retab 0.0.40__py3-none-any.whl → 0.0.42__py3-none-any.whl

retab/resources/processors/automations/outlook.py

@@ -3,7 +3,13 @@ from typing import Any, Literal, List
 from pydantic_core import PydanticUndefined
 
 from ...._resource import AsyncAPIResource, SyncAPIResource
-from ....types.automations.outlook import FetchParams, ListOutlooks, MatchParams, Outlook, UpdateOutlookRequest
+from ....types.automations.outlook import (
+    FetchParams,
+    ListOutlooks,
+    MatchParams,
+    Outlook,
+    UpdateOutlookRequest,
+)
 from ....types.standards import PreparedRequest
 
 
@@ -38,10 +44,15 @@ class OutlooksMixin:
             match_params=match_params,
             fetch_params=fetch_params,
         )
-        return PreparedRequest(method="POST", url=self.outlooks_base_url, data=outlook_data.model_dump(mode="json"))
+        return PreparedRequest(
+            method="POST",
+            url=self.outlooks_base_url,
+            data=outlook_data.model_dump(mode="json"),
+        )
 
     def prepare_list(
         self,
+        processor_id: str,
         before: str | None = None,
         after: str | None = None,
         limit: int = 10,
@@ -50,6 +61,7 @@ class OutlooksMixin:
         webhook_url: str | None = None,
     ) -> PreparedRequest:
         params = {
+            "processor_id": processor_id,
             "before": before,
             "after": after,
             "limit": limit,
@@ -92,7 +104,11 @@ class OutlooksMixin:
             fetch_params=fetch_params,
         )
 
-        return PreparedRequest(method="PUT", url=f"{self.outlooks_base_url}/{outlook_id}", data=update_outlook_request.model_dump(mode="json"))
+        return PreparedRequest(
+            method="PUT",
+            url=f"{self.outlooks_base_url}/{outlook_id}",
+            data=update_outlook_request.model_dump(mode="json"),
+        )
 
     def prepare_delete(self, outlook_id: str) -> PreparedRequest:
         return PreparedRequest(method="DELETE", url=f"{self.outlooks_base_url}/{outlook_id}")
@@ -149,12 +165,13 @@ class Outlooks(SyncAPIResource, OutlooksMixin):
         )
         response = self._client._prepared_request(request)
 
-        print(f"Outlook plugin created. Url: https://www.retab.dev/dashboard/processors/automations/{response['id']}")
+        print(f"Outlook plugin created. Url: https://www.retab.com/dashboard/processors/automations/{response['id']}")
 
         return Outlook.model_validate(response)
 
     def list(
         self,
+        processor_id: str,
         before: str | None = None,
         after: str | None = None,
         limit: int = 10,
@@ -174,7 +191,7 @@ class Outlooks(SyncAPIResource, OutlooksMixin):
         Returns:
             List[Outlook]: List of outlook plugin configurations
         """
-        request = self.prepare_list(before, after, limit, order, name, webhook_url)
+        request = self.prepare_list(processor_id, before, after, limit, order, name, webhook_url)
         response = self._client._prepared_request(request)
         return ListOutlooks.model_validate(response)
 
@@ -280,11 +297,12 @@ class AsyncOutlooks(AsyncAPIResource, OutlooksMixin):
             fetch_params=fetch_params,
         )
         response = await self._client._prepared_request(request)
-        print(f"Outlook plugin created. Url: https://www.retab.dev/dashboard/processors/automations/{response['id']}")
+        print(f"Outlook plugin created. Url: https://www.retab.com/dashboard/processors/automations/{response['id']}")
         return Outlook.model_validate(response)
 
     async def list(
         self,
+        processor_id: str,
         before: str | None = None,
         after: str | None = None,
         limit: int = 10,
@@ -292,7 +310,7 @@ class AsyncOutlooks(AsyncAPIResource, OutlooksMixin):
         name: str | None = None,
         webhook_url: str | None = None,
     ) -> ListOutlooks:
-        request = self.prepare_list(before, after, limit, order, name, webhook_url)
+        request = self.prepare_list(processor_id, before, after, limit, order, name, webhook_url)
         response = await self._client._prepared_request(request)
         return ListOutlooks.model_validate(response)
 

retab/resources/processors/automations/tests.py

@@ -1,4 +1,5 @@
 import json
+import base64
 from io import IOBase
 from pathlib import Path
 
@@ -6,7 +7,7 @@ from PIL.Image import Image
 from pydantic import HttpUrl
 
 from ...._resource import AsyncAPIResource, SyncAPIResource
-from ...._utils.mime import prepare_mime_document
+from ....utils.mime import prepare_mime_document
 from ....types.logs import AutomationLog
 from ....types.mime import MIMEData
 from ....types.standards import PreparedRequest
@@ -15,7 +16,12 @@ from ....types.standards import PreparedRequest
 class TestsMixin:
     def prepare_upload(self, automation_id: str, document: Path | str | IOBase | HttpUrl | Image | MIMEData) -> PreparedRequest:
         mime_document = prepare_mime_document(document)
-        return PreparedRequest(method="POST", url=f"/v1/processors/automations/tests/upload/{automation_id}", data={"document": mime_document.model_dump(mode="json")})
+
+        # Convert MIME document to file upload format (similar to processors client)
+        files = {"file": (mime_document.filename, base64.b64decode(mime_document.content), mime_document.mime_type)}
+
+        # Send as multipart form data with file upload
+        return PreparedRequest(method="POST", url=f"/v1/processors/automations/tests/upload/{automation_id}", files=files)
 
     def prepare_webhook(self, automation_id: str) -> PreparedRequest:
         return PreparedRequest(method="POST", url=f"/v1/processors/automations/tests/webhook/{automation_id}", data=None)

retab/resources/processors/client.py

@@ -1,7 +1,7 @@
 import base64
 from io import IOBase
 from pathlib import Path
-from typing import Any, Dict, List, Literal
+from typing import Any, List, Literal
 
 import PIL.Image
 from openai.types.chat.chat_completion_reasoning_effort import ChatCompletionReasoningEffort
@@ -9,8 +9,9 @@ from pydantic import BaseModel, HttpUrl
 from pydantic_core import PydanticUndefined
 
 from ..._resource import AsyncAPIResource, SyncAPIResource
-from ..._utils.ai_models import assert_valid_model_extraction
-from ..._utils.mime import MIMEData, prepare_mime_document
+from ...utils.ai_models import assert_valid_model_extraction
+from ...utils.json_schema import load_json_schema
+from ...utils.mime import MIMEData, prepare_mime_document
 from ...types.browser_canvas import BrowserCanvas
 from ...types.documents.extractions import RetabParsedChatCompletion
 from ...types.logs import ProcessorConfig, UpdateProcessorRequest
@@ -31,7 +32,7 @@ class ProcessorsMixin:
     def prepare_create(
         self,
         name: str,
-        json_schema: dict[str, Any],
+        json_schema: dict[str, Any] | Path | str,
         modality: Modality = "native",
         model: str = "gpt-4o-mini",
         temperature: float = PydanticUndefined,  # type: ignore[assignment]
@@ -42,9 +43,12 @@ class ProcessorsMixin:
     ) -> PreparedRequest:
         assert_valid_model_extraction(model)
 
+        # Load the JSON schema from file path, string, or dict
+        loaded_schema = load_json_schema(json_schema)
+
         processor_config = ProcessorConfig(
             name=name,
-            json_schema=json_schema,
+            json_schema=loaded_schema,
             modality=modality,
             model=model,
             temperature=temperature,
@@ -104,7 +108,7 @@ class ProcessorsMixin:
         image_resolution_dpi: int | None = None,
         browser_canvas: BrowserCanvas | None = None,
         model: str | None = None,
-        json_schema: dict[str, Any] | None = None,
+        json_schema: dict[str, Any] | Path | str | None = None,
         temperature: float | None = None,
         reasoning_effort: ChatCompletionReasoningEffort | None = None,
         n_consensus: int | None = None,
@@ -112,13 +116,18 @@ class ProcessorsMixin:
         if model is not None:
             assert_valid_model_extraction(model)
 
+        # Load the JSON schema from file path, string, or dict if provided
+        loaded_schema = None
+        if json_schema is not None:
+            loaded_schema = load_json_schema(json_schema)
+
         update_request = UpdateProcessorRequest(
             name=name,
             modality=modality,
             image_resolution_dpi=image_resolution_dpi,
             browser_canvas=browser_canvas,
             model=model,
-            json_schema=json_schema,
+            json_schema=loaded_schema,
             temperature=temperature,
             reasoning_effort=reasoning_effort,
             n_consensus=n_consensus,
@@ -203,7 +212,7 @@ class Processors(SyncAPIResource, ProcessorsMixin):
     def create(
         self,
         name: str,
-        json_schema: Dict[str, Any],
+        json_schema: dict[str, Any] | Path | str,
         modality: Modality = "native",
         model: str = "gpt-4o-mini",
         temperature: float = PydanticUndefined,  # type: ignore[assignment]
@@ -216,7 +225,7 @@ class Processors(SyncAPIResource, ProcessorsMixin):
 
         Args:
             name: Name of the processor
-            json_schema: JSON schema for the processor
+            json_schema: JSON schema for the processor. Can be a dictionary, file path (Path or str), or JSON string.
             image_resolution_dpi: Optional image resolution DPI
             browser_canvas: Optional browser canvas size
             modality: Processing modality (currently only "native" supported)
@@ -239,7 +248,7 @@ class Processors(SyncAPIResource, ProcessorsMixin):
             n_consensus=n_consensus,
         )
         response = self._client._prepared_request(request)
-        print(f"Processor ID: {response['id']}. Processor available at https://www.retab.dev/dashboard/processors/{response['id']}")
+        print(f"Processor ID: {response['id']}. Processor available at https://www.retab.com/dashboard/processors/{response['id']}")
         return ProcessorConfig.model_validate(response)
 
     def list(
@@ -295,7 +304,7 @@ class Processors(SyncAPIResource, ProcessorsMixin):
         image_resolution_dpi: int | None = None,
         browser_canvas: BrowserCanvas | None = None,
         model: str | None = None,
-        json_schema: dict[str, Any] | None = None,
+        json_schema: dict[str, Any] | Path | str | None = None,
         temperature: float | None = None,
         reasoning_effort: ChatCompletionReasoningEffort | None = None,
         n_consensus: int | None = None,
@@ -309,7 +318,7 @@ class Processors(SyncAPIResource, ProcessorsMixin):
             image_resolution_dpi: New image resolution DPI
             browser_canvas: New browser canvas size
             model: New AI model
-            json_schema: New JSON schema for the processor
+            json_schema: New JSON schema for the processor. Can be a dictionary, file path (Path or str), or JSON string.
             temperature: New temperature setting
             reasoning_effort: The effort level for the model to reason about the input data.
             n_consensus: New number of consensus required
@@ -378,7 +387,7 @@ class AsyncProcessors(AsyncAPIResource, ProcessorsMixin):
     async def create(
         self,
         name: str,
-        json_schema: Dict[str, Any],
+        json_schema: dict[str, Any] | Path | str,
         modality: Modality = "native",
         model: str = "gpt-4o-mini",
         temperature: float = PydanticUndefined,  # type: ignore[assignment]
@@ -399,7 +408,7 @@ class AsyncProcessors(AsyncAPIResource, ProcessorsMixin):
             n_consensus=n_consensus,
         )
         response = await self._client._prepared_request(request)
-        print(f"Processor ID: {response['id']}. Processor available at https://www.retab.dev/dashboard/processors/{response['id']}")
+        print(f"Processor ID: {response['id']}. Processor available at https://www.retab.com/dashboard/processors/{response['id']}")
 
         return ProcessorConfig.model_validate(response)
 
@@ -432,7 +441,7 @@ class AsyncProcessors(AsyncAPIResource, ProcessorsMixin):
         image_resolution_dpi: int | None = None,
         browser_canvas: BrowserCanvas | None = None,
         model: str | None = None,
-        json_schema: dict[str, Any] | None = None,
+        json_schema: dict[str, Any] | Path | str | None = None,
         temperature: float | None = None,
         reasoning_effort: ChatCompletionReasoningEffort | None = None,
         n_consensus: int | None = None,
@@ -446,7 +455,7 @@ class AsyncProcessors(AsyncAPIResource, ProcessorsMixin):
             image_resolution_dpi: New image resolution DPI
             browser_canvas: New browser canvas size
             model: New AI model
-            json_schema: New JSON schema for the processor
+            json_schema: New JSON schema for the processor. Can be a dictionary, file path (Path or str), or JSON string.
             temperature: New temperature setting
             reasoning_effort: The effort level for the model to reason about the input data.
             n_consensus: New number of consensus required

retab/resources/prompt_optimization.py

@@ -3,7 +3,7 @@
 # import json
 
 # from .._resource import SyncAPIResource, AsyncAPIResource
-# from .._utils.json_schema import load_json_schema
+# from ..utils.json_schema import load_json_schema
 # from ..types.jobs import JobResponse
 # from ..types.jobs.prompt_optimization import PromptOptimizationObject, PromptOptimizationProps, PromptOptimizationJobInputData, PromptOptimizationJob
 

retab/resources/schemas.py

@@ -7,9 +7,9 @@ from openai.types.chat.chat_completion_reasoning_effort import ChatCompletionRea
 from pydantic import BaseModel
 
 from .._resource import AsyncAPIResource, SyncAPIResource
-from .._utils.ai_models import assert_valid_model_schema_generation
-from .._utils.json_schema import load_json_schema
-from .._utils.mime import prepare_mime_document_list
+from ..utils.ai_models import assert_valid_model_schema_generation
+from ..utils.json_schema import load_json_schema
+from ..utils.mime import prepare_mime_document_list
 from ..types.mime import MIMEData
 from ..types.modalities import Modality
 from ..types.schemas.enhance import EnhanceSchemaConfig, EnhanceSchemaConfigDict, EnhanceSchemaRequest

retab/types/automations/mailboxes.py

@@ -17,7 +17,7 @@ class Mailbox(AutomationConfig):
     def object(self) -> str:
         return "automation.mailbox"
 
-    EMAIL_PATTERN: ClassVar[str] = f".*@{os.getenv('EMAIL_DOMAIN', 'mailbox.retab.dev')}$"
+    EMAIL_PATTERN: ClassVar[str] = f".*@{os.getenv('EMAIL_DOMAIN', 'mailbox.retab.com')}$"
     id: str = Field(default_factory=lambda: "mb_" + nanoid.generate(), description="Unique identifier for the mailbox")
 
     # Email Specific config

retab/types/completions.py

@@ -7,7 +7,7 @@ from openai.types.shared_params.reasoning import Reasoning
 from openai.types.shared_params.response_format_json_schema import ResponseFormatJSONSchema
 from pydantic import BaseModel, ConfigDict, Field
 
-from .._utils.ai_models import get_provider_for_model
+from ..utils.ai_models import get_provider_for_model
 from .ai_models import AIProvider
 from .chat import ChatCompletionRetabMessage
 

retab/types/documents/create_messages.py

@@ -10,10 +10,10 @@ from openai.types.chat.chat_completion_message_param import ChatCompletionMessag
 from openai.types.responses.response_input_param import ResponseInputItemParam
 from pydantic import BaseModel, Field, computed_field
 
-from ..._utils.chat import convert_to_anthropic_format, convert_to_google_genai_format, str_messages
-from ..._utils.chat import convert_to_openai_format as convert_to_openai_completions_api_format
-from ..._utils.display import count_image_tokens, count_text_tokens
-from ..._utils.responses import convert_to_openai_format as convert_to_openai_responses_api_format
+from ...utils.chat import convert_to_anthropic_format, convert_to_google_genai_format, str_messages
+from ...utils.chat import convert_to_openai_format as convert_to_openai_completions_api_format
+from ...utils.display import count_image_tokens, count_text_tokens
+from ...utils.responses import convert_to_openai_format as convert_to_openai_responses_api_format
 from ..chat import ChatCompletionRetabMessage
 from ..mime import MIMEData
 from ..modalities import Modality

retab/types/documents/extractions.py

@@ -15,7 +15,7 @@ from openai.types.responses.response import Response
 from openai.types.responses.response_input_param import ResponseInputItemParam
 from pydantic import BaseModel, ConfigDict, Field, ValidationInfo, computed_field, field_validator, model_validator
 
-from ..._utils.usage.usage import CostBreakdown, compute_cost_from_model, compute_cost_from_model_with_breakdown
+from ...utils.usage.usage import CostBreakdown, compute_cost_from_model, compute_cost_from_model_with_breakdown
 from ..ai_models import Amount
 from ..chat import ChatCompletionRetabMessage
 from ..mime import MIMEData
@@ -91,7 +91,7 @@ class FieldLocation(BaseModel):
     quote: str = Field(..., description="The quote of the field (verbatim from the document)")
     file_id: str | None = Field(default=None, description="The ID of the file")
     page: int | None = Field(default=None, description="The page number of the field (1-indexed)")
-    bboxes_normalized: list[tuple[float, float, float, float]] | None = Field(default=None, description="The normalized bounding boxes of the field")
+    bbox_normalized: tuple[float, float, float, float] | None = Field(default=None, description="The normalized bounding box of the field")
     score: float | None = Field(default=None, description="The score of the field")
     match_level: Literal["token", "line", "block"] | None = Field(default=None, description="The level of the match (token, line, block)")
 
@@ -99,7 +99,7 @@ class FieldLocation(BaseModel):
 class RetabParsedChoice(ParsedChoice):
     # Adaptable ParsedChoice that allows None for the finish_reason
     finish_reason: Literal["stop", "length", "tool_calls", "content_filter", "function_call"] | None = None  # type: ignore
-    field_locations: dict[str, list[FieldLocation]] | None = Field(default=None, description="The locations of the fields in the document, if available")
+    field_locations: dict[str, FieldLocation] | None = Field(default=None, description="The locations of the fields in the document, if available")
     key_mapping: dict[str, Optional[str]] | None = Field(default=None, description="Mapping of consensus keys to original model keys")
 
 

retab/types/documents/parse.py

@@ -3,6 +3,7 @@ from pydantic import BaseModel, Field
 
 from ..mime import MIMEData, BaseMIMEData
 from ..browser_canvas import BrowserCanvas
+from ..ai_models import LLMModel
 
 TableParsingFormat = Literal["markdown", "yaml", "html", "json"]
 
@@ -18,7 +19,7 @@ class ParseRequest(BaseModel):
     """Request model for document parsing."""
 
     document: MIMEData = Field(..., description="Document to parse")
-    fast_mode: bool = Field(default=False, description="Use fast mode for parsing (may reduce quality)")
+    model: LLMModel = Field(default="gemini-2.5-flash", description="Model to use for parsing")
     table_parsing_format: TableParsingFormat = Field(default="html", description="Format for parsing tables")
     image_resolution_dpi: int = Field(default=72, description="DPI for image processing")
     browser_canvas: BrowserCanvas = Field(default="A4", description="Canvas size for document rendering")
@@ -30,3 +31,4 @@ class ParseResult(BaseModel):
     document: BaseMIMEData = Field(..., description="Processed document metadata")
     usage: RetabUsage = Field(..., description="Processing usage information")
     pages: list[str] = Field(..., description="Text content of each page")
+    text: str = Field(..., description="Text content of the document")

retab/types/evals.py

@@ -6,8 +6,8 @@ from typing import Any, List, Literal, Optional
 import nanoid  # type: ignore
 from pydantic import BaseModel, Field, computed_field
 
-from .._utils.json_schema import clean_schema, compute_schema_data_id
-from .._utils.mime import generate_blake2b_hash_from_string
+from ..utils.json_schema import clean_schema, compute_schema_data_id
+from ..utils.mime import generate_blake2b_hash_from_string
 from .ai_models import Amount
 from .inference_settings import InferenceSettings
 from .mime import MIMEData

retab/types/evaluations/iterations.py

@@ -6,8 +6,8 @@ from typing import Any, Optional, Self
 import nanoid  # type: ignore
 from pydantic import BaseModel, Field, computed_field, model_validator
 
-from ..._utils.json_schema import clean_schema
-from ..._utils.mime import generate_blake2b_hash_from_string
+from ...utils.json_schema import clean_schema
+from ...utils.mime import generate_blake2b_hash_from_string
 from ..inference_settings import InferenceSettings
 from ..metrics import MetricResult
 from ..predictions import PredictionData

retab/types/evaluations/model.py

@@ -5,8 +5,8 @@ from typing import Any, Optional
 import nanoid  # type: ignore
 from pydantic import BaseModel, Field, computed_field
 
-from ..._utils.json_schema import compute_schema_data_id
-from ..._utils.mime import generate_blake2b_hash_from_string
+from ...utils.json_schema import compute_schema_data_id
+from ...utils.mime import generate_blake2b_hash_from_string
 from ..inference_settings import InferenceSettings
 from .documents import EvaluationDocument
 from .iterations import Iteration

retab/types/extractions.py

@@ -3,13 +3,19 @@ from typing import Any, Literal, Optional
 
 import nanoid  # type: ignore
 from openai.types.chat import ChatCompletion
-from openai.types.chat.chat_completion_reasoning_effort import ChatCompletionReasoningEffort
+from openai.types.chat.chat_completion_reasoning_effort import (
+    ChatCompletionReasoningEffort,
+)
 from pydantic import BaseModel, Field, computed_field, model_validator
 
 from retab.types.chat import ChatCompletionRetabMessage
 from retab.types.documents.extractions import RetabParsedChatCompletion
 
-from .._utils.usage.usage import CostBreakdown, compute_cost_from_model, compute_cost_from_model_with_breakdown
+from ..utils.usage.usage import (
+    CostBreakdown,
+    compute_cost_from_model,
+    compute_cost_from_model_with_breakdown,
+)
 from .ai_models import Amount
 from .modalities import Modality
 
@@ -17,9 +23,18 @@ ValidationsState = Literal["pending", "validated", "invalid"]
 
 
 class ExtractionSource(BaseModel):
-    type: Literal["api", "annotation", "processor", "automation.link", "automation.mailbox", "automation.cron", "automation.outlook", "automation.endpoint", "schema.extract"] = (
-        Field(description="Type of extraction")
-    )
+    type: Literal[
+        "api",
+        "annotation",
+        "processor",
+        "automation",
+        "automation.link",
+        "automation.mailbox",
+        "automation.cron",
+        "automation.outlook",
+        "automation.endpoint",
+        "schema.extract",
+    ] = Field(description="Type of extraction")
     id: str | None = Field(default=None, description="ID the trigger of the extraction")
 
 
@@ -34,7 +49,10 @@ class ExtractionTimingStep(BaseModel):
 
 
 class Extraction(BaseModel):
-    id: str = Field(default_factory=lambda: "extr_" + nanoid.generate(), description="Unique identifier of the analysis")
+    id: str = Field(
+        default_factory=lambda: "extr_" + nanoid.generate(),
+        description="Unique identifier of the analysis",
+    )
     messages: list[ChatCompletionRetabMessage] = Field(default_factory=list)
     messages_gcs: str = Field(..., description="GCS path to the messages")
     file_gcs_paths: list[str] = Field(..., description="GCS paths to the files")
@@ -51,16 +69,23 @@ class Extraction(BaseModel):
     source: ExtractionSource = Field(..., description="Source of the extraction")
     image_resolution_dpi: int = Field(default=96, description="Resolution of the image sent to the LLM")
     browser_canvas: BrowserCanvas = Field(
-        default="A4", description="Sets the size of the browser canvas for rendering documents in browser-based processing. Choose a size that matches the document type."
+        default="A4",
+        description="Sets the size of the browser canvas for rendering documents in browser-based processing. Choose a size that matches the document type.",
     )
     modality: Modality = Field(default="native", description="Modality of the extraction")
-    reasoning_effort: Optional[ChatCompletionReasoningEffort] = Field(default=None, description="The effort level for the model to reason about the input data.")
+    reasoning_effort: Optional[ChatCompletionReasoningEffort] = Field(
+        default=None,
+        description="The effort level for the model to reason about the input data.",
+    )
     timings: list[ExtractionTimingStep] = Field(default_factory=list, description="Timings of the extraction")
 
     # Infered from the schema
     schema_id: str = Field(..., description="Version of the schema used for the analysis")
     schema_data_id: str = Field(..., description="Version of the schema data used for the analysis")
-    created_at: datetime.datetime = Field(default_factory=lambda: datetime.datetime.now(datetime.timezone.utc), description="Timestamp of the creation of the extraction object")
+    created_at: datetime.datetime = Field(
+        default_factory=lambda: datetime.datetime.now(datetime.timezone.utc),
+        description="Timestamp of the creation of the extraction object",
+    )
     request_at: datetime.datetime | None = Field(default=None, description="Timestamp of the extraction request if provided.")
     organization_id: str = Field(..., description="Organization ID of the user or application")
     validation_state: Optional[ValidationsState] = Field(default=None, description="Validation state of the extraction")

retab/types/jobs/prompt_optimization.py

@@ -1,7 +1,7 @@
 # from typing import Literal, Any
 # from pydantic import BaseModel, computed_field
 # from ..mime import MIMEData
-# from ..._utils.benchmarking import ExtractionAnalysis
+# from ...utils.benchmarking import ExtractionAnalysis
 
 # MAX_CONCURRENCY = 15
 

retab/types/logs.py

@@ -7,9 +7,9 @@ from openai.types.chat.chat_completion import ChatCompletion
 from openai.types.chat.chat_completion_reasoning_effort import ChatCompletionReasoningEffort
 from pydantic import BaseModel, EmailStr, Field, HttpUrl, computed_field, field_validator
 
-from .._utils.json_schema import compute_schema_data_id
-from .._utils.mime import generate_blake2b_hash_from_string
-from .._utils.usage.usage import CostBreakdown, compute_cost_from_model, compute_cost_from_model_with_breakdown
+from ..utils.json_schema import compute_schema_data_id
+from ..utils.mime import generate_blake2b_hash_from_string
+from ..utils.usage.usage import CostBreakdown, compute_cost_from_model, compute_cost_from_model_with_breakdown
 from .ai_models import Amount
 from .documents.extractions import RetabParsedChatCompletion
 from .mime import BaseMIMEData

retab/types/schemas/object.py

@@ -10,9 +10,9 @@ from openai.types.chat.chat_completion_message_param import ChatCompletionMessag
 from openai.types.responses.response_input_param import ResponseInputItemParam
 from pydantic import BaseModel, Field, PrivateAttr, computed_field, model_validator
 
-from ..._utils.chat import convert_to_anthropic_format, convert_to_google_genai_format
-from ..._utils.chat import convert_to_openai_format as convert_to_openai_completions_api_format
-from ..._utils.json_schema import (
+from ...utils.chat import convert_to_anthropic_format, convert_to_google_genai_format
+from ...utils.chat import convert_to_openai_format as convert_to_openai_completions_api_format
+from ...utils.json_schema import (
     convert_basemodel_to_partial_basemodel,
     convert_json_schema_to_basemodel,
     create_reasoning_schema,
@@ -25,7 +25,7 @@ from ..._utils.json_schema import (
     load_json_schema,
     schema_to_ts_type,
 )
-from ..._utils.responses import convert_to_openai_format as convert_to_openai_responses_api_format
+from ...utils.responses import convert_to_openai_format as convert_to_openai_responses_api_format
 from ...types.standards import StreamingBaseModel
 from ..chat import ChatCompletionRetabMessage
 

retab/types/schemas/templates.py

@@ -4,7 +4,7 @@ from typing import Any, Literal, Optional
 import nanoid  # type: ignore
 from pydantic import BaseModel, Field, PrivateAttr, computed_field
 
-from ..._utils.json_schema import generate_schema_data_id, generate_schema_id
+from ...utils.json_schema import generate_schema_data_id, generate_schema_id
 from ...types.mime import MIMEData
 
 

retab/utils/_model_cards/anthropic.yaml

@@ -0,0 +1,59 @@
+- model: "claude-3-5-sonnet-latest"
+  pricing:
+    text:
+      prompt: 3.00
+      cached_discount: 0.5
+      completion: 15.00
+    audio: null
+  capabilities:
+    modalities: ["text", "image"]
+    endpoints: ["chat_completions"]
+    features: ["streaming", "function_calling"]
+  permissions:
+    show_in_free_picker: true
+    show_in_paid_picker: true
+
+- model: "claude-3-5-sonnet-20241022"
+  inherits: "claude-3-5-sonnet-latest"
+  permissions:
+    show_in_free_picker: false
+    show_in_paid_picker: false
+
+- model: "claude-3-opus-20240229"
+  pricing:
+    text:
+      prompt: 15.00
+      cached_discount: 0.5
+      completion: 75.00
+    audio: null
+  capabilities:
+    modalities: ["text", "image"]
+    endpoints: ["chat_completions"]
+    features: ["streaming", "function_calling"]
+  permissions:
+    show_in_free_picker: true
+    show_in_paid_picker: true
+
+- model: "claude-3-sonnet-20240229"
+  pricing:
+    text:
+      prompt: 3.00
+      cached_discount: 0.5
+      completion: 15.00
+    audio: null
+  capabilities:
+    modalities: ["text", "image"]
+    endpoints: ["chat_completions"]
+    features: ["streaming", "function_calling"]
+
+- model: "claude-3-haiku-20240307"
+  pricing:
+    text:
+      prompt: 0.25
+      cached_discount: 0.5
+      completion: 1.25
+    audio: null
+  capabilities:
+    modalities: ["text", "image"]
+    endpoints: ["chat_completions"]
+    features: ["streaming", "function_calling"]