skills/proffesor-for-testing__agentic-qe
1
0
Fork 0
mirror of https://github.com/proffesor-for-testing/agentic-qe.git synced 2026-09-19 08:45:47 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
testing-models
proffesor-for-testing__agen…/docs/qe-reports-3-6-8/02-code-smells-report.md
T
Profa 15a666bfcb plan + aqe init
2026-02-16 09:23:56 +00:00

19 KiB
Raw Permalink Blame History

Code Quality Analysis Report -- v3/src Code Smells

Version: 3.6.8 Date: 2026-02-16 Scope: /workspaces/agentic-qe-new/v3/src/ (942 TypeScript files, 478,814 lines)

Summary

Metric Value
Overall Quality Score 5.5 / 10
Files Analyzed 942
Total Issues Found ~580+
Critical Issues 12
High Severity Issues 38
Medium Severity Issues ~230
Low Severity Issues ~300
Technical Debt Estimate 160-200 hours

The v3 codebase has solid domain-driven structure and good use of Result types for error handling, but suffers from significant God class/file problems, pervasive as any casting that defeats TypeScript's type safety, a repeated consensus mixin anti-pattern across all domain coordinators, and widespread magic numbers. The coordinator classes in particular exhibit both excessive size and duplicated boilerplate.

1. God Classes / God Files (Critical)

Severity: CRITICAL -- 28 files exceed 1,500 lines; 85 files exceed 1,000 lines; 398 files exceed 500 lines (42% of codebase).

The project's CLAUDE.md mandates a 500-line file limit. Over 42% of source files violate this rule.

Worst Offenders (>2,000 lines)

File Lines Concern
v3/src/domains/quality-assessment/coordinator.ts 2,426 Quality assessment orchestration
v3/src/kernel/unified-memory.ts 2,272 Memory management + DB + CRDT
v3/src/domains/security-compliance/services/security-auditor.ts 2,228 Security scanning
v3/src/coordination/workflow-orchestrator.ts 2,219 Workflow orchestration
v3/src/coordination/queen-coordinator.ts 2,202 Agent coordination
v3/src/domains/code-intelligence/coordinator.ts 2,159 Code intelligence
v3/src/domains/visual-accessibility/services/accessibility-tester.ts 2,126 Accessibility testing
v3/src/init/init-wizard.ts 2,113 Init wizard
v3/src/domains/learning-optimization/coordinator.ts 2,094 Learning optimization
v3/src/cli/commands/learning.ts 2,048 CLI learning commands

Analysis

All 13 domain coordinator files are oversized, ranging from 809 lines (enterprise-integration) to 2,426 lines (quality-assessment). These coordinators violate the Single Responsibility Principle by combining initialization, disposal, consensus management, event handling, and domain-specific orchestration in a single class.

Recommendation: Extract consensus lifecycle, event subscription setup, and service orchestration into separate collaborators. Introduce a base coordinator class that handles the common lifecycle pattern (see Section 3 below).

2. any Type Usage (High)

Severity: HIGH -- 103 occurrences across 40 files (typed any declarations); 24 as any) casts; 7 explicit : any parameter/return types.

Pattern A: as any Casting on Consensus Mixin (20 occurrences, 7 files)

// Found in ALL domain coordinators:
await (this.consensusMixin as any).initializeConsensus();
await (this.consensusMixin as any).disposeConsensus();
return (this.consensusMixin as any).isConsensusAvailable?.() ?? false;

Files affected:

  • /workspaces/agentic-qe-new/v3/src/domains/visual-accessibility/coordinator.ts (lines 269, 288, 1701)
  • /workspaces/agentic-qe-new/v3/src/domains/security-compliance/coordinator.ts (lines 300, 373, 1349)
  • /workspaces/agentic-qe-new/v3/src/domains/defect-intelligence/coordinator.ts (lines 219, 231)
  • /workspaces/agentic-qe-new/v3/src/domains/requirements-validation/coordinator.ts (lines 280, 343, 1334)
  • /workspaces/agentic-qe-new/v3/src/domains/test-generation/coordinator.ts (lines 393, 414, 1595)
  • /workspaces/agentic-qe-new/v3/src/domains/enterprise-integration/coordinator.ts (lines 193, 207, 809)
  • /workspaces/agentic-qe-new/v3/src/domains/code-intelligence/coordinator.ts (lines 404, 507, 2016)

This indicates the ConsensusEnabledMixin interface does not properly expose initializeConsensus(), disposeConsensus(), and isConsensusAvailable() on the mixed-in type, forcing every consumer to cast to any.

Pattern B: as any on MCP Tool Registration (6 occurrences, 1 file)

// /workspaces/agentic-qe-new/v3/src/mcp/server.ts (lines 677-702)
this.registry.register(tool.definition, tool.handler as any);

Repeated 6 times in initialize() for CORE_TOOLS, TASK_TOOLS, AGENT_TOOLS, DOMAIN_TOOLS, MEMORY_TOOLS, CROSS_PHASE_TOOLS. The tool handler type does not match the registry's expected signature.

Pattern C: Database Query Results as any[] (15 occurrences)

// Example from multiple files:
`).all(this.config.maxOutcomesInMemory) as any[];

Files affected:

  • /workspaces/agentic-qe-new/v3/src/feedback/test-outcome-tracker.ts (line 130)
  • /workspaces/agentic-qe-new/v3/src/feedback/coverage-learner.ts (line 128)
  • /workspaces/agentic-qe-new/v3/src/routing/routing-feedback.ts (line 114)
  • /workspaces/agentic-qe-new/v3/src/learning/real-qe-reasoning-bank.ts (line 1220)
  • /workspaces/agentic-qe-new/v3/src/integrations/rl-suite/persistence/q-value-store.ts (lines 319, 341, 420, 421)
  • /workspaces/agentic-qe-new/v3/src/integrations/agentic-flow/reasoning-bank/pattern-evolution.ts (lines 487, 606, 790, 806)
  • /workspaces/agentic-qe-new/v3/src/integrations/agentic-flow/reasoning-bank/experience-replay.ts (line 750)
  • /workspaces/agentic-qe-new/v3/src/integrations/agentic-flow/reasoning-bank/trajectory-tracker.ts (lines 397, 648)

Recommendation: Define row-type interfaces for each SQL query and use typed wrappers around better-sqlite3 .all() / .get() calls.

Pattern D: Monkey-Patching with as any (8 occurrences, 1 file)

// /workspaces/agentic-qe-new/v3/src/performance/optimizer.ts
(eventAdapter as any).emit = (event: string, ...args: unknown[]) => { ... };
(eventAdapter as any).getOptimizationStats = () => ({ ... });
(taskManager as any).getTask = (id: string) => { ... };
(surfaceGenerator as any).getSurface = (id: string) => { ... };

This is runtime monkey-patching that completely bypasses the type system. Use interface augmentation or wrapper classes instead.

Pattern E: Typed any Properties (2 occurrences)

// /workspaces/agentic-qe-new/v3/src/governance/continue-gate-integration.ts:43
private guidanceContinueGate: any = null;

// /workspaces/agentic-qe-new/v3/src/governance/memory-write-gate-integration.ts:44
private guidanceMemoryGate: any = null;

3. Code Duplication (High)

Severity: HIGH -- Significant structural duplication across domain coordinators and error handling.

Pattern A: Coordinator Lifecycle Boilerplate (13 files)

All 13 domain coordinators follow the exact same lifecycle pattern:

async initialize(): Promise<void> {
  // ... domain-specific setup ...
  await (this.consensusMixin as any).initializeConsensus();
  // ... event subscriptions ...
}

async dispose(): Promise<void> {
  await (this.consensusMixin as any).disposeConsensus();
  // ... cleanup subscriptions ...
}

Files: All v3/src/domains/*/coordinator.ts (13 files)

This boilerplate should be extracted into a BaseDomainCoordinator abstract class with template method hooks.

Pattern B: Error Coercion Pattern (298 occurrences across 150 files)

error instanceof Error ? error.message : String(error)

Found 298 times. This should be a shared utility function (e.g., toErrorMessage(error: unknown): string).

Pattern C: Error Wrapping Pattern (426 occurrences across 129 files)

error instanceof Error ? error : new Error(String(error))

Found 426 times. Same recommendation -- extract to shared utility.

Pattern D: MCP Tool Registration Loop (6 identical loops)

// /workspaces/agentic-qe-new/v3/src/mcp/server.ts (lines 676-703)
for (const tool of CORE_TOOLS) {
  this.registry.register(tool.definition, tool.handler as any);
}
for (const tool of TASK_TOOLS) {
  this.registry.register(tool.definition, tool.handler as any);
}
// ... repeated 4 more times

Could be a single loop over [...CORE_TOOLS, ...TASK_TOOLS, ...AGENT_TOOLS, ...].

4. Magic Numbers and Strings (Medium)

Severity: MEDIUM -- 60+ occurrences of hardcoded numeric constants scattered across files.

Positive Finding

The codebase has /workspaces/agentic-qe-new/v3/src/kernel/constants.ts which centralizes many constants. However, numerous files still use inline magic numbers instead of referencing these constants.

Examples

// /workspaces/agentic-qe-new/v3/src/validation/parallel-eval-runner.ts:65-67
timeout: 30000,
progressIntervalMs: 5000,

// /workspaces/agentic-qe-new/v3/src/kernel/unified-persistence.ts:52-54
mmapSize: 64 * 1024 * 1024, // 64MB
busyTimeout: 5000,

// /workspaces/agentic-qe-new/v3/src/performance/optimizer.ts:129-143
poolSize: 1000,
maxSize: 10000,
ttl: 60000, // 1 minute
threshold: 1024, // 1KB

// /workspaces/agentic-qe-new/v3/src/integrations/ruvector/persistent-q-router.ts:102-103
lambda: 1000,
consolidationInterval: 5 * 60 * 1000, // 5 minutes

// /workspaces/agentic-qe-new/v3/src/kernel/memory-factory.ts:234-247
cleanupInterval: 60000,
busyTimeout: 10000,

Recommendation: Migrate all timeout, size, and interval values to kernel/constants.ts or to module-level named constants. The constants file already has BUSY_TIMEOUT_MS: 5000 and CLEANUP_INTERVAL_MS: 60000, but not all call sites use them.

5. Swallowed / Silent Error Handling (Medium)

Severity: MEDIUM -- ~15 catch blocks suppress errors silently or return without logging.

Pattern A: "Non-critical" Catch-and-Ignore (10+ occurrences)

// /workspaces/agentic-qe-new/v3/src/kernel/unified-persistence.ts:312-313
} catch (error) {
  // Non-critical: cleanup errors during shutdown
}

// /workspaces/agentic-qe-new/v3/src/kernel/unified-memory.ts:2137-2138
} catch (error) {
  // Non-critical: file stat errors during storage stats
}

// /workspaces/agentic-qe-new/v3/src/kernel/hybrid-backend.ts:376-377
} catch (e) {
  // Non-critical
}

While some of these are legitimately non-critical, the pattern of catching and completely discarding errors makes debugging difficult. At minimum, log at debug level.

Pattern B: Silent Return (1 occurrence)

// /workspaces/agentic-qe-new/v3/src/sync/readers/sqlite-reader.ts:159-160
} catch (error) {
  return 0;
}

Silently returns a default value on error with no logging.

6. TODO/FIXME Markers -- Incomplete Implementation (Medium)

Severity: MEDIUM -- 15 TODO comments indicating unfinished features.

File Line TODO
v3/src/cli/commands/learning.ts 554 // TODO: Implement actual promotion via reasoningBank
v3/src/domains/visual-accessibility/services/browser-swarm-coordinator.ts 522 // TODO: Integrate with axe-core or similar
v3/src/mcp/http-server.ts 489 // TODO: Wire to actual MCP tool handlers
v3/src/mcp/transport/websocket/connection-manager.ts 484 totalReconnections: 0, // TODO: track reconnections
v3/src/integrations/ruvector/server-client.ts 402 // TODO: Parse actual health response
v3/src/integrations/ruvector/server-client.ts 450 // TODO: When server API becomes available
v3/src/integrations/browser/client-factory.ts 64 // TODO: Implement Vibium availability check
v3/src/integrations/browser/client-factory.ts 270 // TODO: Return real Vibium client when implemented
v3/src/integrations/browser/client-factory.ts 279 // TODO: Return real Vibium client when implemented
v3/src/integrations/agentic-flow/reasoning-bank/index.ts 319 similarity: 0.8, // TODO: Calculate actual similarity
v3/src/learning/qe-unified-memory.ts 1129 // TODO: Implement JSON migration
v3/src/learning/v2-to-v3-migration.ts 419 // TODO: Migrate embeddings if available

The // TODO: Wire to actual MCP tool handlers in http-server.ts is concerning as it suggests the HTTP transport may not be functional.

7. Inappropriate Intimacy / Deep Coupling (Medium)

Severity: MEDIUM

Deep Import Paths (23 occurrences)

Files reaching 4+ levels deep into other modules via relative imports:

// /workspaces/agentic-qe-new/v3/src/domains/security-compliance/services/scanners/dast-scanner.ts:7
import { Result, ok, err } from '../../../../shared/types/index.js';

// /workspaces/agentic-qe-new/v3/src/domains/test-execution/services/e2e/e2e-coordinator.ts:11-12
import type { VibiumClient, ... } from '../../../../integrations/vibium';
import type { IBrowserClient } from '../../../../integrations/browser';

// /workspaces/agentic-qe-new/v3/src/domains/code-intelligence/services/c4-model/index.ts:13-15
import { Result, ok, err, Severity, Priority } from '../../../../shared/types';
import type { MemoryBackend, StoreOptions } from '../../../../kernel/interfaces';
import { NomicEmbedder, ... } from '../../../../shared/embeddings';

These deep relative paths are fragile and indicate either missing barrel exports or overly nested directory structures.

Recommendation: Use TypeScript path aliases (e.g., @kernel/interfaces, @shared/types) or ensure barrel exports exist at module boundaries.

8. Console Logging Instead of Structured Logging (Medium)

Severity: MEDIUM -- 216 console.log/warn/error calls spread across 30+ files.

The codebase has a logging module at v3/src/logging/, but many files bypass it and use console.* directly. Worst offenders:

File console.* calls
v3/src/performance/run-gates.ts 29
v3/src/learning/dream/dream-scheduler.ts 22
v3/src/learning/real-qe-reasoning-bank.ts 21
v3/src/strange-loop/strange-loop.ts 15
v3/src/kernel/unified-memory-migration.ts 14
v3/src/learning/qe-reasoning-bank.ts 13
v3/src/learning/qe-unified-memory.ts 13

Recommendation: Replace console.* calls with the structured logger from v3/src/logging/. This enables log level filtering, structured metadata, and consistent formatting.

9. Shotgun Surgery Pattern (Medium)

Severity: MEDIUM -- Changes to the consensus mixin or domain coordinator lifecycle require edits to 13+ files.

Consensus Mixin Changes

Any change to how consensus is initialized, disposed, or queried requires modifying all 13 coordinator files because of the duplicated (this.consensusMixin as any).* pattern. This is the textbook definition of shotgun surgery.

Error Handling Changes

With 298+ copies of error instanceof Error ? error.message : String(error) and 426+ copies of the wrapping variant, any change to error handling strategy requires updating hundreds of files.

Domain Event Schema Changes

Adding a new standard field to domain events would require changes across all coordinator files and their associated service files.

10. Long Parameter Lists (Low)

Severity: LOW -- Only 3 instances found of functions with 5+ parameters.

// /workspaces/agentic-qe-new/v3/src/integrations/rl-suite/algorithms/ddpg.ts:287
private generateReasoning(state: RLState, action: RLAction, value: number, original: number, noisy: number): string

The WASM binding declarations in agent_booster_wasm_bg.wasm.d.ts have 7-parameter functions, but those are auto-generated and not actionable.

The codebase generally uses config objects for complex parameter sets, which is good practice.

Positive Findings

  1. Result Type Usage: The codebase uses Result<T> with ok() / err() wrappers consistently across domain services, providing explicit error handling without exceptions.

  2. Domain-Driven Structure: Clear bounded contexts in v3/src/domains/ with coordinator + services + plugin patterns per domain.

  3. Constants File Exists: v3/src/kernel/constants.ts centralizes many named constants with documentation -- the issue is incomplete adoption.

  4. Event-Driven Architecture: The EventBus and domain event system provides loose coupling between domains for cross-cutting concerns.

  5. Lazy Initialization: Persistent resources (DB connections, CRDT stores) use lazy init() patterns with | null types, avoiding unnecessary resource allocation.

  6. Config Objects: Most complex constructors use typed configuration interfaces rather than long parameter lists.

  7. Explicit Error Logging in Catch Blocks: Most catch blocks log warnings with context ([ModuleName] description:, error) rather than silently swallowing.

Refactoring Priorities

Priority 1: Fix Consensus Mixin Typing (Est. 8 hours)

  • Fix the ConsensusEnabledMixin type definition to properly expose initializeConsensus(), disposeConsensus(), and isConsensusAvailable().
  • Eliminate all 20 (this.consensusMixin as any) casts across 7 coordinator files.
  • This also reduces shotgun surgery risk.

Priority 2: Extract Base Domain Coordinator (Est. 16 hours)

  • Create BaseDomainCoordinator abstract class with template methods for onInitialize(), onDispose().
  • Move consensus lifecycle, event subscription management, and health check patterns into the base class.
  • Reduces 13 coordinator files by 100-200 lines each.

Priority 3: Shared Error Utilities (Est. 4 hours)

  • Create toErrorMessage(error: unknown): string and toError(error: unknown): Error.
  • Replace 700+ duplicated inline patterns.

Priority 4: Typed Database Row Interfaces (Est. 12 hours)

  • Define row-type interfaces for all SQL queries.
  • Replace 15+ as any[] casts on database results with typed alternatives.

Priority 5: Split God Files (Est. 40 hours)

  • Target files > 1,500 lines first (28 files).
  • unified-memory.ts (2,272 lines) should split into memory manager, DB layer, and CRDT integration.
  • workflow-orchestrator.ts (2,219 lines) should split workflow types, execution engine, and step management.
  • Domain coordinators will shrink via Priority 2.

Priority 6: Consolidate Magic Numbers (Est. 8 hours)

  • Audit all timeout/size/interval constants.
  • Move to kernel/constants.ts or domain-specific constant files.
  • Update all call sites to reference named constants.

Priority 7: Replace Console Logging (Est. 12 hours)

  • Replace 216 console.* calls with structured logger.
  • Ensure all catch blocks use the logger rather than console.warn/error.

Technical Debt Distribution by Module

Module Issues Dominant Smell
domains/ (13 coordinators) ~95 God classes, duplicated lifecycle, as any consensus
kernel/ ~25 God files (unified-memory), magic numbers
coordination/ ~30 God files (workflow-orchestrator, queen-coordinator)
learning/ ~25 Console logging, as any[] DB queries
mcp/ ~20 as any tool registration, TODO stubs
performance/ ~15 as any monkey-patching, magic numbers
governance/ ~10 any-typed properties
integrations/ ~25 as any[] DB queries, TODOs
cli/ ~15 as any casts, God file (learning.ts)
feedback/ ~10 Duplicated tracker patterns

Report generated by Code Quality Analyzer on 2026-02-16

Reference in New Issue View Git Blame Copy Permalink