+_RAG_TERMINAL_STATUSES = ("COMPLETED", "FAILED", "CANCELLED", "DELETED")
+
+
+def is_terminal(dto: object) -> bool:
+    """COMPLETED / FAILED are terminal; REGISTERED is not (matches the Java SDK)."""
+    status = getattr(dto, "status", None)
+    value = getattr(status, "value", status)
+    return value in ("COMPLETED", "FAILED")
+
+
+def is_rag_terminal(dto: object) -> bool:
+    """Terminal check for RAG knowledge base polling.
+
+    Reads ``dto.latestOperation.status`` rather than ``dto.status``: the knowledge
+    base's own ``status`` stays ``COMPLETED`` while an append or document-delete
+    operation is in flight, so polling it directly would return prematurely (or
+    never reflect append/delete completion at all). Terminal values are
+    ``COMPLETED``, ``FAILED``, ``CANCELLED``, ``DELETED``.
+    """
+    latest_operation = getattr(dto, "latestOperation", None)
+    if latest_operation is None:
+        return False
+    status = getattr(latest_operation, "status", None)
+    value = getattr(status, "value", status)
+    return value in _RAG_TERMINAL_STATUSES
+
+
+def execute_args(
+    ml_flow_id: str,
+    input: dict[str, Any],
+    output_format: str | None,
+    exec_id: str | None,
+    rag_handle: str | None = None,
+    rag_top_k: int | None = None,
+    rag_min_score: float | None = None,
+    rag_prompt_mode: str | None = None,
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    params: dict[str, Any] = {"mlFlowId": ml_flow_id}
+    if exec_id:
+        params["execId"] = exec_id
+    body: dict[str, Any] = {"input": input or {}, "outputFormat": output_format}
+    if rag_handle is not None:
+        body["ragHandle"] = rag_handle
+    if rag_top_k is not None:
+        body["ragTopK"] = rag_top_k
+    if rag_min_score is not None:
+        body["ragMinScore"] = rag_min_score
+    if rag_prompt_mode is not None:
+        body["ragPromptMode"] = rag_prompt_mode
+    return params, body
+
+
+def fine_tune_args(
+    model_id: str,
+    name: str,
+    input: dict[str, Any],
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    params: dict[str, Any] = {"modelId": model_id}
+    body: dict[str, Any] = {"name": name, "input": input or {}}
+    return params, body
+
+
+def rag_ingest_args(
+    model_id: str,
+    name: str,
+    file_ids: list[str],
+    params: dict[str, Any] | None,
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    """Build the ``(query, body)`` pair for ``POST /v1/rag``."""
+    query: dict[str, Any] = {"modelId": model_id}
+    body: dict[str, Any] = {"name": name, "fileIds": file_ids}
+    if params is not None:
+        body["params"] = params
+    return query, body
+
+
+def rag_file_selection_args(file_ids: list[str]) -> dict[str, Any]:
+    """Build the request body shared by ``append`` and ``delete_documents``."""
+    return {"fileIds": file_ids}

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/resources/executions.py

@@ -36,12 +36,30 @@ class ExecutionsResource:
         output_format: str | None = None,
         exec_id: str | None = None,
         *,
+        rag_handle: str | None = None,
+        rag_top_k: int | None = None,
+        rag_min_score: float | None = None,
+        rag_prompt_mode: str | None = None,
         api_key: str | None = None,
     ) -> ModelExecResponseDto:
-        """Submit a model run. Returns immediately (typically ``PENDING``)."""
+        """Submit a model run. Returns immediately (typically ``PENDING``).
+
+        The ``rag_*`` kwargs enable retrieval-augmented generation against a
+        knowledge base previously created via ``client.rag.ingest`` (identified by
+        its ``ragHandle``).
+        """
         if not ml_flow_id:
             raise KainguruConfigError("ml_flow_id is required")
-        params, body = execute_args(ml_flow_id, input, output_format, exec_id)
+        params, body = execute_args(
+            ml_flow_id,
+            input,
+            output_format,
+            exec_id,
+            rag_handle=rag_handle,
+            rag_top_k=rag_top_k,
+            rag_min_score=rag_min_score,
+            rag_prompt_mode=rag_prompt_mode,
+        )
         response = self._t.request("POST", _PATH, params=params, json=body, api_key=api_key)
         return unwrap(response, ModelExecResponseDto)
 
@@ -52,6 +70,13 @@ class ExecutionsResource:
         response = self._t.request("GET", f"{_PATH}/{id}", api_key=api_key)
         return unwrap(response, ModelExecResponseDto)
 
+    def cancel(self, id: str, *, api_key: str | None = None) -> ModelExecResponseDto:
+        """Cancel an active execution."""
+        if not id:
+            raise KainguruConfigError("id is required")
+        response = self._t.request("POST", f"{_PATH}/{id}/cancel", api_key=api_key)
+        return unwrap(response, ModelExecResponseDto)
+
     def await_completion(
         self,
         id: str,
@@ -86,11 +111,24 @@ class AsyncExecutionsResource:
         output_format: str | None = None,
         exec_id: str | None = None,
         *,
+        rag_handle: str | None = None,
+        rag_top_k: int | None = None,
+        rag_min_score: float | None = None,
+        rag_prompt_mode: str | None = None,
         api_key: str | None = None,
     ) -> ModelExecResponseDto:
         if not ml_flow_id:
             raise KainguruConfigError("ml_flow_id is required")
-        params, body = execute_args(ml_flow_id, input, output_format, exec_id)
+        params, body = execute_args(
+            ml_flow_id,
+            input,
+            output_format,
+            exec_id,
+            rag_handle=rag_handle,
+            rag_top_k=rag_top_k,
+            rag_min_score=rag_min_score,
+            rag_prompt_mode=rag_prompt_mode,
+        )
         response = await self._t.request("POST", _PATH, params=params, json=body, api_key=api_key)
         return unwrap(response, ModelExecResponseDto)
 
@@ -100,6 +138,12 @@ class AsyncExecutionsResource:
         response = await self._t.request("GET", f"{_PATH}/{id}", api_key=api_key)
         return unwrap(response, ModelExecResponseDto)
 
+    async def cancel(self, id: str, *, api_key: str | None = None) -> ModelExecResponseDto:
+        if not id:
+            raise KainguruConfigError("id is required")
+        response = await self._t.request("POST", f"{_PATH}/{id}/cancel", api_key=api_key)
+        return unwrap(response, ModelExecResponseDto)
+
     async def await_completion(
         self,
         id: str,

{kainguru_sdk-0.2.11.dev0 → kainguru_sdk-0.2.13.dev0}/src/kainguru/resources/finetuning.py

@@ -51,6 +51,20 @@ class FineTuningResource:
         response = self._t.request("GET", f"{_PATH}/{id}", api_key=api_key)
         return unwrap(response, ModelFineTunedDetailDto)
 
+    def cancel(self, id: str, *, api_key: str | None = None) -> ModelFineTunedDetailDto:
+        """Cancel an active fine-tuning job."""
+        if not id:
+            raise KainguruConfigError("id is required")
+        response = self._t.request("POST", f"{_PATH}/{id}/cancel", api_key=api_key)
+        return unwrap(response, ModelFineTunedDetailDto)
+
+    def delete(self, id: str, *, api_key: str | None = None) -> ModelFineTunedDetailDto:
+        """Delete a fine-tuned model and its LoRA artifact."""
+        if not id:
+            raise KainguruConfigError("id is required")
+        response = self._t.request("DELETE", f"{_PATH}/{id}", api_key=api_key)
+        return unwrap(response, ModelFineTunedDetailDto)
+
     def await_completion(
         self,
         id: str,
@@ -98,6 +112,18 @@ class AsyncFineTuningResource:
         response = await self._t.request("GET", f"{_PATH}/{id}", api_key=api_key)
         return unwrap(response, ModelFineTunedDetailDto)
 
+    async def cancel(self, id: str, *, api_key: str | None = None) -> ModelFineTunedDetailDto:
+        if not id:
+            raise KainguruConfigError("id is required")
+        response = await self._t.request("POST", f"{_PATH}/{id}/cancel", api_key=api_key)
+        return unwrap(response, ModelFineTunedDetailDto)
+
+    async def delete(self, id: str, *, api_key: str | None = None) -> ModelFineTunedDetailDto:
+        if not id:
+            raise KainguruConfigError("id is required")
+        response = await self._t.request("DELETE", f"{_PATH}/{id}", api_key=api_key)
+        return unwrap(response, ModelFineTunedDetailDto)
+
     async def await_completion(
         self,
         id: str,