anyformat 2.6.0__tar.gz → 2.7.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (26) hide show
  1. {anyformat-2.6.0 → anyformat-2.7.0}/PKG-INFO +1 -1
  2. {anyformat-2.6.0 → anyformat-2.7.0}/pyproject.toml +1 -1
  3. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/_http.py +8 -0
  4. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/__init__.py +2 -3
  5. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/nodes.py +20 -7
  6. anyformat-2.7.0/src/anyformat/workflow/upstream.py +58 -0
  7. {anyformat-2.6.0 → anyformat-2.7.0}/README.md +0 -0
  8. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/__init__.py +0 -0
  9. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/_banner.py +0 -0
  10. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/_source.py +0 -0
  11. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/_viz.py +0 -0
  12. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/builder.py +0 -0
  13. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/cli.py +0 -0
  14. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/client.py +0 -0
  15. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/errors.py +0 -0
  16. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/handles.py +0 -0
  17. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/result.py +0 -0
  18. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/sdk/views.py +0 -0
  19. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/builder.py +0 -0
  20. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/checks.py +0 -0
  21. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/definition.py +0 -0
  22. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/fields.py +0 -0
  23. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/results.py +0 -0
  24. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/schema.py +0 -0
  25. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/template_tokens.py +0 -0
  26. {anyformat-2.6.0 → anyformat-2.7.0}/src/anyformat/workflow/topology.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: anyformat
3
- Version: 2.6.0
3
+ Version: 2.7.0
4
4
  Summary: Python SDK for the AnyFormat API
5
5
  Project-URL: Homepage, https://anyformat.ai
6
6
  Project-URL: Documentation, https://docs.anyformat.ai
@@ -1,6 +1,6 @@
1
1
  [project]
2
2
  name = "anyformat"
3
- version = "2.6.0"
3
+ version = "2.7.0"
4
4
  description = "Python SDK for the AnyFormat API"
5
5
  readme = "README.md"
6
6
  authors = [
@@ -64,6 +64,11 @@ class PacketFile:
64
64
  # otherwise. Only the upload/create response carries it — the pre-rename name
65
65
  # isn't persisted, so later reads expose ``name`` alone.
66
66
  original_name: str | None = None
67
+ # Set only on a packet read, which signs ``download_url`` to the file's original bytes until
68
+ # ``download_url_expires_at``; all three are ``None`` when the bytes cannot be linked.
69
+ content_type: str | None = None
70
+ download_url: str | None = None
71
+ download_url_expires_at: str | None = None
67
72
 
68
73
 
69
74
  @dataclass(frozen=True)
@@ -1080,6 +1085,9 @@ def parse_packet_file(data: dict) -> PacketFile:
1080
1085
  id=data["id"],
1081
1086
  name=data.get("name", ""),
1082
1087
  original_name=data.get("original_name"),
1088
+ content_type=data.get("content_type"),
1089
+ download_url=data.get("download_url"),
1090
+ download_url_expires_at=data.get("download_url_expires_at"),
1083
1091
  )
1084
1092
 
1085
1093
 
@@ -2,9 +2,8 @@
2
2
  # storage-format methods (no to_config / from_config / @model_serializer
3
3
  # beyond what pydantic ships): storage shape == wire shape ==
4
4
  # ``node.model_dump(exclude_none=True)``. Runtime Node->Task construction
5
- # lives in ``anyformat_core.task_builders``; the dramatiq worker actor
6
- # dispatches via ``TASK_BUILDERS`` at message receipt, so the backend
7
- # only emits typed nodes and never imports a Task class.
5
+ # is ``anyformat.domain.node_plan.plan_node``, which the core compiler calls
6
+ # per node, so the backend only emits typed nodes and never imports a Task class.
8
7
  from anyformat.workflow.builder import WorkflowBuilder, WorkflowBuilderError
9
8
  from anyformat.workflow.checks import (
10
9
  AllOfCheck,
@@ -300,7 +300,7 @@ class ExtractionSchema(BaseModel):
300
300
 
301
301
  class ExtractNode(BaseNode):
302
302
  type: Literal["extract"]
303
- # `mode="agentic"` flips ExtractionTask.agentic_enabled. Legacy data
303
+ # `mode="agentic"` plans the agentic extract (`plan_node`). Legacy data
304
304
  # may carry it; the typed builder does not yet expose it as a kwarg.
305
305
  # `mode="lite"` is a standard extract on a cheaper model (gpt-5.6-luna).
306
306
  mode: Literal["standard", "agentic", "max", "lite"] = Field(
@@ -392,7 +392,7 @@ class ExtractNode(BaseNode):
392
392
  @property
393
393
  def no_confidence_reason(self) -> str | None:
394
394
  """Why this extract scores no confidence, or None when it does. An agentic extract
395
- compiles before the row extractor, so it scores either way."""
395
+ wins over the row extractor in `plan_node`, so it scores either way."""
396
396
  if not self.confidence:
397
397
  return "confidence is off"
398
398
  if self.mode == "agentic":
@@ -607,10 +607,15 @@ EmailAddress = Annotated[
607
607
  ]
608
608
 
609
609
 
610
- # Caps each of to and cc, so the email a node sends per split has a bounded fan-out.
610
+ # SES refuses a send naming more addresses than this across to, cc and bcc together.
611
611
  MAX_EMAIL_RECIPIENTS = 50
612
612
 
613
613
 
614
+ def require_within_recipient_cap(to: list[str], cc: list[str]) -> None:
615
+ if len(to) + len(cc) > MAX_EMAIL_RECIPIENTS:
616
+ raise ValueError(f"to and cc together name more than {MAX_EMAIL_RECIPIENTS} addresses")
617
+
618
+
614
619
  class EmailBody(BaseModel):
615
620
  """Where an Email node's body comes from: its own template, or an upstream report."""
616
621
 
@@ -651,16 +656,18 @@ class EmailAlertNode(BaseNode):
651
656
  """
652
657
 
653
658
  type: Literal["email_alert"]
654
- to: list[EmailAddress] = Field(min_length=1, max_length=MAX_EMAIL_RECIPIENTS, description="Recipients.")
659
+ to: list[EmailAddress] = Field(
660
+ min_length=1,
661
+ max_length=MAX_EMAIL_RECIPIENTS,
662
+ description=f"Recipients; to and cc together name at most {MAX_EMAIL_RECIPIENTS} addresses.",
663
+ )
655
664
  cc: list[EmailAddress] = Field(
656
665
  default_factory=list, max_length=MAX_EMAIL_RECIPIENTS, description="Copied recipients."
657
666
  )
658
667
  reply_to: Annotated[EmailAddress | None, BeforeValidator(_empty_str_to_none)] = Field(
659
668
  default=None, description="Where replies go."
660
669
  )
661
- subject_template: str = Field(
662
- min_length=1, description="Subject with ``${field.*}`` tokens, on one line."
663
- )
670
+ subject_template: str = Field(min_length=1, description="Subject with ``${field.*}`` tokens, on one line.")
664
671
  body: EmailBody
665
672
  on_missing: Literal["hold", "send"] = Field(
666
673
  default="hold",
@@ -685,6 +692,11 @@ class EmailAlertNode(BaseNode):
685
692
  raise ValueError("an address appears more than once across to and cc")
686
693
  return self
687
694
 
695
+ @model_validator(mode="after")
696
+ def _within_the_recipient_cap(self) -> "EmailAlertNode":
697
+ require_within_recipient_cap(self.to, self.cc)
698
+ return self
699
+
688
700
  @classmethod
689
701
  def accepts(cls, predecessor: type["BaseNode"]) -> bool:
690
702
  return predecessor in (ExtractNode, ValidateNode, IfElseNode)
@@ -958,6 +970,7 @@ __all__ = [
958
970
  "ExtractNode",
959
971
  "MAX_CODE_BYTES",
960
972
  "MAX_EMAIL_RECIPIENTS",
973
+ "require_within_recipient_cap",
961
974
  "MAX_EXAMPLE_REPORT_BYTES",
962
975
  "PUBLIC_HIDDEN_NODE_KEYS",
963
976
  "strip_public_hidden_keys",
@@ -0,0 +1,58 @@
1
+ """Which upstream node a sink reads: the run compiler (anyformat-core) and the Email preview (backend) share it."""
2
+
3
+ from __future__ import annotations
4
+
5
+ from dataclasses import dataclass
6
+
7
+ # Node types that write no extraction of their own: smart_lookup enriches it in place, validate reads it,
8
+ # and if_else routes on it, so a consumer below them reads the extract above.
9
+ # Frontend mirror: ``services/frontend/src/features/studio/nodes/_shared/hooks/useUpstreamExtractFields.ts``.
10
+ EXTRACTION_TRANSPARENT_TYPES = frozenset({"smart_lookup", "validate", "if_else"})
11
+
12
+
13
+ @dataclass(frozen=True)
14
+ class UpstreamGraph:
15
+ """A graph as each node's type by id and its ``(source, target)`` edges."""
16
+
17
+ types: dict[str, str]
18
+ edges: tuple[tuple[str, str], ...]
19
+
20
+ def single_upstream(self, node_id: str) -> str:
21
+ """``ValueError`` unless exactly one edge enters the node."""
22
+ sources = [source for source, target in self.edges if target == node_id]
23
+ if not sources:
24
+ raise ValueError(f"Node '{node_id}' has no predecessors in the graph")
25
+ if len(sources) > 1:
26
+ raise ValueError(f"Node '{node_id}' has multiple predecessors: {sources}")
27
+ return sources[0]
28
+
29
+ def extraction_source(self, node_id: str) -> str:
30
+ """The node whose extraction ``node_id`` reads, past every extraction-transparent node; ``ValueError``
31
+ when a node on the way has no single predecessor."""
32
+ current = self.single_upstream(node_id)
33
+ seen: set[str] = set()
34
+ while self.types.get(current) in EXTRACTION_TRANSPARENT_TYPES:
35
+ if current in seen:
36
+ raise ValueError(
37
+ f"Cycle through extraction-transparent nodes while resolving extraction source for '{node_id}'"
38
+ )
39
+ seen.add(current)
40
+ current = self.single_upstream(current)
41
+ return current
42
+
43
+ def classification_source(self, node_id: str) -> str | None:
44
+ """The classify node closest above ``node_id``; ``None`` when there is none or the path forks.
45
+
46
+ A nested arm sees every classifier on its path; only the nearest names its document.
47
+ """
48
+ current = node_id
49
+ seen: set[str] = set()
50
+ while current not in seen:
51
+ seen.add(current)
52
+ sources = {source for source, target in self.edges if target == current}
53
+ if len(sources) != 1:
54
+ return None
55
+ current = sources.pop()
56
+ if self.types.get(current) == "classify":
57
+ return current
58
+ return None
File without changes