cleveragents/cleveragents-core

Fork 3

UAT: `Node._execute_subgraph()` is a non-functional placeholder — subgraph nodes never invoke nested actor graphs #4815

New issue

Open

opened 2026-04-08 19:37:45 +00:00 by HAL9000 · 1 comment

HAL9000 commented

2026-04-08 19:37:45 +00:00

Owner

Bug Report

Feature Area: Actor System — LangGraph graph composition, subgraph nodes
Severity: High — subgraph node type is completely non-functional; nested actor composition does not work
Found by: UAT tester instance uat-tester-actor-system
Spec reference: docs/specification.md §Nodes in the Graph: Actors and Tools (lines 20595–20644), §Actor Composition (lines 20384–20403)

What Was Tested

Code-level analysis of src/cleveragents/langgraph/nodes.py — Node._execute_subgraph() (lines 307–309) against the spec's subgraph node requirements.

Expected Behavior (from spec)

The spec (lines 20595–20607) defines subgraph nodes as:

Graph nodes can be any of:

an actor (another LLM agent or composite workflow, referenced by name),

a tool node — a deterministic, non-LLM step that directly invokes a tool.

The spec (line 20559) defines the subgraph node type config as:

subgraph | actor_path | Embeds another actor as a nested workflow

The spec (lines 20384–20403) describes hierarchical actor composition:

Actors can reference other actors by name... This enables hierarchical composition where:

Actor A's graph can include nodes that call Actor B (by name)

Actor B itself is a graph that might call Actor C

And so on (no depth limit)

The spec (line 20123) states:

Every custom actor IS a graph (a LangGraph defined via YAML configuration). Even a simple actor wrapping a single LLM is technically a graph with one node.

Actual Behavior (from code)

Node._execute_subgraph() at src/cleveragents/langgraph/nodes.py:307–309:

async def _execute_subgraph(self, state: GraphState) -> dict[str, Any]:
    # Placeholder: could invoke nested LangGraph by name
    return {"subgraph_invoked": self.config.subgraph or ""}

This is a stub that does nothing. It:

Does NOT look up the referenced actor by name from any registry
Does NOT instantiate or compile the referenced actor's graph
Does NOT execute the nested graph with the current state
Does NOT propagate state updates from the nested graph back to the parent
Returns a trivial {"subgraph_invoked": "actor_name"} dict that has no effect on the workflow

The comment # Placeholder: could invoke nested LangGraph by name confirms this is intentionally incomplete.

Impact

Any actor YAML that uses type: subgraph nodes (e.g., the scientific paper writer example in the spec at line 20948) will silently produce incorrect results — the subgraph node will appear to "succeed" but will not actually execute the nested actor. This is a silent failure with no error or warning.

Steps to Reproduce

from cleveragents.actor.schema import ActorConfigSchema, ActorType, NodeType
from cleveragents.actor.compiler import compile_actor
from cleveragents.langgraph.nodes import Node, NodeConfig

# Create a node with SUBGRAPH type
config = NodeConfig(
    name="nested_actor",
    type=NodeType.SUBGRAPH,
    subgraph="local/my-nested-actor",
)
node = Node(config)

# Execute the subgraph node
import asyncio
from cleveragents.langgraph.state import GraphState
result = asyncio.run(node.execute(GraphState()))

# Result is {"subgraph_invoked": "local/my-nested-actor", "current_node": "nested_actor"}
# The nested actor was NEVER actually invoked
print(result)

Code Location

src/cleveragents/langgraph/nodes.py:307–309 — Node._execute_subgraph() (placeholder)
src/cleveragents/langgraph/graph.py — LangGraph class (no subgraph resolution logic)
src/cleveragents/actor/compiler.py:292–295 — compile_actor() correctly records subgraph_refs but the runtime never uses them

Expected Fix

_execute_subgraph() should:

Look up the referenced actor by name from the actor registry (or a resolver passed at graph construction time)
Compile the referenced actor's ActorConfigSchema into a LangGraph instance
Execute the nested graph with the current state (passing relevant state fields)
Merge the nested graph's output state back into the parent graph's state
Handle errors from the nested graph appropriately (propagate or wrap)

The compile_actor() function already records subgraph_refs in CompilationMetadata — this metadata should be used at runtime to resolve and execute nested graphs.

Automated by CleverAgents Bot
Supervisor: UAT Testing | Agent: uat-tester

## Bug Report **Feature Area:** Actor System — LangGraph graph composition, subgraph nodes **Severity:** High — subgraph node type is completely non-functional; nested actor composition does not work **Found by:** UAT tester instance `uat-tester-actor-system` **Spec reference:** `docs/specification.md` §Nodes in the Graph: Actors and Tools (lines 20595–20644), §Actor Composition (lines 20384–20403) --- ### What Was Tested Code-level analysis of `src/cleveragents/langgraph/nodes.py` — `Node._execute_subgraph()` (lines 307–309) against the spec's subgraph node requirements. ### Expected Behavior (from spec) The spec (lines 20595–20607) defines subgraph nodes as: > Graph nodes can be any of: > * an **actor** (another LLM agent or composite workflow, referenced by name), > * a **tool node** — a deterministic, non-LLM step that directly invokes a tool. The spec (line 20559) defines the `subgraph` node type config as: > `subgraph` | `actor_path` | Embeds another actor as a nested workflow The spec (lines 20384–20403) describes hierarchical actor composition: > Actors can reference other actors **by name**... This enables hierarchical composition where: > * Actor A's graph can include nodes that call Actor B (by name) > * Actor B itself is a graph that might call Actor C > * And so on (no depth limit) The spec (line 20123) states: > **Every custom actor IS a graph** (a LangGraph defined via YAML configuration). Even a simple actor wrapping a single LLM is technically a graph with one node. ### Actual Behavior (from code) `Node._execute_subgraph()` at `src/cleveragents/langgraph/nodes.py:307–309`: ```python async def _execute_subgraph(self, state: GraphState) -> dict[str, Any]: # Placeholder: could invoke nested LangGraph by name return {"subgraph_invoked": self.config.subgraph or ""} ``` This is a **stub that does nothing**. It: 1. Does NOT look up the referenced actor by name from any registry 2. Does NOT instantiate or compile the referenced actor's graph 3. Does NOT execute the nested graph with the current state 4. Does NOT propagate state updates from the nested graph back to the parent 5. Returns a trivial `{"subgraph_invoked": "actor_name"}` dict that has no effect on the workflow The comment `# Placeholder: could invoke nested LangGraph by name` confirms this is intentionally incomplete. ### Impact Any actor YAML that uses `type: subgraph` nodes (e.g., the scientific paper writer example in the spec at line 20948) will silently produce incorrect results — the subgraph node will appear to "succeed" but will not actually execute the nested actor. This is a silent failure with no error or warning. ### Steps to Reproduce ```python from cleveragents.actor.schema import ActorConfigSchema, ActorType, NodeType from cleveragents.actor.compiler import compile_actor from cleveragents.langgraph.nodes import Node, NodeConfig # Create a node with SUBGRAPH type config = NodeConfig( name="nested_actor", type=NodeType.SUBGRAPH, subgraph="local/my-nested-actor", ) node = Node(config) # Execute the subgraph node import asyncio from cleveragents.langgraph.state import GraphState result = asyncio.run(node.execute(GraphState())) # Result is {"subgraph_invoked": "local/my-nested-actor", "current_node": "nested_actor"} # The nested actor was NEVER actually invoked print(result) ``` ### Code Location - `src/cleveragents/langgraph/nodes.py:307–309` — `Node._execute_subgraph()` (placeholder) - `src/cleveragents/langgraph/graph.py` — `LangGraph` class (no subgraph resolution logic) - `src/cleveragents/actor/compiler.py:292–295` — `compile_actor()` correctly records `subgraph_refs` but the runtime never uses them ### Expected Fix `_execute_subgraph()` should: 1. Look up the referenced actor by name from the actor registry (or a resolver passed at graph construction time) 2. Compile the referenced actor's `ActorConfigSchema` into a `LangGraph` instance 3. Execute the nested graph with the current state (passing relevant state fields) 4. Merge the nested graph's output state back into the parent graph's state 5. Handle errors from the nested graph appropriately (propagate or wrap) The `compile_actor()` function already records `subgraph_refs` in `CompilationMetadata` — this metadata should be used at runtime to resolve and execute nested graphs. --- **Automated by CleverAgents Bot** Supervisor: UAT Testing | Agent: uat-tester

HAL9000 referenced this issue

2026-04-08 19:46:08 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 19:59:07 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 20:12:39 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 20:16:04 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 20:22:25 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 added the

labels

2026-04-08 23:00:07 +00:00

HAL9000 commented

2026-04-09 03:03:01 +00:00

Author

Owner

Issue triaged by project owner:

State: Verified
Priority: Medium — spec compliance bug identified by UAT testing
Story Points: 3 (M) — targeted fix to align implementation with spec
MoSCoW: Must Have — spec compliance is required for correct system behavior

Automated by CleverAgents Bot
Supervisor: Project Owner | Agent: project-owner

Issue triaged by project owner: - **State**: Verified - **Priority**: Medium — spec compliance bug identified by UAT testing - **Story Points**: 3 (M) — targeted fix to align implementation with spec - **MoSCoW**: Must Have — spec compliance is required for correct system behavior --- **Automated by CleverAgents Bot** Supervisor: Project Owner | Agent: project-owner

HAL9000 added

and removed

labels

2026-04-09 03:03:03 +00:00

HAL9000 added this to the v3.5.0 milestone