Self-improving AI Employees for everyone
Python ai-agentsai-employeeai-employeesassistantautonomous-agentsco-employeescompanygrok-botinstinctloop-agentsmuseown-your-datapersonalself-hosted
refactor: remove dead code across execution, llm, and event layers
trohitgdbd8a2e27880
diff --git a/client/package.json b/client/package.jsonindex d8d4e703..99b1d1aa 100644--- a/client/package.json+++ b/client/package.json@@ -47,7 +47,6 @@"globals": "^16.5.0","idb-keyval": "^6.2.2","lucide-react": "^1.14.0",- "next-themes": "^0.4.6","partysocket": "1.1.18","postcss": "^8.5.14","prismjs": "^1.30.0",diff --git a/pnpm-lock.yaml b/pnpm-lock.yamlindex 407023f2..0fb1ed66 100644--- a/pnpm-lock.yaml+++ b/pnpm-lock.yaml@@ -137,9 +137,6 @@ importers:lucide-react:specifier: ^1.14.0version: 1.14.0([email protected])- next-themes:- specifier: ^0.4.6partysocket:specifier: 1.1.18version: 1.1.18([email protected])@@ -4685,12 +4682,6 @@ packages:resolution: {integrity: sha512-8Ofs/AUQh8MaEcrlq5xOX0CQ9ypTF5dl78mjlMNfOK08fzpgTHQRQPBxcPlEtIw0yRpws+Zo/3r+5WRby7u3Gg==}engines: {node: '>= 0.6'}- resolution: {integrity: sha512-pZvgD5L0IEvX5/9GWyHMf3m8BKiVQwsCMHfoFosXtXBMnaS0ZnIJ9ST4b4NqLVKDEm8QBxoNNGNaBv2JNF6XNA==}- peerDependencies:- react: ^16.8 || ^17 || ^18 || ^19 || ^19.0.0-rc- react-dom: ^16.8 || ^17 || ^18 || ^19 || ^19.0.0-rc-resolution: {integrity: sha512-/jKZoMpw0F8GRwl4/eLROPA3cfcXtLApP0QzLmUT/HuPCZWyB7IY9ZrMeKw2O/nFIqPQB3PVM9aYm0F312AXDQ==}engines: {node: '>=10.5.0'}@@ -11535,11 +11526,6 @@ snapshots:- dependencies:- react: 19.2.6- react-dom: 19.2.6([email protected])-diff --git a/server/constants.py b/server/constants.pyindex 9d3fc7c1..d25f8faf 100644--- a/server/constants.py+++ b/server/constants.py@@ -77,11 +77,6 @@ SKILL_NODE_TYPES: FrozenSet[str] = frozenset(])-# Toolkit node types (aggregate sub-nodes, n8n Sub-Node pattern)-# Sub-nodes connected to toolkits should not execute independently --# they only execute when called via the toolkit's tool interface-TOOLKIT_NODE_TYPES: FrozenSet[str] = frozenset()-# All AI-related node types (for API key injection)AI_MODEL_TYPES: FrozenSet[str] = AI_AGENT_TYPES | AI_CHAT_MODEL_TYPESdiff --git a/server/services/ai.py b/server/services/ai.pyindex 7af5c624..274ce549 100644--- a/server/services/ai.py+++ b/server/services/ai.py@@ -134,11 +134,9 @@ logger = get_logger(__name__)# ---------------------------------------------------------------------------# Native LLM provider imports (dual-path: native for chat, LangChain for agents)# ----------------------------------------------------------------------------# ``create_provider`` / ``is_native_provider`` are no longer called from-# this module — ``ChatUnifier`` (injected via DI) owns provider-# dispatch. The imports stay deleted; Phase D removes the legacy-# factory functions entirely once the LangChain fallback for-# groq/cerebras lands behind the unifier.+# ``ChatUnifier`` (injected via DI) owns provider dispatch. The legacy+# ``services.llm.factory`` module (``create_provider`` /+# ``is_native_provider``) was removed after the unifier cutover.from services.llm.protocol import (Message as NativeMessage,ThinkingConfig as NativeThinkingConfig,diff --git a/server/services/deployment/manager.py b/server/services/deployment/manager.pyindex 7b3e53e7..0df370fd 100644--- a/server/services/deployment/manager.py+++ b/server/services/deployment/manager.py@@ -14,7 +14,7 @@ from datetime import datetimefrom typing import Dict, Any, List, Optional, Callable, TYPE_CHECKINGfrom core.logging import get_logger-from constants import POLLING_TRIGGER_TYPES, TOOLKIT_NODE_TYPES, WORKFLOW_TRIGGER_TYPES+from constants import POLLING_TRIGGER_TYPES, WORKFLOW_TRIGGER_TYPESfrom services import event_waiterfrom services.events.envelope import equivalent_event_typesfrom .state import DeploymentState, TriggerInfo@@ -1201,20 +1201,6 @@ class DeploymentManager:continuedownstream_ids.add(source)- # Include sub-nodes connected to toolkit nodes (n8n Sub-Node pattern).- # Service nodes connect to a toolkit's input-main (not a config- # handle) and need to be included so the toolkit can discover- # them. ``TOOLKIT_NODE_TYPES`` is the canonical set; today only- # toolkit node types are in it.- toolkit_node_ids = {n["id"] for n in nodes if n.get("type") in TOOLKIT_NODE_TYPES and n["id"] in downstream_ids}- for edge in edges:- target = edge.get("target")- source = edge.get("source")- # Include nodes that connect to toolkit nodes- if target in toolkit_node_ids and source not in downstream_ids:- downstream_ids.add(source)- logger.debug(f"[Deployment] Including sub-node {source} connected to toolkit {target}")-# Include tool nodes connected to AI Agent nodes (for capability discovery)# When a child agent is included, we need its connected tools so the parent# can discover what capabilities the child hasdiff --git a/server/services/event_waiter.py b/server/services/event_waiter.pyindex 931050d2..c03e2fa5 100644--- a/server/services/event_waiter.py+++ b/server/services/event_waiter.py@@ -363,19 +363,14 @@ async def run_trigger_precheck(node_type: str, parameters: Dict) -> Any:@dataclassclass Waiter:- """Single event waiter.-- In memory mode: uses asyncio.Future- In Redis mode: uses stream polling with stored metadata- """+ """Single event waiter, resolved via asyncio.Future."""id: str = field(default_factory=lambda: str(uuid.uuid4()))node_id: str = ""node_type: str = ""event_type: str = ""- params: Dict = field(default_factory=dict) # Store params for Redis mode filter rebuildfilter_fn: Callable[[Dict], bool] = field(default_factory=lambda: lambda x: True)- future: Optional[asyncio.Future] = None # Only used in memory mode+ future: Optional[asyncio.Future] = Nonecancelled: bool = Falsecreated_at: float = field(default_factory=time.time)@@ -409,7 +404,6 @@ async def register(node_type: str, node_id: str, params: Dict) -> Waiter:node_id=node_id,node_type=node_type,event_type=config.event_type,- params=params,filter_fn=build_filter(node_type, params),)@@ -605,5 +599,8 @@ def clear_all() -> int:def get_backend_mode() -> str:- """Get current backend mode for debugging."""+ """Get current backend mode for debugging.++ Always "memory" — the Redis-Streams backend was retired in Wave 15.3.+ """return "memory"diff --git a/server/services/execution/executor.py b/server/services/execution/executor.pyindex bcd3e4a5..f135f515 100644--- a/server/services/execution/executor.py+++ b/server/services/execution/executor.py@@ -455,151 +455,6 @@ class WorkflowExecutor:ctx.status = WorkflowStatus.FAILED- # =========================================================================- # PARALLEL EXECUTION (Legacy - Fork/Join with FIRST_COMPLETED pattern)- # =========================================================================-- async def _execute_parallel_nodes(self, ctx: ExecutionContext, nodes: List[NodeExecution], enable_caching: bool) -> None:- """Execute multiple nodes in parallel using asyncio.wait with FIRST_COMPLETED.-- Uses the standard asyncio pattern for mixed task types:- - Regular nodes complete quickly- - Trigger nodes wait indefinitely for external events- - If a regular node fails, cancel remaining trigger nodes immediately-- This follows Python asyncio best practices:- https://docs.python.org/3/library/asyncio-task.html#asyncio.wait-- Args:- ctx: ExecutionContext- nodes: List of NodeExecution to run in parallel- enable_caching: Whether to use result caching- """- # Mark all as scheduled- for node in nodes:- node.status = TaskStatus.SCHEDULED- await self._notify_status(node.node_id, "scheduled", {})-- # Create named tasks for parallel execution- # Using dict to track node -> task mapping for proper result handling- node_to_task: Dict[str, asyncio.Task] = {}- task_to_node: Dict[asyncio.Task, NodeExecution] = {}-- for node in nodes:- task = asyncio.create_task(self._execute_node_with_retry(ctx, node, enable_caching), name=f"node_{node.node_id}")- node_to_task[node.node_id] = task- task_to_node[task] = node-- pending: Set[asyncio.Task] = set(node_to_task.values())- workflow_failed = False-- # Process tasks as they complete using FIRST_COMPLETED pattern- while pending:- # Wait for any task to complete- done, pending = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)-- # Process completed tasks- for task in done:- node = task_to_node[task]-- try:- result = task.result()-- if isinstance(result, Exception):- # Task raised exception- node.status = TaskStatus.FAILED- node.error = str(result)- node.completed_at = time.time()- ctx.errors.append(- {- "node_id": node.node_id,- "error": str(result),- "timestamp": time.time(),- }- )- await self._notify_status(node.node_id, "error", {"error": str(result)})- logger.error("Parallel node failed", node_id=node.node_id, error=str(result))- workflow_failed = True-- elif result.get("retries_exhausted"):- # Node failed after all retries - already in DLQ- node.status = TaskStatus.FAILED- node.error = result.get("error", "Unknown error")- node.completed_at = time.time()- ctx.errors.append(- {- "node_id": node.node_id,- "error": node.error,- "retries_exhausted": True,- "timestamp": time.time(),- }- )- workflow_failed = True-- elif not result.get("success"):- # Node returned failure without exhausting retries- node.status = TaskStatus.FAILED- node.error = result.get("error", "Unknown error")- node.completed_at = time.time()- ctx.errors.append(- {- "node_id": node.node_id,- "error": node.error,- "timestamp": time.time(),- }- )- await self._notify_status(node.node_id, "error", {"error": node.error})- logger.error("Parallel node failed", node_id=node.node_id, error=node.error)- workflow_failed = True-- except asyncio.CancelledError:- # Task was cancelled (by us or externally)- node.status = TaskStatus.CANCELLED- node.completed_at = time.time()- logger.info("Parallel node cancelled", node_id=node.node_id)-- except Exception as e:- # Unexpected exception from task.result()- node.status = TaskStatus.FAILED- node.error = str(e)- node.completed_at = time.time()- ctx.errors.append(- {- "node_id": node.node_id,- "error": str(e),- "timestamp": time.time(),- }- )- await self._notify_status(node.node_id, "error", {"error": str(e)})- logger.error("Parallel node exception", node_id=node.node_id, error=str(e))- workflow_failed = True-- # If workflow failed, cancel remaining pending tasks- # This prevents trigger nodes from blocking forever when a regular node fails- if workflow_failed and pending:- logger.info("Workflow failed, cancelling remaining tasks", pending_count=len(pending))-- for task in pending:- task.cancel()-- # Wait for cancelled tasks to finish- if pending:- cancelled_done, _ = await asyncio.wait(pending, return_when=asyncio.ALL_COMPLETED)-- # Mark cancelled nodes- for task in cancelled_done:- node = task_to_node[task]- if node.status not in (TaskStatus.COMPLETED, TaskStatus.FAILED):- node.status = TaskStatus.CANCELLED- node.completed_at = time.time()- logger.info("Cancelled pending node", node_id=node.node_id)-- pending = set() # All done now-- # Mark workflow as failed if any node failed- if workflow_failed:- ctx.status = WorkflowStatus.FAILED-async def _execute_single_node(self, ctx: ExecutionContext, node: NodeExecution, enable_caching: bool) -> None:"""Execute a single node with retry logic.@@ -879,7 +734,7 @@ class WorkflowExecutor:workflow. They listen for specific events/conditions and initiatethe execution of the entire workflow.- Config nodes and toolkit sub-nodes are excluded from layers since+ Config nodes and AI-agent sub-nodes are excluded from layers sincethey don't execute as independent workflow nodes.Args:@@ -889,14 +744,11 @@ class WorkflowExecutor:Returns:List of layers, where each layer is a list of node IDs"""- from constants import CONFIG_NODE_TYPES, TOOLKIT_NODE_TYPES, AI_AGENT_TYPES+ from constants import CONFIG_NODE_TYPES, AI_AGENT_TYPES# Build node type lookup for trigger detectionnode_types: Dict[str, str] = {node["id"]: node.get("type", "unknown") for node in nodes}- # Find toolkit sub-nodes (nodes that connect TO a toolkit)- toolkit_node_ids = {n.get("id") for n in nodes if n.get("type") in TOOLKIT_NODE_TYPES}-# Find AI Agent nodes (all agent types have config handles)ai_agent_node_ids = {n.get("id") for n in nodes if n.get("type") in AI_AGENT_TYPES}@@ -906,10 +758,6 @@ class WorkflowExecutor:target = edge.get("target")target_handle = edge.get("targetHandle")- # Any node that connects TO a toolkit is a sub-node- if target in toolkit_node_ids and source:- subnode_ids.add(source)-# Nodes connected to AI Agent config handles are sub-nodes# These handles: input-memory, input-tools, input-skill, input-teammatesif target in ai_agent_node_ids and source and target_handle:diff --git a/server/services/execution/models.py b/server/services/execution/models.pyindex c8f0fa58..52139beb 100644--- a/server/services/execution/models.py+++ b/server/services/execution/models.py@@ -315,10 +315,10 @@ class ExecutionContext:Config nodes (memory, tools, model configs) are excluded from executionas they provide configuration to other nodes via special handles.- Toolkit sub-nodes (nodes connected to an aggregator) are also- excluded - they execute only when called via the toolkit's tool interface.+ Sub-nodes connected to an AI Agent's config handles are also+ excluded - they execute only when called via the agent's tool interface."""- from constants import CONFIG_NODE_TYPES, TOOLKIT_NODE_TYPES, AI_AGENT_TYPES+ from constants import CONFIG_NODE_TYPES, AI_AGENT_TYPESexecution_id = str(uuid.uuid4())ctx = cls(@@ -329,10 +329,6 @@ class ExecutionContext:edges=edges or [],)- # Find toolkit sub-nodes (nodes that connect TO a toolkit node)- # These should only execute when called via the toolkit, not as workflow nodes- toolkit_node_ids = {n.get("id") for n in (nodes or []) if n.get("type") in TOOLKIT_NODE_TYPES}-# Find AI Agent nodes (all agent types have config handles)ai_agent_node_ids = {n.get("id") for n in (nodes or []) if n.get("type") in AI_AGENT_TYPES}@@ -342,10 +338,6 @@ class ExecutionContext:target = edge.get("target")target_handle = edge.get("targetHandle")- # Any node that connects TO a toolkit is a sub-node- if target in toolkit_node_ids and source:- subnode_ids.add(source)-# Nodes connected to AI Agent config handles are sub-nodes# These handles: input-memory, input-tools, input-skill, input-teammatesif target in ai_agent_node_ids and source and target_handle:diff --git a/server/services/execution/recovery.py b/server/services/execution/recovery.pyindex d679edd9..0d8685a7 100644--- a/server/services/execution/recovery.py+++ b/server/services/execution/recovery.py@@ -194,8 +194,6 @@ class RecoverySweeper:return needs_recovery- return cleaned-# Global sweeper instance (initialized by main.py)_sweeper: Optional[RecoverySweeper] = Nonediff --git a/server/services/llm/__init__.py b/server/services/llm/__init__.pyindex e2331f47..6bcb6a59 100644--- a/server/services/llm/__init__.py+++ b/server/services/llm/__init__.py@@ -1,4 +1,4 @@-"""LLM service layer -- native provider SDKs, config, protocol types, factory."""+"""LLM service layer -- native provider SDKs, config, protocol types."""from services.llm.config import (ProviderConfig,@@ -21,7 +21,6 @@ from services.llm.protocol import (LLMProvider,)from services.llm.messages import is_valid_message_content, filter_empty_messages-from services.llm.factory import create_provider, is_native_provider, NATIVE_PROVIDERSfrom services.llm.registry import (ProviderSpec,register_provider,@@ -59,10 +58,6 @@ __all__ = [# Messages"is_valid_message_content","filter_empty_messages",- # Factory (legacy — replaced by unifier in Phase A3; kept for back-compat)- "create_provider",- "is_native_provider",- "NATIVE_PROVIDERS",# Plugin registry + unifier (Phase A — single SERVICE facade)"ProviderSpec","register_provider",diff --git a/server/services/llm/factory.py b/server/services/llm/factory.pydeleted file mode 100644index 4440db8c..00000000--- a/server/services/llm/factory.py+++ /dev/null@@ -1,75 +0,0 @@-"""Provider factory -- lazy-imports native SDK providers."""--from __future__ import annotations--from typing import Optional--from core.logging import get_logger-from services.llm.protocol import LLMProvider--logger = get_logger(__name__)--# Providers with dedicated SDK implementations-_DEDICATED_PROVIDERS = frozenset({"anthropic", "openai", "gemini", "openrouter"})--# All providers supported by native SDK path (dedicated + OpenAI-compatible via base_url).-# Local servers (ollama, lmstudio) expose OpenAI-compatible /v1 endpoints, so they fall-# through to OpenAIProvider with base_url from llm_defaults.json — same path as deepseek/-# kimi/mistral. The user's custom URL rides via the existing {provider}_proxy credential.-NATIVE_PROVIDERS = _DEDICATED_PROVIDERS | frozenset(- {- "xai",- "deepseek",- "kimi",- "mistral",- "ollama",- "lmstudio",- }-)---def create_provider(- provider: str,- api_key: str,- *,- proxy_url: Optional[str] = None,-) -> LLMProvider:- """Create a native LLM provider instance.-- Dedicated providers (anthropic, gemini, openrouter) use their own classes.- OpenAI-compatible providers use OpenAIProvider with base_url from config.- """- if provider == "anthropic":- from services.llm.providers.anthropic import AnthropicProvider-- return AnthropicProvider(api_key, proxy_url=proxy_url)-- if provider == "openai":- from services.llm.providers.openai import OpenAIProvider-- return OpenAIProvider(api_key, proxy_url=proxy_url)-- if provider == "gemini":- from services.llm.providers.gemini import GeminiProvider-- return GeminiProvider(api_key, proxy_url=proxy_url)-- if provider == "openrouter":- from services.llm.providers.openrouter import OpenRouterProvider-- return OpenRouterProvider(api_key, proxy_url=proxy_url)-- # OpenAI-compatible providers: use OpenAIProvider with base_url from config- from services.llm.config import get_provider_config-- config = get_provider_config(provider)- if config and config.base_url:- from services.llm.providers.openai import OpenAIProvider-- return OpenAIProvider(api_key, base_url=config.base_url, proxy_url=proxy_url)-- raise ValueError(f"Unknown provider: {provider}")---def is_native_provider(provider: str) -> bool:- return provider in NATIVE_PROVIDERSdiff --git a/server/services/temporal/trigger_listener_workflow.py b/server/services/temporal/trigger_listener_workflow.pyindex ce9a394a..277c41b9 100644--- a/server/services/temporal/trigger_listener_workflow.py+++ b/server/services/temporal/trigger_listener_workflow.py@@ -353,7 +353,6 @@ def _build_run_graph("""from constants import (AI_AGENT_TYPES,- TOOLKIT_NODE_TYPES,WORKFLOW_TRIGGER_TYPES,)@@ -388,14 +387,6 @@ def _build_run_graph(continuedownstream_ids.add(source)- # Sub-nodes connected to toolkit nodes (n8n Sub-Node pattern).- toolkit_ids = {n["id"] for n in nodes if n.get("type") in TOOLKIT_NODE_TYPES and n["id"] in downstream_ids}- for edge in edges:- target = edge.get("target")- source = edge.get("source")- if target in toolkit_ids and source not in downstream_ids:- downstream_ids.add(source)-# Tool nodes connected to AI Agent's input-tools handle.agent_ids = {n["id"] for n in nodes if n.get("type") in AI_AGENT_TYPES and n["id"] in downstream_ids}for edge in edges:diff --git a/server/services/temporal/workflow.py b/server/services/temporal/workflow.pyindex 67820502..b759c525 100644--- a/server/services/temporal/workflow.py+++ b/server/services/temporal/workflow.py@@ -69,7 +69,7 @@ AGENT_WORKFLOW_TYPES = frozenset(["aiAgent","chatAgent",- # Specialized agents (12)+ # Specialized agents (11)"android_agent","coding_agent","web_agent",@@ -578,9 +578,6 @@ class MachinaWorkflow:return node_id, {"success": False, "error": str(e)}# Wait for first completion using Temporal's wait- [h for _, h in items]-- # Use asyncio.wait pattern via workflow.waitawait workflow.wait_condition(lambda: any(h.done() for _, h in items))# Find the completed onediff --git a/server/tests/llm/test_factory.py b/server/tests/llm/test_factory.pydeleted file mode 100644index 795e771d..00000000--- a/server/tests/llm/test_factory.py+++ /dev/null@@ -1,79 +0,0 @@-"""Test create_provider() factory and provider detection."""--from unittest.mock import patch--from services.llm.factory import create_provider, is_native_provider, NATIVE_PROVIDERS-from services.llm.config import detect_provider_from_model-from services.llm.protocol import LLMProvider---def test_native_providers_set():- assert NATIVE_PROVIDERS == {- "anthropic",- "openai",- "gemini",- "openrouter",- "xai",- "deepseek",- "kimi",- "mistral",- "ollama",- "lmstudio",- }---def test_is_native_provider():- assert is_native_provider("openai") is True- assert is_native_provider("groq") is False---def test_create_anthropic_provider():- with patch("anthropic.AsyncAnthropic"):- p = create_provider("anthropic", "sk-test")- assert p.provider_name == "anthropic"- assert isinstance(p, LLMProvider)---def test_create_openai_provider():- with patch("openai.AsyncOpenAI"):- p = create_provider("openai", "sk-test")- assert p.provider_name == "openai"- assert isinstance(p, LLMProvider)---def test_create_gemini_provider():- with patch("google.genai.Client"):- p = create_provider("gemini", "key")- assert p.provider_name == "gemini"- assert isinstance(p, LLMProvider)---def test_create_openrouter_provider():- with patch("openai.AsyncOpenAI"):- p = create_provider("openrouter", "or-key")- assert p.provider_name == "openrouter"---def test_create_xai_uses_openai_with_base_url():- with patch("openai.AsyncOpenAI") as mock_cls:- p = create_provider("xai", "xai-key")- assert p.provider_name == "openai" # reuses OpenAI class- mock_cls.assert_called_once()- call_kwargs = mock_cls.call_args[1]- assert call_kwargs["base_url"] == "https://api.x.ai/v1"---def test_create_unknown_raises():- import pytest-- with pytest.raises(ValueError, match="Unknown provider"):- create_provider("nonexistent", "key")---def test_detect_provider_from_model():- assert detect_provider_from_model("gpt-5.2") == "openai"- assert detect_provider_from_model("o3-mini") == "openai"- assert detect_provider_from_model("claude-sonnet-4-6") == "anthropic"- assert detect_provider_from_model("gemini-2.5-flash") == "gemini"- assert detect_provider_from_model("grok-3") == "xai"- assert detect_provider_from_model("unknown-xyz") == "openai" # fallback