Files
dify/api/services/dify_builder/agent/graph_translate.py
T
Charles YaoandClaude Opus 5 f0c7b7c413 feat(dify-builder): carry nesting and layout through translation
to_intents read only id and data, so every node-level ReactFlow key --
parentId, extent, zIndex, position -- was dropped on each Builder build.
Iteration and loop children were written un-nested, not inside their
container at all, and the generator's computed layout was replaced by
default positions. The data-level markers (isInIteration, iteration_id)
survived via config, which is why the graphs looked almost right.

apply_create_node already took position; parent_id is new and implies
extent="parent" + zIndex=1002, matching what the generator computes.

Unfiled; found while tracing ESQ1-288. Skipping the synthetic start node
stops the canvas crash, but an iteration whose children are not inside it
still does not work.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-09-20 22:09:26 -07:00

79 lines
3.4 KiB
Python

"""Translate a WorkflowGenerator graph into Dify Builder MutationIntents.
The generator emits ReactFlow nodes ({id, type:"custom", data:{type:<real>,...}})
and edges. We emit one create_node per node (config = node.data minus the
redundant "type" key -- apply_create_node re-sets it) followed by one connect
per edge, so every connect endpoint already exists at apply time. Node ids
are preserved verbatim -- including the generator's start node id -- so
intra-node references (variable selectors, prompt {{#id.var#}}) stay valid;
reconciling the generator's start node with the draft's existing start is the
handler's job, not this translation step."""
from typing import Any
from core.dify_builder.models import MutationIntent
# Container-start markers the frontend synthesizes for iteration/loop bodies.
# The generator emits them correctly typed, but a Builder create_node re-creates
# them as generic `custom` nodes with no matching component, which crashes the
# canvas on open and never recovers (ESQ1-288). The canvas rebuilds them from
# the container itself, so dropping them here loses nothing.
_SYNTHETIC_NODE_TYPES = frozenset({"iteration-start", "loop-start"})
# The frontend regenerates container start markers at render time and
# rewrites start_node_id to the marker it created (workflow-init.ts:124-141,
# :177-181). Persisting our own pointer leaves it dangling once the marker
# node is skipped, and workflow-init.ts:147 dereferences it without a guard.
_FRONTEND_OWNED_KEYS = frozenset({"start_node_id"})
def to_intents(graph: dict[str, Any]) -> list[MutationIntent]:
nodes = graph.get("nodes") or []
edges = graph.get("edges") or []
creates: list[MutationIntent] = []
skipped_node_ids: set[str] = set()
for node in nodes:
data = node.get("data") or {}
node_type = data.get("type")
if node_type in _SYNTHETIC_NODE_TYPES:
skipped_node_ids.add(node.get("id"))
continue
node_id = node.get("id")
if not node_type or not node_id:
continue
config = {k: v for k, v in data.items() if k != "type" and k not in _FRONTEND_OWNED_KEYS}
args: dict[str, Any] = {
"node_type": str(node_type),
"config": config,
"node_id": str(node_id),
}
# The ReactFlow wrapper carries nesting and layout OUTSIDE `data`, so
# reading only id+data dropped them on every build: children landed
# un-nested and the generator's layout was replaced by defaults.
parent_id = node.get("parentId")
if parent_id:
args["parent_id"] = str(parent_id)
position = node.get("position")
if isinstance(position, dict):
args["position"] = dict(position)
creates.append(MutationIntent(op="create_node", args=args))
connects: list[MutationIntent] = []
for edge in edges:
src, tgt = edge.get("source"), edge.get("target")
if not src or not tgt:
continue
if src in skipped_node_ids or tgt in skipped_node_ids:
continue
args: dict[str, Any] = {"from_node": str(src), "to_node": str(tgt)}
source_handle = edge.get("sourceHandle")
if source_handle:
args["source_handle"] = source_handle
target_handle = edge.get("targetHandle")
if target_handle:
args["target_handle"] = target_handle
connects.append(MutationIntent(op="connect", args=args))
return creates + connects