Animus Known Issues & Gaps¶

Last updated: 2026-01-27 (v0.6.0)

Testing Gaps¶

HIGH: No Offline/Mock LLM Testing RESOLVED¶

Status: Resolved (v0.6.0) — MockModel added to animus/cognitive.py

Solution implemented: - MockModel(ModelInterface) with configurable default response and response map - ModelConfig.mock() classmethod for easy construction - MOCK provider in ModelProvider enum + create_model() factory - Call history recording (calls: list[dict]) for test assertions - Tests in tests/test_cognitive_mock.py covering MockModel, CognitiveLayer, and DecisionFramework

MEDIUM: API Server Test Coverage RESOLVED¶

Status: Resolved (v0.6.0) — tests/test_api.py added

Solution implemented: - 30 tests covering all endpoints (status, chat, memory CRUD, tools, tasks CRUD, decisions, brief, integrations, learning, auth) - Uses MockModel backend for deterministic responses - Skips gracefully when FastAPI not installed

MEDIUM: Asyncio Deprecation Warnings RESOLVED¶

Status: Resolved (v0.6.0) — All asyncio.get_event_loop().run_until_complete() replaced with asyncio.run()

Files fixed: - animus/__main__.py — 11 call sites - tests/test_phase4.py — 7 call sites - All 6 previously failing tests now pass

LOW: Voice Interface Testing¶

Problem: Voice interface requires audio hardware, cannot test in CI.

Affected Components: - VoiceInput - Microphone capture - VoiceOutput - Audio playback - Whisper transcription

Proposed Solution: - Add audio file input mode for testing - Mock sounddevice for unit tests - Integration test with pre-recorded audio files

Priority: Low - Hardware dependent

API Inconsistencies¶

Memory.create() Missing RESOLVED¶

Status: Resolved (v0.6.0) — Memory.create() classmethod added to animus/memory.py

Solution: Factory method auto-generates id (UUID), created_at, updated_at. All other fields have sensible defaults. Tests in tests/test_core.py::TestMemoryCreate.

Inconsistent Method Names¶

Problem: Similar operations have different names across modules.

Module	Add	Complete/Apply
TaskTracker	`add()`	`complete()`
MemoryStore	`store()`	N/A
LearningLayer	N/A	`approve_learning()`

Proposed: Standardize on add/remove or document conventions.

Priority: Low - Documentation

Feature Gaps¶

No Web Dashboard¶

Problem: Learning transparency data has no visual interface.

Current: CLI-only via /learning command

Proposed: - Simple HTML dashboard served by API - Or separate React/Vue frontend - Real-time updates via WebSocket

Priority: Medium - Usability

No Scheduled Tasks RESOLVED¶

Status: Resolved (v0.6.0) — Background scheduler added to LearningLayer

Solution: threading.Timer-based scheduler with start_auto_scan(interval_hours), stop_auto_scan(), and auto_scan_running property. Daemon thread, no external dependencies. Tests in tests/test_phase5.py::TestAutoScanScheduler.

No Data Export¶

Problem: No way to export memories, learnings, or decisions.

Proposed: - /export memories - JSON/CSV export - /export learnings - Learning history - /export decisions - Decision log - API endpoints for same

Priority: Low - Nice to have

Performance¶

ChromaDB Cold Start¶

Problem: First memory operation downloads 79MB embedding model.

Impact: ~3s delay on first use

Mitigation: Pre-download during install or first-run setup

Priority: Low - One-time cost

Security¶

No Input Sanitization Audit RESOLVED¶

Status: Resolved (v0.6.0) — Security hardening applied to animus/tools.py

Changes: - _validate_command(): Whitespace normalization to prevent bypass; blocks $(), backtick subshells, | sh, | bash - _tool_read_file(): Path resolution via .resolve() before I/O (prevents symlink traversal) - _tool_list_files(): Same .resolve() fix - _tool_web_search(): Control character stripping, 500-char query limit - Tests in tests/test_security.py::TestShellInjection and TestWebSearchSanitization

Next Actions¶

Add MockModel for testing without LLM
Add API integration tests
Fix asyncio deprecation warnings
Add Memory.create() factory method
Implement background scheduler for auto-scan
Security audit of tool inputs