Skip to content

EvoSpikeNet project feature implementation status

Author: Masahiro Aoki

  • Last updated: May 6, 2026

This document reorganizes the description of Remaining_Functionality in the following order.

  1. Update history
  2. Implemented/Not yet implemented
  3. Implementation completed

Judgment criteria (based on source code)

  • Implemented: An implementation exists in the repository and is available in at least one of unit/integration/e2e.
  • Under implementation: There are placeholders, partial implementations, waiting for integration, and unfinished production/optimization items.
  • Not implemented: There is no implementation, or an external contract/operation procedure is a completion condition and the code alone cannot be considered complete.

1. Update history

1.1 Operation hardening update (2026-04-24)

  • evospikenet.api abolished pseudo-degraded startup when PyTorch is not installed and changed to fail-fast startup.
  • Video analysis refuses to start if VIDEO_ANALYSIS_FAIL_CLOSED=true and real backend cannot be satisfied.
  • Return source_backend / fallback_reason / quality_level / degraded in degraded response.
  • Changed the specification of AvailabilityMonitor to not automatically start the monitoring thread when importing.
  • Changed the specification to not silently complete missing genome/chromosome/topology in strict mode (EVOSPIKENET_STRICT_FUNCTIONAL_MODULES=true).

1.2 Video Scene Analysis added (2026-04-22)

-API: - POST /analyze/video - GET /results/{job_id} - Pipeline: - Frame extraction, audio extraction, detector inference, Whisper STT, event merging, Japanese narrative generation -SDK: - VideoSceneSDK (sync) - VideoSceneAsyncSDK (asynchronous) - Test: - API input validation - SDK synchronous/asynchronous unit testing

1.3 Phase E Connectome Integration (2026-03-18 ~ 2026-03-19)

  • E-0: Added design, schema, configuration, and test template.
  • E-1/E-2: Implemented connectome_loader.py, node_mapping.py, delay_buffer.py, zenoh_connectome_publisher.py.
  • E-3 (internal implementation): Implemented sync_connectome.py, brain_routing.py, auto_node_mapper.py, and E2E validation.

1.4 Phase D Distributed Node Integration (2026-03-11)

  • Added BrainSimulation compatible alias.
  • Implemented InstantiatedBrain.apply_weight_delta().
  • DistributedBrainNode.deploy_genome() implementation.
  • Implemented DistributedEvolutionEngine.deploy_to_nodes().

1.5 Biomimetic integration (2026-03-06/09)

  • Phase A/B completed.
  • Phase C (C-1 to C-4) completed.
  • Expanded Biomimetic API, SDK integration, and distributed_brain integration.

1.6 EEG/Security fix (2026-03-05)

  • Numerical stabilization of comparative_analysis.py.
  • Compatibility fixes for eeg_translator.py, spectrum_converter.py, device_interface.py.
  • Clarified HTTP 403 judgment when rejecting OPA.

1.7 Implementation updates (2026-05-06)

  • Video analysis: prod/staging default fail-closed, backend startup checks, and DeepSORT adapter updates reflected.
  • EEG: added gRPC streaming definitions (eeg_streaming_pb2.py / eeg_streaming_pb2_grpc.py).
  • Embodied PLA: reflected ROS2-backed _ROSRobotInterface implementation.

1.8 Document History

Date Version Change summary
2026-02-02 v1.0 Remaining_Functionality.md New creation
2026-02-25 v1.1 Organizing biomimicry 11-1 to 11-19 and implementation plan
2026-03-18 v1.2 Added Phase E (Section 17) and design artifacts
2026-03-19 v1.3 E-0/E-1/E-2 completion reflection
2026-03-19 v1.4 E-3 Internal implementation completion reflected
2026-04-24 v2.0 This reorganized version (update history → Implemented/Not implemented → Implemented)
2026-04-25 v2.1 2.1.1 Transfer of functional module layer to completion of implementation
2026-04-25 v2.2 2.1.2 Embodied PLA loop transferred to implementation completion
2026-04-25 v2.3 2.1.4 Distributed Facade Layer / 2.1.5 Transferring LLM learning hook to implementation completion
2026-04-25 v2.4 2.1.9 Foundation extension / 2.1.10 Transfer of language binding base to completion of implementation
2026-04-27 v2.5 3.21 GPU/CPU device consistency correction (AEG / ClosedLoopController) and SDK-URL all file replacement added to implementation completion
2026-04-27 v2.6 3.22 Evaluate_System.md Added implementation of all unsupported items (ALERT-001 / TEST-001 / TYPE-001 / MOCK-001 remaining / tasks.py High / IoU Low / translator Low / CRED-001 confirmation)
2026-05-01 v2.7 4.1 Reflecting fail-closed mock/placeholder detection points and enhanced observability (communication, compatible API, video analysis, PFC, time synchronization, service fallback, test-status)
2026-05-06 v2.8 Video-analysis strict/backends updates, EEG gRPC definitions, ROS2 integration reflection

2. Implemented/Not yet implemented

2.1 Implementation in progress (including partial implementation and placeholders)

2.1.1 Production backend integration of video analysis (partial implementation)

  • Target: evospikenet/video_analysis/backends.py, evospikenet/video_analysis/asr.py
  • Condition:
  • The real backend branch of MoveNetRealPoseBackend / STGCNRealActionBackend / WhisperRealASRBackend has been implemented.
  • Use stub/fallback when there are dependencies or models are not placed.
  • Unfinished points:
  • Production model deployment, operational monitoring, and SLO verification must be completed.

2.1.2 Complete type safety (ongoing)

  • Target: Feature 12 set
  • Condition:
  • Added type hints, mypy settings, and CI integration have been implemented.
  • However, type error convergence is still ongoing (gradual improvement phase).

2.1.3 Higher precision and faster space generation (ongoing)

  • Target:
  • evospikenet/services/spatial_generation_service.py
  • evospikenet/spatial/recognition.py
  • evospikenet/spatial/generation.py
  • evospikenet/spatial/attention.py
  • Condition:
  • Basic services, /generate, /quantum/infer, and high precision mode switching have been implemented.
  • Final optimization and production tuning of 50ms/node threshold continues.

2.1.4 Security operations integration (ongoing)

  • Target: evospikenet/security.py, evospikenet/foundation_extensions.py
  • Condition:
  • API key authentication, rate limiting, mTLS header validation, and certificate rotation functions have been implemented.
  • Production operation design for encryption at rest/KMS/Secrets linkage (Vault linkage, rotation daemon) is ongoing.

2.2 Not implemented/not started

2.2.1 External procedure dependence

  • E-3-5 HCP DUC acquisition procedures and data placement (data/hcp/ linkage) are dependent on external contracts/procedures and have not been started yet.

2.2.2 Long-term research themes (planning stage)

  • Quantum real machine collaboration (IBM Quantum / Google Quantum)
  • Production operation of hybrid quantum-classical system
  • BMI/BCI bidirectional closed loop integration (EEG→Brain Language infrastructure has been implemented)
  • Actual robot evaluation pipeline (including online evolution)
  • Enhanced automatic orchestration of cloud/edge mixture (distributed infrastructure has already been implemented)
  • Large-scale distributed demonstration of production operation of several thousand nodes

2.2.3 Expanded themes (mainly not yet started)

  • Production implementation of bias detection and mitigation
  • Operational integration of ethical assessment frameworks
  • Olfactory/Taste Sensor Integration
  • Advanced biosensor integration
  • Sophistication of continuous learning and memory management (large-scale operation optimization, strategy advancement, automatic parameter optimization)

3. Implementation completed

3.1 Core SNN Engine

  • Neuron layer:
  • LIF NeuronLayer
  • IzhikevichNeuronLayer
  • EntangledSynchronyLayer
  • Synapse management:
  • SynapseMatrixCSR
  • Attention:
  • ChronoSpikeAttention
  • Encoding: -TAS-Encoding
  • RateEncoder

3.2 Learning/Plasticity

  • STDP
  • Meta-STDP
  • Homeostasis
  • MetaPlasticity
  • EnergyManager
  • EnergyConstrainedPlasticityController
  • Surrogate gradients
  • HierarchicalPlasticityController series

3.3 Distributed processing/communication

  • Zenoh asynchronous communication platform
  • PFC Cognitive Control Routing
  • Raft consensus building
  • Node discovery/cleanup
  • Load balancing/dynamic capacity management
  • Geographically distributed node management (GeoNodeManager)
  • Communication compression framework (compression.py + CUDA wrapper)

3.4 Connectome Integration (Phase E)

  • E-0: Design/schema/configuration/test template
  • E-1: connectome_loader.py and ConnectomeLIFLayer
  • E-2: node_mapping.py, SparseDelayBuffer, ConnectomeMetadataPublisher
  • E-3 (internal implementation): sync_connectome.py, brain_routing.py, auto_node_mapper.py, E2E validation

3.5 Brain function simulation

  • Functional modules (visual/auditory/language/speech/motor/calculation)
  • Federated learning (Async-FedAvg)
  • Embodied PLA integration loop
  • Continuous learning and memory management (foundation):
  • LongTermMemoryModule
  • EpisodicMemory memory retention/integration/forgetting processing
  • MemoryIntegratorNode
  • ForgettingController
  • Storage node:
  • EpisodicMemoryNode
  • SemanticMemoryNode
  • MemoryIntegratorNode

3.6 Biomimetic functions (11-1 to 11-19)

The implementation completion of the following has been confirmed by checking the source.

  • 11-1 Delay/Brain wave rhythm
  • 11-2 Cell/synapse diversification
  • 11-3 Layered/hierarchical topology
  • 11-4 Regulator gating
  • 11-5 Storage system expansion
  • 11-6 Strengthening sensory-motor closed loop
  • 11-7 Energy homeostasis
  • 11-8 Developmental Dynamics
  • 11-9 Intention expression module
  • 11-10 Creativity generation
  • 11-11 Self-awareness/introspection
  • 11-12 Dynamic target selection
  • 11-13 Emotional system
  • 11-14 Sleep phase memory consolidation
  • 11-15 Mirror Neuron
  • 11-16 Acetylcholine system
  • 11-17 NAcc/VTA Reward Loop
  • 11-18 DMN dedicated module
  • 11-19 Curriculum learning scheduler

3.7 Feature 13 Advanced spatial recognition/generation

  • SpatialWhereNode (Where path)
  • SpatialWhatNode (What route)
  • SpatialIntegrationNode (What-Where integration)
  • SpatialAttentionControlNode (Attention Control)
  • A set of distributed brain (Rank 12-15) integration and related tests

3.8 Feature 20 Video scene description/3D spatial recognition/speech to text

Implemented modules:

  • event_schema.py (Unified Event Schema)
  • shot_boundary_detector.py
  • vad.py
  • depth_estimation.py
  • spatial_relations.py
  • narrative_generator.py
  • temporal_action_localizer.py
  • pose.py
  • tracking.py
  • action_recognition.py
  • asr.py
  • asr_policy.py
  • fusion.py
  • pipeline.py
  • backends.py
  • backends_real.py
  • postprocess.py
  • postprocess_polite.py
  • job_queue.py
  • privacy.py
  • metrics.py
  • runtime_metrics.py

API/SDK Implemented:

  • evospikenet/api_modules/video_analysis_api.py
  • evospikenet/video_analysis/worker.py
  • evospikenet.sdk.video_sdk.VideoSceneSDK
  • evospikenet.sdk.video_sdk.VideoSceneAsyncSDK

3.9 EEG Integration

  • Data conversion/analysis:
  • spike_extractor.py
  • eeg_converter.py
  • spectrum_converter.py
  • Streaming/Time Sync:
  • streaming.py
  • eeg_streaming_pb2.py
  • eeg_streaming_pb2_grpc.py
  • time_sync.py
  • Device integration:
  • OpenBCI / Muse / Emotiv / Neurosky driver
  • Brain Language Conversion:
  • eeg_translator.py
  • brain_language_decoder.py (includes scaffold + bench/learning steps)

3.10 RAG/Document management

  • rag_milvus.py / rag_backends.py / rag_client.py
  • PDF/Word/Excel/PPT/Markdown import
  • Version history/difference/job monitoring
  • Splitting large uploads/background processing

3.11 UI/API/SDK

  • Dash UI / Evolution dashboard / 3D visualization
  • FastAPI base API group
  • Python SDK (type safety, Jupyter integration)
  • Additional language proxy SDKs:
  • Go (sdk_go.py)
  • TypeScript (sdk_ts.py)
  • Swift (sdk_swift.py)

3.12 Security/Monitoring/Operations

-API Key Authentication - rate limit - CORS control - SpikeEncryption (PSK/DH/AES-256-GCM) - Audit log (audit_log.py) - Automatic recovery (auto_recovery.py) - Prometheus integration monitoring - CI/static analysis/coverage infrastructure

3.13 Patented technology implementation (MT25-EV001~MT25-EV030)

  • For the old ledger MT25-EV001 to MT25-EV030, it has been confirmed that the corresponding module has been implemented in the source.
  • Representative support:
  • ChronoSpikeAttention -TAS-Encoding
  • Quantum PFC / Q-PFC Loop
  • Energy Plasticity -Embodied PLA
  • Conscience Circuit
  • SNN-RAG Hybrid -Spike Encryption
  • Sleep/regulators/development/creativity/introspection/goal selection/coevolution, etc.

3.14 Other completed items

  • Sensor integration plug-in mechanism (USB/Stereo/IR/ONVIF/LiDAR/Environment/GPIO/Audio)
  • Document parser extension including markdown_parser.py
  • quantum/quantum_interface.py proxy
  • BiomimeticAdapter integration in distributed_brain_executor.py
  • Improved communication wrapper (communication.py, zenoh_comm.py)

3.15 Functional module layer (evospikenet/functional_modules.py)

  • Check the implementation and management mechanism of VisualModule / AuditoryModule / SpeechGenerationModule / WordEmbeddingLayer / TAS_Encoder / FunctionalModuleManager.
  • docstring TODO (TODO: document this function/class) has been resolved.
  • EdgeDetector.forward() has been updated to Sobel-based edge extraction.
  • MFCCExtractor.forward() has been updated to STFT + DCT based MFCC approximate extraction.
  • AuditoryModule has already implemented a 1D Conv-based audio encoder.
  • SpeechGenerationModule / WaveformSynthesizer has implemented ConvTranspose1d-based waveform decoding path.
  • The "In a real implementation ..." note in functional_modules.py has been resolved.
  • MilvusRetriever uses Milvus search if there is a dependency such as EVOSPIKENET_MILVUS_URI, and uses local fallback retrieval if not set.
  • Additional information:
  • pass is an abstract method declaration of the abstract base class FunctionalModuleBase.forward(), and is not a missing bug.
  • The Milvus connection path and failure fallback path of MilvusRetriever have been verified in tests/unit/test_functional_modules_milvus.py.

3.16 Embodied PLA Loop (evospikenet/embodied_pla_loop.py)

  • Cleaned up duplicate stub definitions of PerceptualLanguageEncoder / LanguageActionDecoder and enabled only the main implementation.
  • Implemented SNN-based visual processing in Rank1-9 (integrated SNNConv2d/LIF approximation processing).
  • Integrate SpikingLanguageDecoder as a voice/NLP production route and implement hybrid inference with rule base.
  • Connect inverse kinematics/cerebellar coordination/PWM transformation with TrajectoryPlanner / CerebellumCoordinator.
  • Introduced _RobotInterface as an abstraction for real robot cooperation, and implemented _MockRobotInterface / _ROSRobotInterface.
  • Implement forward kinematics-based current position estimation and ensure control loop continuity by treating idle actions as no-op successes.
  • Corrected search for vocab_table.json to prioritize package relative path (resolves Docker fixed path dependency).
  • Verification:
  • Fixed a closed loop hang in tests/unit/test_embodied_pla_loop.py.
  • Added the following new validation to the same file:
    • SNN visual Rank1 output verification
    • Voice/NLP integrated decoder verification
    • Inverse kinematics → cerebellar coordination → PWM flow verification
    • Robot IF mode switching verification
  • Execution result: 20 passed with pytest tests/unit/test_embodied_pla_loop.py tests/unit/test_embodied_pla_phase_modules.py -q -o addopts=''.

3.17 Distributed facade layer (evospikenet/distributed.py)

  • Expand the lightweight facade for research to a dependent lightweight facade for actual operations.
  • Implemented the following in ZenohCommunicator:
  • send_message / receive_message
  • publish / subscribe
  • encode_message / decode_message
  • close
  • Implemented node ledger and health management in NodeDiscoveryService:
  • register_node / discover_nodes
  • update_node_status / heartbeat
  • get_failed_nodes
  • Implemented real load based distribution in DistributedLoadBalancer:
  • register_node / assign_task
  • mark_task_done / handle_node_failure
  • Implemented snapshot generation and restoration in SnapshotManager:
  • create_snapshot
  • restore_from_snapshot
  • Improved DistributedBrainNetwork from random output to stable and reproducible inference (shape normalization + deterministic projection + node number gain).
  • Improved communication/computation cost model of DistributedManager (from simple proportional model to non-linear scaling).
  • Verification:
  • tests/unit/test_distributed_facade.py New addition
  • Execution result: pytest tests/unit/test_distributed_facade.py -q -o addopts='' passed

3.18 LLM backend learning hook (evospikenet/llm_backend.py)

  • Enhanced train() to an executable hook while maintaining the inference path (CPU retry when device mismatch).
  • Enhanced implementation of train():
  • dataset_path existence verification
  • Learning script solution with trainer=spiking|standard
  • CLI argument assembly (epochs / batch_size / learning_rate etc.)
  • dry_run support
  • subprocess Return summary of execution results (exit code, stdout/stderr tail)
  • timeout processing (exit code 124)
  • This makes it possible to safely determine whether learning can be started from the SDK/API while maintaining the premise of external script operation.
  • Verification:
  • tests/unit/test_llm_backend.py New addition
  • Execution result: pytest tests/unit/test_llm_backend.py -q -o addopts='' passed

3.19 Foundation extensions (evospikenet/foundation_extensions.py etc.)

  • Bias mitigation operationalization:
  • Implemented BiasMitigationEngine / BiasMitigationResult.
  • Implemented calculation of Demographic parity / Equal opportunity gap and lightweight reweighing when the threshold is exceeded.
  • Operationalization of ethical evaluation:
  • Implemented EthicsComplianceEvaluator / EthicsEvaluationResult.
  • Implemented operational gates for safety, privacy, fairness, and human override.
  • Large-scale multi-agent collaboration:
  • Implemented ScalableMultiAgentCoordinator.
  • Implemented experience accumulation, aggregation, and recommendation through shard distribution.
  • Olfactory/Taste/Biological Sensor Integration:
  • Implemented MultiModalSensorData / MultiSensoryToBrainLanguageAdapter.
  • Implemented fusion vectorization of tactile existing route and olfactory/gustatory/biosignal.
  • API extensions:
  • /api/foundation/bias-mitigation/evaluate
  • /api/foundation/ethics/evaluate
  • /api/foundation/social/scale
  • Verification:
  • Added unit test for bias/ethics/scalable social to tests/unit/test_foundation_extensions.py
  • Added new API integration test to tests/integration/test_foundation_api_integration.py
  • Added multisensory adapter test to tests/unit/test_tactile_to_language.py

3.20 Language binding/decompilation infrastructure (sdk_* / brain_language_decoder.py)

  • Added automatic generation of SDK distribution templates for each language:
  • export_go_sdk_package() (go.mod generation)
  • export_ts_sdk_package() (package.json / dist/* generation)
  • export_swift_sdk_package() (Package.swift / Sources / Tests generation)
  • Added unit test for Swift proxy:
  • tests/unit/test_sdk_swift_stub.py
  • Strengthen Brain Language decompilation operation evaluation:
  • BrainLanguageDecoder.integrate_pretrained_artifacts()
  • BrainLanguageDecoder.evaluate_dataset()
  • Verification:
  • Added package export test to tests/unit/test_sdk_stubs.py
  • Pretrained integration + dataset evaluation test added to tests/unit/test_brain_language_decoder.py

3.21 GPU/CPU device consistency fix (2026-04-27)

3.21.1 AEG device alignment (BUG-CPU-001)

  • Target: evospikenet/control.py — AEG class
  • Problem: Four state tensors in __init__ were defined without register_buffer, so moving the module with .to(device) / .cuda() did not move the tensors to the device and caused a device mismatch error. Also, in some cases, calling reset_statistics() before update() resulted in an AttributeError.
  • Correction details:
  • self.energy(torch.ones(num_neurons)) → register_buffer("energy", ...)
  • self.load_history (torch.zeros(100)) → register_buffer("load_history", ...)
  • self.total_spikes_processed (Python int) → register_buffer("total_spikes_processed", torch.zeros(1, dtype=torch.long))
  • self.total_energy_saved (Python float) → register_buffer("total_energy_saved", torch.zeros(1))
  • Changed the comparison in get_power_statistics() to if self.total_spikes_processed > 0: → .item() > 0 (tensor bool conversion UserWarning resolved)
  • Verification:
  • Add the following to tests/unit/test_fail_closed_regressions.py:
    • test_aeg_buffers_move_with_module_to_cpu
    • test_aeg_buffers_move_with_module_to_cuda (run only when CUDA is available)
    • test_aeg_reset_statistics_before_any_update
    • test_aeg_get_power_statistics_before_any_update
  • CPU (CUDA_VISIBLE_DEVICES=''): 17 passed, 2 skipped / GPU (CUDA_VISIBLE_DEVICES=0): 19 passed

3.21.2 ClosedLoopController Device automatic movement (BUG-CPU-002)

  • Target: evospikenet/control.py — ClosedLoopController.forward()
  • Problem: Even if the model was moved to the GPU, the input tensors (sensor_input, target) and internal buffers (integral, prev_error, spike_buffer) remained on the CPU, causing a device mismatch error.
  • Correction details:
  • Get model_device = next(self.controller_net.parameters()).device at the beginning of forward() and automatically move sensor_input / target / 3 internal buffers with .to(model_device).
  • Verification:
  • Add the following to tests/unit/test_fail_closed_regressions.py:
    • test_closed_loop_controller_cpu_gpu_input_alignment
    • test_closed_loop_controller_moves_cpu_inputs_to_cuda (executed only when CUDA is available)
  • All test results: CPU 17 passed / GPU 25 passed (including existing tests)

3.21.3 SDK-URL Replace all files at once (Sprint-2 SDK-URL)

  • Target: 40+ files under future_apps/ in the entire repository
  • Contents: tools/replace_sdk_url.py has been executed to replace hard-coded URLs such as http://localhost:8000 with environment variable references (such as EVOSPIKENET_SDK_URL).
  • Status: Confirm no hits (excluding documentation and comments) with grep -R "http://localhost:8000" ..

3.21.4 AvailabilityMonitor delayed initialization (MONITOR-001)

  • Target: evospikenet/availability_monitor.py — AvailabilityMonitor
  • Contents:
  • autostart=False By default, the monitoring thread does not start automatically when importing.
  • EVOSPIKENET_AVAIL_MONITOR_LAZY=true Delay can also be controlled by environment variable.
  • Health endpoint setting (FALLBACK-004) using the EVOSPIKENET_HEALTH_URL environment variable has also been implemented.
  • Status: Source verified and tested.

3.21.5 Evaluate_System.md Source Controlled Audit

  • Corrected all statuses (✅ / ❌ / 🔄) of Evaluate_System.md against source code:
  • Sprint-2 SDK-URL: ❌ → ✅
  • Sprint-3 MONITOR-001: ❌ → ✅
  • BUG-005 (hidden_size ONNX): ✅ → 🔄 (TensorRT path is a GPU-only class, so .cuda() is appropriate and is a partial response rather than a complete modification)
  • BUG-CPU-001 / BUG-CPU-002: Added new entry
  • Newly added Section 9 “GPU/CPU Device Integrity Verification Summary”

3.22 Evaluate_System.md Implementation of all unsupported items (2026-04-27)

3.22.1 Unified alert module (ALERT-001)

  • Target: evospikenet/alerting.py (new creation)
  • Contents:
  • Implemented a public API that unified 3 channels: Slack Incoming Webhook / PagerDuty Events API v2 / SMTP email.
  • send_alert(message, level, component, details) → Dict[str, bool]
  • send_fallback_alert(component, reason, details) — Fallback canned alert
  • send_critical_alert(component, message, details) — Critical failure alert
  • Enabling each channel is controlled by environment variables:
    • EVOSPIKENET_SLACK_WEBHOOK_URL — Slack Webhook URL
    • EVOSPIKENET_PAGERDUTY_ROUTING_KEY — PagerDuty routing key
    • EVOSPIKENET_ALERT_EMAIL_TO / EVOSPIKENET_ALERT_EMAIL_FROM / EVOSPIKENET_ALERT_SMTP_HOST / EVOSPIKENET_ALERT_SMTP_PORT — SMTP settings
    • EVOSPIKENET_ALERT_CHANNELS — Explicit specification of valid channels (comma separated)
  • Fallback to logger.warning (does not throw exception) when channel is not set.
  • Verification: python -c "from evospikenet.alerting import ..." has confirmed that all three functions are working normally (channel not set → {} return, logger.warning output).

3.22.2 SDK mock package (TEST-001)

  • Target: future_apps/tools/evospikenet_sdk_mock/ (new package)
  • Contents:
  • __init__.py — MockServer(port) class: start() / stop() / base_url properties
  • server.py — FastAPI app: /api/health / /v1/evolve / /v1/predict / /v1/spatial/generate / /v1/spatial/infer / /v1/sessions/{id} / Implement catch-all
  • conftest.py — pytest fixtures: evospikenet_mock_server (session scope) / evospikenet_mock_url
  • pyproject.toml — installable package, evospikenet-mock CLI entry point
  • tests/test_sdk_mock.py — 6 tests (import, check all routes, health, evolve, catch-all)
  • Usage: Test execution pointing to EVOSPIKENET_API_URL=http://localhost:18000 in a CI environment. Prevent Mock from being mixed into production code.
  • Verification: pytest future_apps/tools/evospikenet_sdk_mock/tests/ -q → 6 passed.

3.22.3 GPUAdapter TensorRT hidden_size fix (TYPE-001)

  • Target: evospikenet/universal_integration.py — GPUAdapter.convert_format()
  • Problem: Dummy input during TensorRT compilation is hardcoded to torch.randn(1, 768).cuda(), causing shape mismatch in models with hidden_size other than 768.
  • Correction details:python _hidden = 768 try: _hidden = int(getattr(getattr(model, 'config', None), 'hidden_size', _hidden)) except Exception: pass model = torch_tensorrt.compile(model, inputs=[torch.randn(1, _hidden).cuda()])- Status: Source corrected. BUG-005 / TYPE-001 is fully supported and Evaluate_System.md has been updated.

3.22.4 JobQueue migration warning (MOCK-001 remaining)

  • Target: evospikenet/video_analysis/job_queue.py — JobQueue.__init__
  • Contents: JobQueue is an in-memory only legacy class. Added logger.warning to prompt migration to the production equivalent VideoAnalysisJobStore (Redis / SQLite / JSON fallback supported).
  • Status: Updated MOCK-001 in Evaluate_System.md to ✅.

3.22.5 mineral_exploration task warning enhancement (tasks.py High)

  • Target: future_apps/mineral_exploration/workers/tasks.py
  • Problem: Even though preprocess_task / predict_task uses placeholder output (.tif text file, "0\n" file), quality degradation is not propagated downstream.
  • Correction details:
  • preprocess_task: Added logger.warning("[preprocess_task][job=%s] Using placeholder stack...") and added "synthetic": True / "warning": "..." fields to the result.
  • predict_task: Added logger.warning and "synthetic": True / "warning": "..." fields as well.
  • Status: API consumers can now check result["synthetic"] to detect data quality degradation.

3.22.6 Added IoU exception log (Low — video_scene_app)

  • Target: future_apps/video_scene_app/main.py — _iou() function
  • Contents: Changed except Exception block to except Exception as _e: and added logger.debug("_iou: failed to compute IoU for boxA=%s boxB=%s — %s", boxA, boxB, _e).
  • Condition: Easier debugging when division fails.

3.22.7 Translation Service Neural Engine Integration (Low — translator)

  • Target: future_apps/real_time_language_translation/src/services/translator.py
  • Problem: Although the app name is "real_time_language_translation", the implementation is only rule-based.
  • Correction details:
  • Added _load_neural_backend(src_lang, tgt_lang) function (lazy loading with global cache). Use MarianTokenizer + MarianMTModel of Helsinki-NLP/OPUS-MT.
  • Attempt neural translation if the EVOSPIKENET_TRANSLATION_BACKEND=neural environment variable is set in Translator.translate(), and fallback to rule-based on failure.
  • Added "engine": "neural" | "rule" field to response metadata.
  • Depends on: transformers Automatic fallback to rule base (no exceptions) if not installed.

3.22.8 Evaluate_System.md Reflect all status

  • Updated Evaluate_System.md with completion of implementation of 3.22.1 to 3.22.7 above:
  • Sprint-3: ALERT-001 / TEST-001 / CRED-001 ❌ → ✅
  • Sprint-4: TYPE-001 ❌ → ✅
  • Section 3 MOCK-001 Remaining items (job_queue.py) ❌ → ✅
  • Section 4 workers/tasks.py High ❌ → ✅
  • Section 4.14 IoU Low 🟢 → ✅
  • Section 4.15 translator.py Low 🟢 → ✅
  • BUG-005 / TYPE-001 (Section 8 Summary) 🔄 → ✅
  • Added v2 to the update note at the top of the document

3.23 4.1 Improved mock/placeholder detection points (2026-05-01)

3.23.1 Communication fallback fail-closed control

  • Target: evospikenet/communication.py
  • Contents:
  • Added EVOSPIKENET_FAIL_ON_COMM_FALLBACK.
  • Explicit communication backend state backend / is_fallback_backend.
  • Added warning log when in-memory fallback publish.
  • Status: Implemented.

3.23.2 Compatibility API fallback stricter

  • Target: evospikenet/api_modules/future_apps_compat_api.py
  • Contents:
  • Added EVOSPIKENET_ALLOW_COMPAT_FALLBACKS.
  • Returns 503 compat_fallback_disabled when fallback is not allowed in strict environment.
  • Add headers is_synthetic / data_provenance to fallback response.
  • Status: Implemented.

3.23.3 Abolition of video analysis placeholder string

  • Target: evospikenet/video_scene_service.py
  • Contents:
  • Removed TRANSCRIPT_PLACEHOLDER / NARRATIVE_GENERATION_FAILED dependency.
  • Explicit status with transcript_status / narrative_status and synthetic flag.
  • Make it fail-closed with EVOSPIKENET_ALLOW_MEDIA_PLACEHOLDERS=0.
  • Status: Implemented.

3.23.4 PFC placeholder implementation replacement

  • Target: evospikenet/pfc.py
  • Contents:
  • Replaced unconnected placeholder in remote backup with _backup_to_remote_site implementation.
  • Fixed leader task result Removed task_completed string and returned structured result.
  • Status: Implemented.

3.23.5 Visualization of time synchronization deterioration status

  • Target: evospikenet/eeg_integration/time_sync.py
  • Contents:
  • Added require_high_precision / allow_ntp_fallback_for_ptp.
  • Added sync_source / degraded / degraded_reason to status.
  • Changed NTP fallback when PTP is unavailable to be set to fail-closed.
  • Status: Implementation complete (type error resolved).

3.23.6 Operational control of dummy service fallback

  • Target: evospikenet/services/__init__.py
  • Contents:
  • Added EVOSPIKENET_ALLOW_DUMMY_SERVICES.
  • Prohibited dummy scene generation in strict environment and changed to throw an exception.
  • Status: Implemented.

3.23.7 ​​Test-status execution status

  • Target: evospikenet/api_modules/test_api.py
  • Contents:
  • Added TEST_EXECUTION_STATE.
  • Update status on start/completion/failure/timeout of /run-tests.
  • /test-status returns runtime state instead of placeholder.
  • Status: Implemented.

Reference documentation

  • docs-dev/connectome_evospikenet_implementation_policy.ja.md
  • docs-dev/connectome_schema.md
  • docs/DISTRIBUTED_BRAIN_SPATIAL_NODES.md
  • docs/SDK_API_REFERENCE.md
  • Docs/VIDEO_AUDIO_ANALYSIS_SPEC.md

This reorganized version is an operational ledger that has integrated duplicate sections and corrected the status notation to comply with the source code.