lda.chat/lda-wf
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

first recipe: extract textcontent from output

Browse Source
This commit is contained in:
lda
2026-05-29 22:58:08 +07:00 Verified
parent 4211d1b7fa
commit 222447f821
5 changed files with 191 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/wf_authoring/__init__.py
+14
View File
@@ -75,6 +75,14 @@ from .nodes import (
outcome,
)
from .reducers import AuthoredReducer, ReducerCatalog, reducer
from .recipes import (
ExtractTextContentInput,
ExtractTextContentOutput,
ExtractTextContentState,
build_extract_text_content_spec,
build_extract_text_content_workflow,
extract_text_content,
)
from .schemas import StateFieldMetadata, state_field
from .subgraph import async_subgraph_node, subgraph_node, subgraph_ref
@@ -89,6 +97,9 @@ __all__ = [
"ConstantInput",
"CountOutput",
"ExtractFieldInput",
"ExtractTextContentInput",
"ExtractTextContentOutput",
"ExtractTextContentState",
"FilterItemsInput",
"FilterItemsPresentInput",
"GraphPath",
@@ -117,12 +128,15 @@ __all__ = [
"bind_fields",
"build_async_registry",
"build_registry",
"build_extract_text_content_spec",
"build_extract_text_content_workflow",
"bind_state",
"coalesce",
"concat",
"constant",
"default_if_none",
"extract_field",
"extract_text_content",
"filter_items",
"filter_items_present",
"merge_maps",
src/wf_authoring/recipes.py
+104
View File
@@ -0,0 +1,104 @@
from __future__ import annotations
from typing import Any
from pydantic import BaseModel
from wf_authoring.builder import WorkflowBuilder
from wf_authoring.dsl import input_from, input_path, input_value, output_to, state_path
from wf_authoring.nodes import NodeSpec, build_registry
from wf_authoring.ops import concat, extract_field, filter_items
from wf_authoring.subgraph import subgraph_node
from wf_core import END
class ExtractTextContentInput(BaseModel):
"""Input for extracting text from generic content-block dictionaries."""
content: list[dict[str, Any]]
separator: str = ""
class ExtractTextContentState(BaseModel):
"""Internal recipe state for staged content extraction."""
text_items: list[dict[str, Any]] = []
texts: list[str] = []
text: str = ""
class ExtractTextContentOutput(BaseModel):
"""Output text joined from all text content blocks."""
text: str
def build_extract_text_content_workflow():
"""Build the first-party text-content extraction recipe workflow.
Current limitation: this recipe is exposed as a wrapper-node capability via
`subgraph_node`, not as a native `SubgraphNode`. Callers can use it like any
other NodeSpec, but parent traces see one node call and child frames,
interrupts, and step-level diagnostics are not promoted to the parent run.
"""
builder = WorkflowBuilder(
name="extract_text_content",
input_schema=ExtractTextContentInput,
state_schema=ExtractTextContentState,
output_schema=ExtractTextContentOutput,
)
filter_text = builder.use(
filter_items,
id="filter_text",
input=[
input_from(input_path("content"), "items"),
input_value("key", "type"),
input_value("value", "text"),
],
output=[output_to("items", state_path("text_items"))],
)
extract_text = builder.use(
extract_field,
id="extract_text",
input=[
input_from(state_path("text_items"), "items"),
input_value("field", "text"),
],
output=[output_to("values", state_path("texts"))],
)
join_text = builder.use(
concat,
id="join_text",
input=[
input_from(state_path("texts"), "items"),
input_from(input_path("separator"), "separator"),
],
output=[output_to("text", state_path("text"))],
)
builder.set_entry_point(filter_text)
builder.connect(filter_text, "ok", extract_text)
builder.connect(extract_text, "ok", join_text)
builder.connect(join_text, "ok", END)
return builder.compile()
def build_extract_text_content_spec() -> NodeSpec[
ExtractTextContentInput, ExtractTextContentOutput
]:
"""Return the recipe as a normal NodeSpec for first-party capability sources."""
workflow = build_extract_text_content_workflow()
return subgraph_node(
name="authoring.extract_text_content",
workflow=workflow,
registry=build_registry(filter_items, extract_field, concat),
input_model=ExtractTextContentInput,
output_model=ExtractTextContentOutput,
description=(
"Extract text fields from content blocks with type='text' and join "
"them using separator."
),
)
extract_text_content = build_extract_text_content_spec()
"""First-party recipe composed from generic sequence/value ops."""
src/wf_mcp/broker/service/builtins.py
+42 -6
View File
@@ -7,6 +7,7 @@ from wf_authoring import extract_field, filter_items, filter_items_present, firs
from wf_authoring import first_item_maybe, first_item_or_none, is_empty, last_item
from wf_authoring import last_item_or_none, length, node, pick_key, pick_path
from wf_authoring import project_fields, rename_fields, runtime_error, truthy
from wf_authoring import extract_text_content
from wf_core.runtime.ops.merges import DEFAULT_REDUCER_DEFINITIONS
from wf_platform import (
@@ -26,6 +27,9 @@ BUILTIN_CONNECTION_ID = "wf.std"
MCP_SOURCE_ID = "wf.mcp"
"""Reserved source id for future workflow-safe MCP utility node specs."""
RECIPE_SOURCE_ID = "wf.recipes"
"""Internal source id for first-party composed workflow recipes."""
AUTHORING_STD_SPECS: tuple[NodeSpec[Any, Any], ...] = (
coalesce,
@@ -52,16 +56,36 @@ AUTHORING_STD_SPECS: tuple[NodeSpec[Any, Any], ...] = (
"""Existing authoring ops that are also exposed through the workflow stdlib."""
def builtin_specs() -> dict[str, NodeSpec[Any, Any]]:
"""Return built-in NodeSpecs available to raw broker workflow plans."""
specs = [
node(spec, name=spec.name.removeprefix("authoring."))
for spec in AUTHORING_STD_SPECS
RECIPE_SPECS: tuple[NodeSpec[Any, Any], ...] = (extract_text_content,)
"""Composed first-party recipes exposed as capabilities."""
def _qualified_specs(
source_id: str,
specs: tuple[NodeSpec[Any, Any], ...],
) -> dict[str, NodeSpec[Any, Any]]:
"""Return specs with authoring names rewritten under one source id."""
local_specs = [
node(spec, name=spec.name.removeprefix("authoring.")) for spec in specs
]
qualified_specs = [qualify_spec(BUILTIN_CONNECTION_ID, spec) for spec in specs]
qualified_specs = [qualify_spec(source_id, spec) for spec in local_specs]
return {spec.name: spec for spec in qualified_specs}
def builtin_specs() -> dict[str, NodeSpec[Any, Any]]:
"""Return primitive built-in NodeSpecs available to raw broker workflow plans."""
return _qualified_specs(BUILTIN_CONNECTION_ID, AUTHORING_STD_SPECS)
def recipe_specs() -> dict[str, NodeSpec[Any, Any]]:
"""Return composed first-party recipe specs.
Recipes are wrapper-node subgraphs today. They are useful workflow-facing
capabilities, but parent runs do not yet see their child graph frames.
"""
return _qualified_specs(RECIPE_SOURCE_ID, RECIPE_SPECS)
def builtin_reducers() -> dict[str, ReducerSpec]:
"""Return built-in reducers owned by the workflow standard library."""
return {
@@ -94,4 +118,16 @@ def builtin_sources() -> dict[str, CapabilitySource]:
permissions=SourcePermissions(safe_for_workflow=True),
description="Workflow standard-library nodes.",
),
RECIPE_SOURCE_ID: CapabilitySource(
id=RECIPE_SOURCE_ID,
kind="system",
capabilities=CapabilityBuckets(node_specs=recipe_specs()),
visibility=SourceVisibility(
planner=True,
mcp_client=True,
admin_dashboard=True,
),
permissions=SourcePermissions(safe_for_workflow=True),
description="First-party workflow recipes composed from standard nodes.",
),
}
tests/authoring/test_ops.py
+20
View File
@@ -10,6 +10,7 @@ from wf_authoring import (
default_if_none,
concat,
extract_field,
extract_text_content,
filter_items,
filter_items_present,
first_item,
@@ -381,6 +382,25 @@ def test_concat_joins_strings_with_separator() -> None:
assert result["output"]["text"] == "a\nb\nc"
def test_extract_text_content_recipe_filters_extracts_and_joins_text_blocks() -> None:
registry = build_registry(extract_text_content)
result = registry["authoring.extract_text_content"](
{
"content": [
{"type": "text", "text": "hello"},
{"type": "image", "url": "img://1"},
{"type": "text", "text": "world"},
],
"separator": " ",
},
RuntimeContext(current_node_id="extract_text_content"),
)
assert result["outcome"] == "ok"
assert result["output"]["text"] == "hello world"
def test_truthy_routes_truthy_and_falsey_outcomes() -> None:
registry = build_registry(truthy)
ctx = RuntimeContext(current_node_id="truthy")
tests/wf_mcp/test_service.py
+11
View File
@@ -273,6 +273,16 @@ def test_service_sources_have_visibility_and_capability_buckets() -> None:
assert not std_source.capabilities.tools
def test_wf_recipes_source_contains_composed_capabilities() -> None:
service = WfMcpService(store=FileStore(local_temp_root() / "recipes_source_store"))
specs = service.capability_sources["wf.recipes"].capabilities.node_specs
assert set(specs) == {"wf.recipes.extract_text_content"}
assert (
service.capability_sources["wf.recipes"].permissions.safe_for_workflow is True
)
def test_wf_admin_source_exists_but_is_not_planner_visible() -> None:
service = WfMcpService(store=FileStore(local_temp_root() / "admin_source_store"))
source = service.capability_sources["wf.admin"]
@@ -298,6 +308,7 @@ def test_service_can_disable_builtin_stdlib_specs() -> None:
)
assert "wf.std" not in service.capability_sources
assert "wf.recipes" not in service.capability_sources
def test_service_planner_catalog_excludes_hidden_sources() -> None: