Create ARCHITECTURE.md

Author: BryceWDesign

docs/ARCHITECTURE.md

@@ -0,0 +1,42 @@
+# Architecture (SynapDrive-AI)
+
+## Purpose
+SynapDrive-AI is a **simulation-first** reference implementation of an intent-to-action pipeline:
+input → decode/intent → context optimization → safety gate → actuation → evaluation → memory.
+
+## Canonical wiring
+The single source of truth is:
+- `synapdrive_ai/pipeline.py`
+
+## Data contracts
+### Intent packet (minimum)
+- `intent: str`
+- `confidence: float (0..1)`
+- `source: str`
+- `memory_context: list`
+
+### Result packet (normalized)
+Produced by `DecisionRouter.route()`:
+- `status: "success" | "failed"`
+- `intent: str`
+- `confidence: float`
+- `duration: float`
+- `raw_status: str`
+
+### Telemetry log entry (contract)
+Produced by `ActuationEngine`:
+- `timestamp`
+- `intent`
+- `confidence`
+- `status`
+- `duration`
+- `source`
+- `memory`
+- `memory_context`
+
+## Extensibility
+Integrations are intentionally optional:
+- `synapdrive_ai/integrations/brainflow_adapter.py` (BrainFlow)
+- `synapdrive_ai/integrations/lsl_adapter.py` (pylsl / LSL)
+
+They both output an intent packet and then call the canonical pipeline.