refactor(backend): complete orchestrator consolidation (7→4 target)

Status

Continuation of #3393. Orchestrator consolidation is ~40% complete.

Current State

Still 7 orchestrator files (target: 4)
Backward-compat shims added in first pass
Full consolidation deferred

Remaining Files

autobot-backend/agents/agent_orchestrator.py
autobot-backend/api/orchestration.py
autobot-backend/enhanced_multi_agent_orchestrator.py
autobot-backend/enhanced_orchestration/orchestrator.py
autobot-backend/orchestrator.py
autobot-backend/services/advanced_workflow/orchestrator.py
autobot-backend/services/slm/deployment_orchestrator.py

Consolidation Plan

Phase 2: Remove backward-compat shims (breaking change for 1-2 releases)
Phase 3: Evaluate remaining specialized orchestrators (agent_orchestration, advanced_workflow, slm deployment)
Phase 4: Consolidate into single canonical pattern or justified service-specific variants

Related: #3393

refactor(backend): consolidate orchestrator files 5→4 (#4048)

refactor(backend): clean up stale EnhancedOrchestrator docstrings (#4048

Merge pull request #5027 from mrveiss/issue-4048

✅ Closing with proof — 7→4 target exceeded (now 2 production files, 1 actual Orchestrator class)

Exact commit chain that resolved the 7-file consolidation

Original file	Resolution commit	Date	PR
`agents/agent_orchestrator.py`	`acf060502` refactor(backend): remove backward-compat shim orchestrator files (#3393) (#4869)	2026-04-16	#4869
`enhanced_multi_agent_orchestrator.py`	`8493bb7f8` refactor(orchestrator): merge EnhancedMultiAgentOrchestrator into ConsolidatedOrchestrator (#5040)	2026-04-17	#5040
`enhanced_orchestration/orchestrator.py`	`2d9e4d61f` refactor(orchestrator): decompose god class into collaborators (#5058)	2026-04-27	#5058
`services/advanced_workflow/orchestrator.py`	`27d704020` refactor(services): rename specialized orchestrators (#5040)	2026-04-17	#5040
`services/slm/deployment_orchestrator.py`	`27d704020` (same commit, same PR)	2026-04-17	#5040
`orchestrator.py`	kept as canonical single conductor — `00919e750` rename `ConsolidatedOrchestrator` → `Orchestrator`	2026-04-17	#5040
`api/orchestration.py`	kept as FastAPI router only (delegates to canonical Orchestrator)	—	—

This issue's own work: 350e5d511 refactor(backend): consolidate orchestrator files 5→4 (#4048) — landed via PR #5027 on 2026-04-17. Plus dc413f324 cleanup of stale EnhancedOrchestrator docstrings (also #4048).

Subsequent cleanup: dc5342f70 refactor(orchestration): 5 targeted fixes from discovery audit (#6404, 2026-04-28).

Final state — verified 2026-05-04

$ grep -rn "^class.*Orchestrator" autobot-backend --include="*.py" | grep -v __pycache__ | grep -v test
autobot-backend/models/settings.py:207:class OrchestratorSettings(BaseSettings):     # config dataclass
autobot-backend/orchestrator.py:146:class OrchestratorConfig:                          # internal config
autobot-backend/orchestrator.py:175:class Orchestrator:                                # ← ONLY actual orchestrator class

Of the 7 originally-listed files, 5 are gone, 1 is the canonical (orchestrator.py), and 1 is a thin FastAPI router (api/orchestration.py). Production-orchestrator-class count = 1. Original target was 4. Target exceeded.

Phase 2 / Phase 3 / Phase 4 status (from original plan)

✅ Phase 2 (remove backward-compat shims) — completed by refactor(backend): complete orchestrator consolidation (#3393) #4869 (acf060502)
✅ Phase 3 (evaluate remaining specialized orchestrators) — completed by refactor(backend): one conductor — single Orchestrator class, rename 6 misnamed '*Orchestrator' classes #5040 (renamed/folded all of agent_orchestration, advanced_workflow, slm/deployment)
✅ Phase 4 (consolidate into single canonical pattern) — completed by refactor(backend): one conductor — single Orchestrator class, rename 6 misnamed '*Orchestrator' classes #5040 (Orchestrator is the single conductor) + refactor(orchestrator): three overlapping workflow execution paths in single Orchestrator class #5058 (decomposed into reusable collaborators)

Deep-audit follow-ups filed today (2026-05-04)

A read-only deep-audit of the four remaining orchestration packages (agents/agent_orchestration/, enhanced_orchestration/, orchestration/, services/orchestration/) surfaced 7 separate consolidation opportunities. None of these are part of issue #4048's scope — they are new tech debt found during closure verification:

#	Severity	Finding
#6816	P1	Wire in or retire abandoned causal subsystem (5 files, 1,722 LOC, 0 production callers)
#6817	P2	Dual `WorkflowPlanner` name collision — orphaned canonical vs live `StrategyPlanner`
#6818	P2	Dual `AgentCapability` definitions — Enum vs dataclass, same name
#6819	P2	Dual `AgentRouter` classes (522 LOC vs 68 LOC) — same name, different scope
#6820	P2	4 orphaned orchestration modules (1,260 LOC, 0 production callers)
#6821	P3	Topology modules unwired (336 LOC, #2137/#2138)
#6822	P3	`subagent_dispatcher.py` lives in `services/`, test-only — layering violation

Total tech debt surfaced: ~3,654 LOC of unwired or duplicated orchestration code, scoped into discrete actionable issues.

Acceptance criteria — final check

✅ Phase 2 complete (backward-compat shims removed in refactor(backend): complete orchestrator consolidation (#3393) #4869)
✅ Phase 3 complete (specialized orchestrators evaluated/folded in refactor(backend): one conductor — single Orchestrator class, rename 6 misnamed '*Orchestrator' classes #5040)
✅ Phase 4 complete (single canonical Orchestrator in orchestrator.py, decomposed into collaborators in refactor(orchestrator): three overlapping workflow execution paths in single Orchestrator class #5058)
✅ Original 7→4 target met (now 2 files, 1 actual orchestrator class)
✅ Deep-audit run; 7 discovery issues filed for new tech debt

Self-audit follow-up — 7 additional discovery issues filed

After the user pushed back on whether the deep-audit findings were fully captured, I cross-checked the 4 audit reports against the 7 issues already filed and found I had missed 7 more. Filed now:

#	Sev	Topic
#6826	P1	#3228 closed prematurely — executor fragmentation persists (5 files, 3,223 LOC, no unified strategy)
#6827	P2	`WorkflowExecutor` god class (1,178 LOC) — decompose into StepOrchestrator/ParallelCoordinator/CheckpointResumer
#6828	P2	`AgentRegistry` fragmentation — 4 implementations + 1 Protocol with same name
#6829	P2	`RLRouter` (Issue #2092) infrastructure-complete but unreachable in production
#6830	P2	`ExecutionStrategyHandler` god class — 6 strategies in one 317-LOC class
#6831	P3	TYPE_CHECKING import cycle: `orchestration.performance_tracker` ↔ `enhanced_orchestration.types`
#6832	P3	Composability gap — `SuccessCriteriaEvaluator` + `CollaborationCoordinator` are single-caller modules

Total tech debt surfaced by this issue's closure audit: 14 follow-up issues (#6816–#6822 + #6826–#6832).

The most actionable: #6826 (P1) reveals that #3228 was closed before the migration was complete — code comments still reference it as "tracked in #3228" for work that hasn't happened.

Framing correction — "orphans" was wrong; these are unwired-by-design completed features

User pushed back: the modules I called "orphans" all have explicit tracker issue references in their docstrings. Checking each:

Module	Tracker	Tracker state
causal_executor.py + 4 siblings	#2154	CLOSED
workflow_planner.py, workflow_documentation.py, agent_registry.py	#381	CLOSED
workflow_memory.py	#3019	CLOSED
topology.py	#2137	CLOSED
topology_routing.py	#2138	CLOSED
rl_router.py	#2092	CLOSED
success_criteria.py	#3293	CLOSED

Every tracker is closed but the code is not reachable from a production path. That's not orphan code — that's systemic premature-closure, the same pattern this very issue (#4048) almost fell into until the user pushed back on my first proposed closure.

Filed as #6836 (P1) with all 3,906 LOC of evidence and recommended CLAUDE.md updates to the closure verification gate.

The 14 follow-up issues (#6816–#6822, #6826–#6832) remain valid — they describe the wire-in work for each module — but their framing should be "complete the integration that the prematurely-closed parent tracker missed," not "audit and remove orphan." The issue bodies use "wire it in or retire," which is correct; only my user-facing summary used the wrong word.

feat(closure-gate): layered defense against premature feature-issue c…

Navigation Menu

refactor(backend): complete orchestrator consolidation (7→4 target) #4048

Status

Current State

Remaining Files

Consolidation Plan

✅ Closing with proof — 7→4 target exceeded (now 2 production files, 1 actual Orchestrator class)

Exact commit chain that resolved the 7-file consolidation

Final state — verified 2026-05-04

Phase 2 / Phase 3 / Phase 4 status (from original plan)

Deep-audit follow-ups filed today (2026-05-04)

Acceptance criteria — final check

Self-audit follow-up — 7 additional discovery issues filed

Framing correction — "orphans" was wrong; these are unwired-by-design completed features

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

refactor(backend): complete orchestrator consolidation (7→4 target) #4048

Description

Status

Current State

Remaining Files

Consolidation Plan

Activity

mrveiss commented on May 4, 2026

✅ Closing with proof — 7→4 target exceeded (now 2 production files, 1 actual Orchestrator class)

Exact commit chain that resolved the 7-file consolidation

Final state — verified 2026-05-04

Phase 2 / Phase 3 / Phase 4 status (from original plan)

Deep-audit follow-ups filed today (2026-05-04)

Acceptance criteria — final check

mrveiss commented on May 4, 2026

Self-audit follow-up — 7 additional discovery issues filed

mrveiss commented on May 4, 2026

Framing correction — "orphans" was wrong; these are unwired-by-design completed features

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions