seekrai 0.3.2__tar.gz → 0.4.1__tar.gz

{seekrai-0.3.2 → seekrai-0.4.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: seekrai
-Version: 0.3.2
+Version: 0.4.1
 Summary: Python client for SeekrAI
 Home-page: https://gitlab.cb.ntent.com/ml/seekr-py
 License: Apache-2.0

{seekrai-0.3.2 → seekrai-0.4.1}/pyproject.toml

@@ -14,7 +14,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "seekrai"
-version = "0.3.2"
+version = "0.4.1"
 authors = [
     "SeekrFlow <support@seekr.com>"
 ]
@@ -73,6 +73,7 @@ tox = "^4.14.1"
 [tool.poetry.group.dev.dependencies]
 ipdb = "^0.13.13"
 pre-commit = "^3.7.1"
+jupyter = "^1.0.0"
 
 [tool.poetry.urls]
 "Homepage" = "https://www.seekr.com/"

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/abstract/api_requestor.py

@@ -524,6 +524,8 @@ class APIRequestor:
                 if rbody.strip().endswith("[DONE]"):
                     # TODO
                     rbody = rbody.replace("data: [DONE]", "")
+                if rbody.startswith("data: "):
+                    rbody = rbody[len("data: ") :]
                 data = json.loads(rbody)
         except (JSONDecodeError, UnicodeDecodeError) as e:
             raise error.APIError(

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/filemanager.py

@@ -91,10 +91,9 @@ def _prepare_output(
 
     content_type = str(headers.get("content-type"))
 
-    assert remote_name, (
-        "No model name found in fine_tune object. "
-        "Please specify an `output` file name."
-    )
+    assert (
+        remote_name
+    ), "No model name found in fine_tune object. Please specify an `output` file name."
 
     if step > 0:
         remote_name += f"-checkpoint-{step}"

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/resources/alignment.py

@@ -3,9 +3,12 @@ from typing import List
 from seekrai.abstract import api_requestor
 from seekrai.seekrflow_response import SeekrFlowResponse
 from seekrai.types import (
+    AlignmentEstimationRequest,
+    AlignmentEstimationResponse,
     AlignmentList,
     AlignmentRequest,
     AlignmentResponse,
+    AlignmentType,
     SeekrFlowClient,
     SeekrFlowRequest,
 )
@@ -19,14 +22,14 @@ class Alignment:
         self,
         instructions: str,
         files: List[str],
+        type: AlignmentType = AlignmentType.PRINCIPLE,
     ) -> AlignmentResponse:
         requestor = api_requestor.APIRequestor(
             client=self._client,
         )
 
         parameter_payload = AlignmentRequest(
-            instructions=instructions,
-            files=files,
+            instructions=instructions, files=files, type=type
         ).model_dump()
 
         response, _, _ = requestor.request(
@@ -92,6 +95,27 @@ class Alignment:
 
         return AlignmentResponse(**response.data)
 
+    def estimate(self, files: List[str]) -> AlignmentEstimationResponse:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        parameter_payload = AlignmentEstimationRequest(
+            files=files,
+        ).model_dump()
+
+        response, _, _ = requestor.request(
+            options=SeekrFlowRequest(
+                method="POST",
+                url="flow/alignment/estimate",
+                params=parameter_payload,
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, SeekrFlowResponse)
+        return AlignmentEstimationResponse(**response.data)
+
 
 class AsyncAlignment:
     def __init__(self, client: SeekrFlowClient) -> None:
@@ -101,14 +125,14 @@ class AsyncAlignment:
         self,
         instructions: str,
         files: List[str],
+        type: AlignmentType = AlignmentType.PRINCIPLE,
     ) -> AlignmentResponse:
         requestor = api_requestor.APIRequestor(
             client=self._client,
         )
 
         parameter_payload = AlignmentRequest(
-            instructions=instructions,
-            files=files,
+            instructions=instructions, files=files, type=type
         ).model_dump()
 
         response, _, _ = await requestor.arequest(
@@ -173,3 +197,24 @@ class AsyncAlignment:
         assert isinstance(response, SeekrFlowResponse)
 
         return AlignmentResponse(**response.data)
+
+    async def estimate(self, files: List[str]) -> AlignmentEstimationResponse:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        parameter_payload = AlignmentEstimationRequest(
+            files=files,
+        ).model_dump()
+
+        response, _, _ = await requestor.arequest(
+            options=SeekrFlowRequest(
+                method="POST",
+                url="flow/alignment/estimate",
+                params=parameter_payload,
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, SeekrFlowResponse)
+        return AlignmentEstimationResponse(**response.data)

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/resources/chat/completions.py

@@ -30,13 +30,13 @@ class ChatCompletions:
         repetition_penalty: float = 1,
         stream: bool = False,
         logprobs: bool | None = False,
-        top_logprobs: int | None = 0,
+        top_logprobs: int | None = None,
         echo: bool = False,
         n: int = 1,
         safety_model: str | None = None,
         response_format: Dict[str, str | Dict[str, Any]] | None = None,
         tools: Dict[str, str | Dict[str, Any]] | None = None,
-        tool_choice: str | Dict[str, str | Dict[str, str]] | None = None,
+        tool_choice: str | Dict[str, str | Dict[str, str]] | None = "auto",
     ) -> ChatCompletionResponse | Iterator[ChatCompletionChunk]:
         """
         Method to generate completions based on a given prompt using a specified model.
@@ -89,7 +89,6 @@ class ChatCompletions:
             ChatCompletionResponse | Iterator[ChatCompletionChunk]: Object containing the completions
             or an iterator over completion chunks.
         """
-
         requestor = api_requestor.APIRequestor(
             client=self._client,
         )
@@ -110,7 +109,7 @@ class ChatCompletions:
             n=n,
             safety_model=safety_model,
             response_format=response_format,
-            tools=tools,
+            tools=tools or [],
             tool_choice=tool_choice,
         ).model_dump()
 

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/types/__init__.py

@@ -1,9 +1,12 @@
 from seekrai.types.abstract import SeekrFlowClient
 from seekrai.types.alignment import (
+    AlignmentEstimationRequest,
+    AlignmentEstimationResponse,
     AlignmentJobStatus,
     AlignmentList,
     AlignmentRequest,
     AlignmentResponse,
+    AlignmentType,
 )
 from seekrai.types.chat_completions import (
     ChatCompletionChunk,
@@ -87,9 +90,12 @@ __all__ = [
     "ModelResponse",
     "ModelList",
     "AlignmentRequest",
+    "AlignmentEstimationRequest",
+    "AlignmentEstimationResponse",
     "AlignmentResponse",
     "AlignmentJobStatus",
     "AlignmentList",
+    "AlignmentType",
     "Project",
     "ProjectWithRuns",
     "GetProjectsResponse",

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/types/alignment.py

@@ -9,6 +9,12 @@ from pydantic import Field
 from seekrai.types.abstract import BaseModel
 
 
+class AlignmentType(str, Enum):
+    PRINCIPLE = "principle"
+    CHAIN_OF_THOUGHT = "chain_of_thought"
+    RAFT = "raft"
+
+
 class AlignmentRequest(BaseModel):
     instructions: str = Field(
         default=..., description="Task description/instructions for the alignment task"
@@ -16,6 +22,28 @@ class AlignmentRequest(BaseModel):
     files: List[str] = Field(
         default=..., description="List of file ids to use for alignment"
     )
+    type: AlignmentType = Field(
+        default=AlignmentType.PRINCIPLE,
+        description="Type of alignment task (principle, chain_of_thought, or raft)",
+    )
+
+
+class AlignmentEstimationRequest(BaseModel):
+    files: List[str] = Field(
+        default=...,
+        description="List of file ids to use to generate an alignment estimate",
+    )
+
+
+class AlignmentEstimationResponse(BaseModel):
+    input_tokens: int = Field(
+        default=...,
+        description="Estimated number of input tokens for the given files",
+    )
+    output_tokens: int = Field(
+        default=...,
+        description="Estimated number of output tokens for the given files",
+    )
 
 
 class AlignmentJobStatus(str, Enum):

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/types/finetune.py

@@ -172,6 +172,7 @@ class FinetuneResponse(BaseModel):
     inference_available: bool = False
     project_id: Optional[int] = None  # TODO - fix this
     completed_at: datetime | None = None
+    description: str | None = None
 
     # dataset token count
     # TODO

seekrai-0.4.1/src/seekrai/types/ingestion.py

@@ -0,0 +1,8 @@
+from enum import Enum
+
+
+class IngestionJobStatus(Enum):
+    QUEUED = "queued"
+    RUNNING = "running"
+    COMPLETED = "completed"
+    FAILED = "failed"

{seekrai-0.3.2 → seekrai-0.4.1}/src/seekrai/utils/files.py

@@ -47,7 +47,7 @@ def check_file(
 
     if file_size > MAX_FILE_SIZE_GB * NUM_BYTES_IN_GB:
         report_dict["message"] = (
-            f"Maximum supported file size is {MAX_FILE_SIZE_GB} GB. Found file with size of {round(file_size / NUM_BYTES_IN_GB ,3)} GB."
+            f"Maximum supported file size is {MAX_FILE_SIZE_GB} GB. Found file with size of {round(file_size / NUM_BYTES_IN_GB, 3)} GB."
         )
         report_dict["is_check_passed"] = False
     elif file_size == 0:
@@ -117,7 +117,7 @@ def _check_jsonl(file: Path) -> Dict[str, Any]:
                         report_dict["key_value"] = False
                         report_dict["message"] = (
                             f'Invalid value type for "text" key on line {idx + 1}. '
-                            f'Expected string. Found {type(json_line["text"])}.'
+                            f"Expected string. Found {type(json_line['text'])}."
                         )
 
                         report_dict["is_check_passed"] = False