retab 0.0.87__py3-none-any.whl → 0.0.89__py3-none-any.whl

retab/client.py

@@ -10,7 +10,7 @@ import backoff.types
 import httpx
 import truststore
 
-from .resources import documents, models, schemas, projects, extractions, edit
+from .resources import documents, models, schemas, projects, extractions, edit, workflows
 from .types.standards import PreparedRequest, FieldUnset
 
 
@@ -189,7 +189,7 @@ class Retab(BaseRetab):
         self.models = models.Models(client=self)
         self.schemas = schemas.Schemas(client=self)
         self.edit = edit.Edit(client=self)
-
+        self.workflows = workflows.Workflows(client=self)
     def _request(
         self,
         method: str,
@@ -487,7 +487,8 @@ class AsyncRetab(BaseRetab):
         self.models = models.AsyncModels(client=self)
         self.schemas = schemas.AsyncSchemas(client=self)
         self.edit = edit.AsyncEdit(client=self)
-
+        self.workflows = workflows.AsyncWorkflows(client=self)
+        
     def _parse_response(self, response: httpx.Response) -> Any:
         """Parse response based on content-type.
 

retab/resources/workflows/runs/client.py

@@ -1,6 +1,6 @@
 from io import IOBase
 from pathlib import Path
-from typing import Any, Dict
+from typing import Any, Dict, Optional
 
 import PIL.Image
 from pydantic import HttpUrl
@@ -21,15 +21,19 @@ class WorkflowRunsMixin:
     def prepare_create(
         self,
         workflow_id: str,
-        documents: Dict[str, DocumentInput],
+        documents: Optional[Dict[str, DocumentInput]] = None,
+        json_inputs: Optional[Dict[str, Dict[str, Any]]] = None,
+        text_inputs: Optional[Dict[str, str]] = None,
     ) -> PreparedRequest:
-        """Prepare a request to run a workflow with input documents.
+        """Prepare a request to run a workflow with input documents, JSON data, and/or text data.
 
         Args:
             workflow_id: The ID of the workflow to run
             documents: Mapping of start node IDs to their input documents.
                        Each document can be a file path, bytes, file-like object,
                        MIMEData, PIL Image, or HttpUrl.
+            json_inputs: Mapping of start_json node IDs to their input JSON data.
+            text_inputs: Mapping of start_text node IDs to their input text.
 
         Returns:
             PreparedRequest: The prepared request
@@ -40,20 +44,37 @@ class WorkflowRunsMixin:
             ...     documents={
             ...         "start-node-1": Path("invoice.pdf"),
             ...         "start-node-2": Path("receipt.pdf"),
+            ...     },
+            ...     json_inputs={
+            ...         "json-node-1": {"key": "value"},
+            ...     },
+            ...     text_inputs={
+            ...         "text-node-1": "Hello, world!",
             ...     }
             ... )
         """
+        data: Dict[str, Any] = {}
+
         # Convert each document to MIMEData and then to the format expected by the backend
-        documents_payload: Dict[str, Dict[str, Any]] = {}
-        for node_id, document in documents.items():
-            mime_data = prepare_mime_document(document)
-            documents_payload[node_id] = {
-                "filename": mime_data.filename,
-                "content": mime_data.content,
-                "mime_type": mime_data.mime_type,
-            }
-
-        data = {"documents": documents_payload}
+        if documents:
+            documents_payload: Dict[str, Dict[str, Any]] = {}
+            for node_id, document in documents.items():
+                mime_data = prepare_mime_document(document)
+                documents_payload[node_id] = {
+                    "filename": mime_data.filename,
+                    "content": mime_data.content,
+                    "mime_type": mime_data.mime_type,
+                }
+            data["documents"] = documents_payload
+
+        # Add JSON inputs directly
+        if json_inputs:
+            data["json_inputs"] = json_inputs
+
+        # Add text inputs directly
+        if text_inputs:
+            data["text_inputs"] = text_inputs
+
         return PreparedRequest(method="POST", url=f"/v1/workflows/{workflow_id}/run", data=data)
 
     def prepare_get(self, run_id: str) -> PreparedRequest:
@@ -77,9 +98,11 @@ class WorkflowRuns(SyncAPIResource, WorkflowRunsMixin):
     def create(
         self,
         workflow_id: str,
-        documents: Dict[str, DocumentInput],
+        documents: Optional[Dict[str, DocumentInput]] = None,
+        json_inputs: Optional[Dict[str, Dict[str, Any]]] = None,
+        text_inputs: Optional[Dict[str, str]] = None,
     ) -> WorkflowRun:
-        """Run a workflow with the provided input documents.
+        """Run a workflow with the provided inputs.
 
         This creates a workflow run and starts execution in the background.
         The returned WorkflowRun will have status "running" - use get()
@@ -90,25 +113,37 @@ class WorkflowRuns(SyncAPIResource, WorkflowRunsMixin):
             documents: Mapping of start node IDs to their input documents.
                        Each document can be a file path, bytes, file-like object,
                        MIMEData, PIL Image, or HttpUrl.
+            json_inputs: Mapping of start_json node IDs to their input JSON data.
+            text_inputs: Mapping of start_text node IDs to their input text.
 
         Returns:
             WorkflowRun: The created workflow run with status "running"
 
         Raises:
             HTTPException: If the request fails (e.g., workflow not found,
-                          missing input documents for start nodes)
+                          missing inputs for start nodes)
 
         Example:
             >>> run = client.workflows.runs.create(
             ...     workflow_id="wf_abc123",
             ...     documents={
             ...         "start-node-1": Path("invoice.pdf"),
-            ...         "start-node-2": Path("receipt.pdf"),
+            ...     },
+            ...     json_inputs={
+            ...         "json-node-1": {"key": "value"},
+            ...     },
+            ...     text_inputs={
+            ...         "text-node-1": "Hello, world!",
             ...     }
             ... )
             >>> print(f"Run started: {run.id}, status: {run.status}")
         """
-        request = self.prepare_create(workflow_id=workflow_id, documents=documents)
+        request = self.prepare_create(
+            workflow_id=workflow_id,
+            documents=documents,
+            json_inputs=json_inputs,
+            text_inputs=text_inputs,
+        )
         response = self._client._prepared_request(request)
         return WorkflowRun.model_validate(response)
 
@@ -138,9 +173,11 @@ class AsyncWorkflowRuns(AsyncAPIResource, WorkflowRunsMixin):
     async def create(
         self,
         workflow_id: str,
-        documents: Dict[str, DocumentInput],
+        documents: Optional[Dict[str, DocumentInput]] = None,
+        json_inputs: Optional[Dict[str, Dict[str, Any]]] = None,
+        text_inputs: Optional[Dict[str, str]] = None,
     ) -> WorkflowRun:
-        """Run a workflow with the provided input documents.
+        """Run a workflow with the provided inputs.
 
         This creates a workflow run and starts execution in the background.
         The returned WorkflowRun will have status "running" - use get()
@@ -151,25 +188,37 @@ class AsyncWorkflowRuns(AsyncAPIResource, WorkflowRunsMixin):
             documents: Mapping of start node IDs to their input documents.
                        Each document can be a file path, bytes, file-like object,
                        MIMEData, PIL Image, or HttpUrl.
+            json_inputs: Mapping of start_json node IDs to their input JSON data.
+            text_inputs: Mapping of start_text node IDs to their input text.
 
         Returns:
             WorkflowRun: The created workflow run with status "running"
 
         Raises:
             HTTPException: If the request fails (e.g., workflow not found,
-                          missing input documents for start nodes)
+                          missing inputs for start nodes)
 
         Example:
             >>> run = await client.workflows.runs.create(
             ...     workflow_id="wf_abc123",
             ...     documents={
             ...         "start-node-1": Path("invoice.pdf"),
-            ...         "start-node-2": Path("receipt.pdf"),
+            ...     },
+            ...     json_inputs={
+            ...         "json-node-1": {"key": "value"},
+            ...     },
+            ...     text_inputs={
+            ...         "text-node-1": "Hello, world!",
             ...     }
             ... )
             >>> print(f"Run started: {run.id}, status: {run.status}")
         """
-        request = self.prepare_create(workflow_id=workflow_id, documents=documents)
+        request = self.prepare_create(
+            workflow_id=workflow_id,
+            documents=documents,
+            json_inputs=json_inputs,
+            text_inputs=text_inputs,
+        )
         response = await self._client._prepared_request(request)
         return WorkflowRun.model_validate(response)
 

retab/types/documents/split.py

@@ -5,6 +5,7 @@ from ..mime import MIMEData
 class Category(BaseModel):
     name: str = Field(..., description="The name of the category")
     description: str = Field(..., description="The description of the category")
+    partition_key: str | None = Field(default=None, description="The key to partition the category")
 
 
 class SplitRequest(BaseModel):
@@ -13,20 +14,32 @@ class SplitRequest(BaseModel):
     model: str = Field(default="retab-small", description="The model to use to split the document")
 
 
+class Partition(BaseModel):
+    key: str = Field(..., description="The partition key value (e.g., property ID, invoice number)")
+    pages: list[int] = Field(..., description="The pages of the partition (1-indexed)")
+    first_page_y_start: float = Field(default=0.0, description="The y coordinate of the first page of the partition")
+    last_page_y_end: float = Field(default=1.0, description="The y coordinate of the last page of the partition")
+
 class SplitResult(BaseModel):
     name: str = Field(..., description="The name of the category")
-    start_page: int = Field(..., description="The start page of the category (1-indexed)")
-    end_page: int = Field(..., description="The end page of the category (1-indexed, inclusive)")
+    pages: list[int] = Field(..., description="The pages of the category (1-indexed)")
+    partitions: list[Partition] = Field(default_factory=list, description="The partitions of the category")
 
 
 class SplitResponse(BaseModel):
     splits: list[SplitResult] = Field(..., description="The list of document splits with their page ranges")
 
 
+class SplitOutputItem(BaseModel):
+    """Internal schema item for LLM structured output validation."""
+    name: str = Field(..., description="The name of the category")
+    start_page: int = Field(..., description="The start page of the category (1-indexed)")
+    end_page: int = Field(..., description="The end page of the category (1-indexed, inclusive)")
+
 
 class SplitOutputSchema(BaseModel):
     """Schema for LLM structured output."""
-    splits: list[SplitResult] = Field(
-        ..., 
+    splits: list[SplitOutputItem] = Field(
+        ...,
         description="List of document sections, each classified into one of the provided categories with their page ranges"
     )

{retab-0.0.87.dist-info → retab-0.0.89.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: retab
-Version: 0.0.87
+Version: 0.0.89
 Summary: Retab official python library
 Home-page: https://github.com/retab-dev/retab
 Author: Retab

{retab-0.0.87.dist-info → retab-0.0.89.dist-info}/RECORD

@@ -1,6 +1,6 @@
 retab/__init__.py,sha256=s4GawWTRBYz4VY-CyAV5-ZdFtdw8V5oopGIYm9GgdSo,188
 retab/_resource.py,sha256=JfAU4UTa05ugWfbrpO7fsVr_pFewht99NkoIfK6kBQM,577
-retab/client.py,sha256=VrOzEtZQPR4uydO8QJJYkMOoAiC1TfPbkXmTnatSQ0w,30172
+retab/client.py,sha256=ExQLR-xwFKIwqA1DoH3JxI1BU2RB7kWAiMbwR073w1c,30311
 retab/generate_types.py,sha256=cUu1IX65uU__MHivmEb_PZtzAi8DYsvppZvcY30hj90,8425
 retab/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 retab/resources/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -21,7 +21,7 @@ retab/resources/projects/client.py,sha256=5LPAhJt5-nqBP4VWYvo0k7cW6HLGF6K9xMiHKQ
 retab/resources/workflows/__init__.py,sha256=-I0QNX7XKEr8ZJTV4-awMyKxZqGlSkKMdibiHiB7cZ0,89
 retab/resources/workflows/client.py,sha256=G1dYV66Wsas_QWQ9O2N7s1VUt72TP1W1ZG-_cEWEURM,755
 retab/resources/workflows/runs/__init__.py,sha256=5hPZ-70StN0U8bOlhm9H_ZXFljBjy8VoWQRu1_cGAVM,101
-retab/resources/workflows/runs/client.py,sha256=8l87Sf5RNNLIJNyhCwCprqA9ffq3J9zSlwoQHdyrEN4,6771
+retab/resources/workflows/runs/client.py,sha256=GopedV363XnGl0mL3bZHWaOay12uAeTqq4iIEJSadMA,8739
 retab/types/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 retab/types/chat.py,sha256=x9VbtPMa4w6Gc0HrFC3ILl6cCnfEn5ytDnwJtZmlcys,1436
 retab/types/inference_settings.py,sha256=wIivYffvEE7v6lhbjbhAZGssK4uYr64Oq6cZKxzY5_M,1131
@@ -36,7 +36,7 @@ retab/types/documents/create_messages.py,sha256=Uym0SnVUGkyt1C5AOD37BsZ3puyeu_ig
 retab/types/documents/edit.py,sha256=b6UcYLOJkClpMu4QyYmdp-X4WtN8U_3oiMBc1KLklVY,5663
 retab/types/documents/extract.py,sha256=x_59fm69-icsxxGRgpFd0NN-SLRoMYqbvfCZuG7zyGc,18033
 retab/types/documents/parse.py,sha256=MXe7zh3DusWQhGe0Sr95nPy6cB8DRX8MA4Hmjj_AP7E,1300
-retab/types/documents/split.py,sha256=xRdJ6IpSRAPi_ZtAG2FNqg5A-v5tzfb1QQkW5UfO2pY,1246
+retab/types/documents/split.py,sha256=Bjk5iJdS3v7I3rCvqpFUPlzgO4HINqh3uMPQJg-MqPc,2166
 retab/types/edit/__init__.py,sha256=M8hF97h7fX8RP9IsB6qpkw0eyvO0DFQvP6FmWL8caCQ,331
 retab/types/edit/templates.py,sha256=RLRIMdXzU-5_3XPf0iMSozjRTAP5Tliq0nrjlZn0l8E,2412
 retab/types/extractions/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -59,7 +59,7 @@ retab/utils/hashing.py,sha256=_BMVUvftOcJav68QL0rLkH2dbhW9RRJPzeGC2akR0fc,757
 retab/utils/json_schema.py,sha256=zP4pQLpVHBKWo_abCjb_dU4kA0azhHopd-1TFUgVEvc,20655
 retab/utils/mime.py,sha256=mTP_lqSPttOP5DYJxopiWaeFXrUCPjhwd7y53nCVGO4,6189
 retab/utils/stream_context_managers.py,sha256=gI1gVQSj3nWz6Mvjz7Ix5AiY0g6vSL-c2tPfuP04izo,2314
-retab-0.0.87.dist-info/METADATA,sha256=Rz6B3ctJWOHF0hcaFxc2hEyBgpeBRgvScGxFNGjALMg,4532
-retab-0.0.87.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-retab-0.0.87.dist-info/top_level.txt,sha256=waQR0EGdhLIQtztoE3AXg7ik5ONQ9q_bsKVpyFuJdq0,6
-retab-0.0.87.dist-info/RECORD,,
+retab-0.0.89.dist-info/METADATA,sha256=1ppp_sgtdC53grfu4xxD91N_-BDa7FBdofWz_Vd1WTw,4532
+retab-0.0.89.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+retab-0.0.89.dist-info/top_level.txt,sha256=waQR0EGdhLIQtztoE3AXg7ik5ONQ9q_bsKVpyFuJdq0,6
+retab-0.0.89.dist-info/RECORD,,