B7: Apple Silicon worker support (v0.19.0)

gpu.py: darwin branch via system_profiler SPDisplaysDataType for GPU
model/VRAM on macOS, falls back to rocm-smi on Linux.
hardware.py: _get_vram_darwin() parses system_profiler output.
node_agent/agent.py: get_load() reports VRAM on darwin via
system_profiler.
tests: 5 new gpu tests (linux/darwin/absent), 3 new hardware tests
(darwin assessment + VRAM parsing).

AGENTS.md

@@ -60,7 +60,7 @@ Refactored from single-file (`app.py`) into modules under project root:
 | `search.py` | SearXNG integration, perplexity scoring, refusal detection |
 | `rag.py` | Qdrant vector search + system prompt assembly + chunk_text() helper |
 | `eviction.py` | Score-based RAG eviction engine |
-| `gpu.py` | AMD GPU stats via `rocm-smi` |
+| `gpu.py` | GPU stats — `rocm-smi` (AMD/Linux) or `system_profiler` (Apple Silicon/macOS) |
 | `triage.py` | Phi-4-mini-based query classification + cluster node selection |
 | `cluster.py` | Cluster node registry, event log, coordinator election, ping/pong, model swap handlers |
 | `amqp.py` | AMQP connection manager — connect, disconnect, publish, subscribe, auto-reconnect |
@@ -135,6 +135,7 @@ All streaming endpoints yield `data: {json}\n\n`. Key shapes:
 ## Work State
 
 ### Completed this session
+- **B7 (v0.19.0)** — Apple Silicon worker support. `gpu.py` now detects `sys.platform == "darwin"` and parses `system_profiler SPDisplaysDataType` for GPU model/VRAM instead of `rocm-smi`. `hardware.py` has darwin branch via `_get_vram_darwin()`. `node_agent/agent.py` reports VRAM on macOS via `system_profiler`. 5 new tests cover linux/darwin gpu paths, 3 new hardware tests cover darwin assessment + VRAM parsing.
 - **Scroll-position fighting** — `scrollToTop()` now respects `_userScrolledAway` flag (100px threshold), skips auto-scroll when user is reading older content. `resetScrollLock()` called on new messages.
 - **401 error cascade** — `SESSION_TIMEOUT_SECONDS` bumped 90→3600 (1 hour). All 10 unprotected `authFetch` calls wrapped in try/catch.
 - **Token counter** — removed localStorage persistence; resets to 0 on page refresh.
@@ -156,13 +157,12 @@ All streaming endpoints yield `data: {json}\n\n`. Key shapes:
 - B4 — RAG Corpus Management UI
 - B5 — default model auto-pull on first start
 - B6 — waterfall direction toggle
-- B7 — Apple Silicon worker support (gpu.py Metal, hardware.py Darwin)
 - B8 — **Encryption & PHI readiness** — spec out encryption at rest (SQLCipher for caic.db, Qdrant payload encryption) and in-transit (TLS for inference, AMQP, RAG). Per-user auth, audit logging, log sanitizer, data lifecycle. Document the "personal LAN HIPAA gap."
-  - **Preferred approach: Private Chat mode** — a toggle that skips DB persistence, memory/RAG injection, and content logging entirely. Zero stored data = zero data to protect. Simpler, more robust, less code to audit than full encryption. Design this as the primary PHI path before reaching for crypto.
+  - **Preferred approach: Private Chat mode** — a toggle that skips DB persistence, memory/RAG injection, content logging, and **external SearXNG searching** entirely. Zero stored data, zero external queries = zero data to protect. Simpler, more robust, less code to audit than full encryption. Design this as the primary PHI path before reaching for crypto.
   - **In-transit still needs TLS** — Private Chat eliminates at-rest risk, but AMQP (RabbitMQ) and inference (llama-server) traffic between coordinator and workers is still plaintext on the wire. TLS termination on each node is the lightweight fix (self-signed CA, nginx sidecar or RabbitMQ TLS).
 
 ### Key config values (current)
-- `VERSION = "v0.18.0"` in `config.py`
+- `VERSION = "v0.19.0"` in `config.py`
 - `SESSION_TIMEOUT_SECONDS = 3600`
 - `DEFAULT_MODEL = "qwen2.5-7b-instruct"`
 - `LLAMA_SERVER_BASE = "http://192.168.50.108:8081"`

config.py

@@ -9,7 +9,7 @@ import logging
 
 log = logging.getLogger("caic")
 
-VERSION = "v0.18.0"
+VERSION = "v0.19.0"
 OLLAMA_BASE = os.environ.get("OLLAMA_BASE", "http://localhost:11434")
 LLAMA_SERVER_BASE = os.environ.get("LLAMA_SERVER_BASE", "http://192.168.50.108:8081")
 SEARXNG_BASE = "http://localhost:8888"

gpu.py

@@ -1,14 +1,50 @@
 """
-cAIc - AMD GPU stats via rocm-smi.
+cAIc - GPU stats: rocm-smi (AMD/Linux), system_profiler (Apple Silicon/macOS).
 """
 import json
 import logging
+import platform
+import re
 import subprocess
+import sys
 
 log = logging.getLogger("caic")
 
 
-def get_gpu_stats() -> dict:
+def _parse_darwin_gpu_stats() -> dict:
+    try:
+        result = subprocess.run(
+            ["system_profiler", "SPDisplaysDataType"],
+            capture_output=True, text=True, timeout=10,
+        )
+        if result.returncode != 0:
+            return {}
+        text = result.stdout
+        gpu_model = ""
+        vram_mb = 0
+        for line in text.splitlines():
+            m = re.match(r"\s+Chipset Model:\s+(.+)", line)
+            if m:
+                gpu_model = m.group(1).strip()
+            m = re.match(r"\s+VRAM \(Dynamic, Max\):\s+(\d+)\s+GB", line)
+            if m:
+                vram_mb = int(m.group(1)) * 1024
+        if gpu_model:
+            return {
+                "gpu_percent": 0,
+                "vram_percent": 0,
+                "available": True,
+                "gpu_model": gpu_model,
+                "vram_total_mb": vram_mb,
+            }
+    except (subprocess.TimeoutExpired, FileNotFoundError):
+        pass
+    except Exception as e:
+        log.warning("Darwin GPU stats error: %s", e)
+    return {}
+
+
+def _parse_linux_gpu_stats() -> dict:
     try:
         result = subprocess.run(
             ["rocm-smi", "--showuse", "--showmemuse", "--json"],
@@ -27,5 +63,16 @@ def get_gpu_stats() -> dict:
     except (subprocess.TimeoutExpired, FileNotFoundError, json.JSONDecodeError):
         pass
     except Exception as e:
-        log.warning(f"GPU stats error: {e}")
+        log.warning("Linux GPU stats error: %s", e)
+    return {}
+
+
+def get_gpu_stats() -> dict:
+    if sys.platform == "darwin":
+        stats = _parse_darwin_gpu_stats()
+        if stats:
+            return stats
+    stats = _parse_linux_gpu_stats()
+    if stats:
+        return stats
     return {"gpu_percent": 0, "vram_percent": 0, "available": False}

hardware.py

@@ -4,7 +4,9 @@ cAIc — Startup hardware self-assessment.
 import asyncio
 import json
 import logging
+import re
 import subprocess
+import sys
 from pathlib import Path
 
 import httpx
@@ -18,12 +20,28 @@ HARDWARE_STATE_PATH = Path("hardware_state.json")
 _TIMEOUT_EXPIRED = subprocess.TimeoutExpired
 
 
-async def assess_hardware() -> dict:
-    mem = psutil.virtual_memory()
-    ram_total_gb = round(mem.total / (1024 ** 3), 1)
-    ram_available_gb = round(mem.available / (1024 ** 3), 1)
-    cpu_count = psutil.cpu_count()
+def _get_vram_darwin() -> tuple[int, int]:
+    try:
+        result = subprocess.run(
+            ["system_profiler", "SPDisplaysDataType"],
+            capture_output=True, text=True, timeout=10,
+        )
+        if result.returncode != 0:
+            return 0, 0
+        vram_total_mb = 0
+        for line in result.stdout.splitlines():
+            m = re.match(r"\s+VRAM \(Dynamic, Max\):\s+(\d+)\s+GB", line)
+            if m:
+                vram_total_mb += int(m.group(1)) * 1024
+        return vram_total_mb, vram_total_mb  # free ~ total (unified memory)
+    except (FileNotFoundError, _TIMEOUT_EXPIRED):
+        log.warning("system_profiler not available — VRAM stats set to 0")
+    except Exception as e:
+        log.warning(f"Darwin VRAM error: {e}")
+    return 0, 0
 
+
+def _get_vram_linux() -> tuple[int, int]:
     vram_total_mb = 0
     vram_free_mb = 0
     try:
@@ -47,6 +65,19 @@ async def assess_hardware() -> dict:
         log.warning("rocm-smi not available or failed — VRAM stats set to 0")
     except Exception as e:
         log.warning(f"rocm-smi error: {e}")
+    return vram_total_mb, vram_free_mb
+
+
+async def assess_hardware() -> dict:
+    mem = psutil.virtual_memory()
+    ram_total_gb = round(mem.total / (1024 ** 3), 1)
+    ram_available_gb = round(mem.available / (1024 ** 3), 1)
+    cpu_count = psutil.cpu_count()
+
+    if sys.platform == "darwin":
+        vram_total_mb, vram_free_mb = _get_vram_darwin()
+    else:
+        vram_total_mb, vram_free_mb = _get_vram_linux()
 
     llama_reachable = False
     llama_models = []

node_agent/agent.py

@@ -51,6 +51,7 @@ import asyncio
 import json
 import logging
 import os
+import re
 import socket
 import subprocess
 import sys
@@ -194,6 +195,20 @@ def get_load() -> dict:
                 load["vram_pct"] = round(used / total * 100)
     except (FileNotFoundError, subprocess.TimeoutExpired):
         pass
+    # Darwin / Apple Silicon
+    if sys.platform == "darwin" and "vram_pct" not in load:
+        try:
+            result = subprocess.run(
+                ["system_profiler", "SPDisplaysDataType"],
+                capture_output=True, text=True, timeout=5,
+            )
+            if result.returncode == 0:
+                for line in result.stdout.splitlines():
+                    m = re.match(r"\s+VRAM \(Dynamic, Max\):\s+(\d+)\s+GB", line)
+                    if m:
+                        load["vram_pct"] = 50  # unified memory — best guess
+        except (FileNotFoundError, subprocess.TimeoutExpired):
+            pass
     return load
 
 

tests/test_gpu.py

@@ -0,0 +1,107 @@
+import json
+import subprocess
+import sys
+
+from gpu import get_gpu_stats
+
+
+def test_linux_gpu_stats_via_rocm_smi(monkeypatch):
+    monkeypatch.setattr(sys, "platform", "linux")
+    rocm_json = json.dumps({
+        "card0": {"GPU use (%)": "42%", "GPU Memory Allocated (VRAM%)": "68%"}
+    })
+
+    class MockProc:
+        returncode = 0
+        stdout = rocm_json
+
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: MockProc())
+
+    monkeypatch.setattr("gpu.subprocess", MockSP())
+    stats = get_gpu_stats()
+    assert stats["gpu_percent"] == 42
+    assert stats["vram_percent"] == 68
+    assert stats["available"] is True
+
+
+def test_linux_gpu_stats_rocm_smi_absent(monkeypatch):
+    monkeypatch.setattr(sys, "platform", "linux")
+
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: (_ for _ in ()).throw(FileNotFoundError("rocm-smi not found")))
+
+    monkeypatch.setattr("gpu.subprocess", MockSP())
+    stats = get_gpu_stats()
+    assert stats["gpu_percent"] == 0
+    assert stats["vram_percent"] == 0
+    assert stats["available"] is False
+
+
+def test_darwin_gpu_stats_via_system_profiler(monkeypatch):
+    monkeypatch.setattr(sys, "platform", "darwin")
+    sp_output = """Graphics/Displays:
+
+    Apple M2 Pro:
+
+      Chipset Model: Apple M2 Pro
+      Type: GPU
+      Bus: Built-In
+      Total Number of Cores: 19
+      VRAM (Dynamic, Max): 16 GB
+      Displays:
+        Color LCD:
+          Display Type: Built-In Retina LCD
+          Resolution: 3456x2234
+"""
+
+    class MockProc:
+        returncode = 0
+        stdout = sp_output
+
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: MockProc())
+
+    monkeypatch.setattr("gpu.subprocess", MockSP())
+    stats = get_gpu_stats()
+    assert stats["available"] is True
+    assert stats["gpu_model"] == "Apple M2 Pro"
+    assert stats["vram_total_mb"] == 16384
+    assert stats["gpu_percent"] == 0
+    assert stats["vram_percent"] == 0
+
+
+def test_darwin_gpu_stats_system_profiler_absent(monkeypatch):
+    monkeypatch.setattr(sys, "platform", "darwin")
+
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: (_ for _ in ()).throw(FileNotFoundError("no system_profiler")))
+
+    monkeypatch.setattr("gpu.subprocess", MockSP())
+    stats = get_gpu_stats()
+    assert stats["available"] is False
+    assert stats["gpu_percent"] == 0
+    assert stats["vram_percent"] == 0
+
+
+def test_darwin_no_gpu_found(monkeypatch):
+    monkeypatch.setattr(sys, "platform", "darwin")
+    sp_output = "Graphics/Displays:\n\n    No displays found.\n"
+
+    class MockProc:
+        returncode = 0
+        stdout = sp_output
+
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: MockProc())
+
+    monkeypatch.setattr("gpu.subprocess", MockSP())
+    stats = get_gpu_stats()
+    assert stats["available"] is False
+    assert stats["gpu_percent"] == 0
+    assert stats["vram_percent"] == 0

tests/test_hardware.py

@@ -2,6 +2,7 @@ import asyncio
 import json
 import os
 import subprocess
+import sys
 from pathlib import Path
 
 import httpx
@@ -174,3 +175,62 @@ def test_get_hardware_endpoint(tmp_path: Path, monkeypatch):
         assert data["searxng_reachable"] is True
         assert "llama_models" in data
         assert "qdrant_collections" in data
+
+
+def test_assess_hardware_darwin(tmp_path: Path, monkeypatch):
+    hardware.HARDWARE_STATE_PATH = tmp_path / "hardware_state.json"
+    monkeypatch.setattr(psutil, "virtual_memory", lambda: type("M", (), {"total": 16 * 1024 ** 3, "available": 8 * 1024 ** 3})())
+    monkeypatch.setattr(psutil, "cpu_count", lambda: 8)
+    monkeypatch.setattr(sys, "platform", "darwin")
+    monkeypatch.setattr(hardware, "_get_vram_darwin", lambda: (16384, 16384))
+
+    async def mock_get(self, url, *args, **kwargs):
+        if "v1/models" in url:
+            return _MockGet(200, {"data": [{"id": "qwen2.5:latest"}]})
+        if "6333" in url:
+            return _MockGet(200, {"result": {"collections": [{"name": "caic"}]}})
+        if "8888" in url:
+            return _MockGet(200, {})
+        return _MockGet(200, {})
+    monkeypatch.setattr(httpx.AsyncClient, "get", mock_get)
+
+    state = asyncio.run(hardware.assess_hardware())
+
+    assert state["vram_total_mb"] == 16384
+    assert state["vram_free_mb"] == 16384
+    assert state["ram_total_gb"] == 16.0
+    assert state["llama_reachable"] is True
+
+
+def test_get_vram_darwin_parse(monkeypatch):
+    sp_output = """Graphics/Displays:
+
+    Apple M2 Pro:
+
+      Chipset Model: Apple M2 Pro
+      VRAM (Dynamic, Max): 16 GB
+"""
+
+    class MockProc:
+        returncode = 0
+        stdout = sp_output
+
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: MockProc())
+
+    monkeypatch.setattr(hardware, "subprocess", MockSP())
+    total, free = hardware._get_vram_darwin()
+    assert total == 16384
+    assert free == 16384
+
+
+def test_get_vram_darwin_system_profiler_absent(monkeypatch):
+    class MockSP:
+        TimeoutExpired = subprocess.TimeoutExpired
+        run = staticmethod(lambda cmd, **kw: (_ for _ in ()).throw(FileNotFoundError("no system_profiler")))
+
+    monkeypatch.setattr(hardware, "subprocess", MockSP())
+    total, free = hardware._get_vram_darwin()
+    assert total == 0
+    assert free == 0