fix: include correct __version__ in python packages

hindsight-api/hindsight_api/__init__.py

@@ -46,4 +46,4 @@ __all__ = [
     "RemoteTEICrossEncoder",
     "LLMConfig",
 ]
-__version__ = "0.1.0"
+__version__ = "0.4.0"

hindsight-api/hindsight_api/engine/cross_encoder.py

@@ -137,15 +137,26 @@ class LocalSTCrossEncoder(CrossEncoderModel):
         # which can cause issues when accelerate is installed but no GPU is available.
         # Note: We do NOT use device_map because CrossEncoder internally calls .to(device)
         # after loading, which conflicts with accelerate's device_map handling.
+        import os
+
         import torch
 
-        # Check for GPU (CUDA) or Apple Silicon (MPS)
-        has_gpu = torch.cuda.is_available() or (hasattr(torch.backends, "mps") and torch.backends.mps.is_available())
+        # Force CPU mode if HINDSIGHT_FORCE_CPU is set (used in daemon mode to avoid MPS/XPC issues)
+        force_cpu = os.getenv("HINDSIGHT_FORCE_CPU", "0") == "1"
 
-        if has_gpu:
-            device = None  # Let sentence-transformers auto-detect GPU/MPS
-        else:
+        if force_cpu:
             device = "cpu"
+            logger.info("Reranker: forcing CPU mode (HINDSIGHT_FORCE_CPU=1)")
+        else:
+            # Check for GPU (CUDA) or Apple Silicon (MPS)
+            has_gpu = torch.cuda.is_available() or (
+                hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
+            )
+
+            if has_gpu:
+                device = None  # Let sentence-transformers auto-detect GPU/MPS
+            else:
+                device = "cpu"
 
         self._model = CrossEncoder(
             self.model_name,
@@ -211,12 +222,21 @@ class LocalSTCrossEncoder(CrossEncoderModel):
             )
 
         # Determine device based on hardware availability
-        has_gpu = torch.cuda.is_available() or (hasattr(torch.backends, "mps") and torch.backends.mps.is_available())
+        import os
 
-        if has_gpu:
-            device = None  # Let sentence-transformers auto-detect GPU/MPS
-        else:
+        force_cpu = os.getenv("HINDSIGHT_FORCE_CPU", "0") == "1"
+
+        if force_cpu:
             device = "cpu"
+        else:
+            has_gpu = torch.cuda.is_available() or (
+                hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
+            )
+
+            if has_gpu:
+                device = None  # Let sentence-transformers auto-detect GPU/MPS
+            else:
+                device = "cpu"
 
         self._model = CrossEncoder(
             self.model_name,

hindsight-api/hindsight_api/engine/embeddings.py

@@ -132,15 +132,26 @@ class LocalSTEmbeddings(Embeddings):
         # Determine device based on hardware availability.
         # We always set low_cpu_mem_usage=False to prevent lazy loading (meta tensors)
         # which can cause issues when accelerate is installed but no GPU is available.
+        import os
+
         import torch
 
-        # Check for GPU (CUDA) or Apple Silicon (MPS)
-        has_gpu = torch.cuda.is_available() or (hasattr(torch.backends, "mps") and torch.backends.mps.is_available())
+        # Force CPU mode if HINDSIGHT_FORCE_CPU is set (used in daemon mode to avoid MPS/XPC issues)
+        force_cpu = os.getenv("HINDSIGHT_FORCE_CPU", "0") == "1"
 
-        if has_gpu:
-            device = None  # Let sentence-transformers auto-detect GPU/MPS
-        else:
+        if force_cpu:
             device = "cpu"
+            logger.info("Embeddings: forcing CPU mode (HINDSIGHT_FORCE_CPU=1)")
+        else:
+            # Check for GPU (CUDA) or Apple Silicon (MPS)
+            has_gpu = torch.cuda.is_available() or (
+                hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
+            )
+
+            if has_gpu:
+                device = None  # Let sentence-transformers auto-detect GPU/MPS
+            else:
+                device = "cpu"
 
         self._model = SentenceTransformer(
             self.model_name,
@@ -199,12 +210,21 @@ class LocalSTEmbeddings(Embeddings):
             )
 
         # Determine device based on hardware availability
-        has_gpu = torch.cuda.is_available() or (hasattr(torch.backends, "mps") and torch.backends.mps.is_available())
+        import os
 
-        if has_gpu:
-            device = None  # Let sentence-transformers auto-detect GPU/MPS
-        else:
+        force_cpu = os.getenv("HINDSIGHT_FORCE_CPU", "0") == "1"
+
+        if force_cpu:
             device = "cpu"
+        else:
+            has_gpu = torch.cuda.is_available() or (
+                hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
+            )
+
+            if has_gpu:
+                device = None  # Let sentence-transformers auto-detect GPU/MPS
+            else:
+                device = "cpu"
 
         self._model = SentenceTransformer(
             self.model_name,

hindsight-api/hindsight_api/main.py

@@ -140,6 +140,12 @@ def main():
         args.port = DEFAULT_DAEMON_PORT
         args.host = "127.0.0.1"  # Only bind to localhost for security
 
+        # Force CPU mode for daemon to avoid macOS MPS/XPC issues
+        # MPS (Metal Performance Shaders) has unstable XPC connections in background processes
+        # that can cause assertion failures and process crashes at the C++ level
+        # (which Python exception handlers cannot catch)
+        os.environ["HINDSIGHT_FORCE_CPU"] = "1"
+
         # Check if another daemon is already running
         daemon_lock = DaemonLock()
         if not daemon_lock.acquire():

scripts/release.sh

@@ -77,6 +77,22 @@ for package in "${PYTHON_PACKAGES[@]}"; do
     fi
 done
 
+# Update __version__ in Python __init__.py files
+PYTHON_INIT_FILES=(
+    "hindsight-api/hindsight_api/__init__.py"
+    "hindsight-embed/hindsight_embed/__init__.py"
+    "hindsight-clients/python/hindsight_client_api/__init__.py"
+)
+for init_file in "${PYTHON_INIT_FILES[@]}"; do
+    if [ -f "$init_file" ]; then
+        print_info "Updating __version__ in $init_file"
+        sed -i.bak "s/^__version__ = \".*\"/__version__ = \"$VERSION\"/" "$init_file"
+        rm "${init_file}.bak"
+    else
+        print_warn "File $init_file not found, skipping"
+    fi
+done
+
 # Update Rust CLI
 CARGO_FILE="hindsight-cli/Cargo.toml"
 if [ -f "$CARGO_FILE" ]; then

uv.lock

@@ -1295,7 +1295,7 @@ wheels = [
 
 [[package]]
 name = "hindsight-all"
-version = "0.3.0"
+version = "0.4.0"
 source = { editable = "hindsight" }
 dependencies = [
     { name = "hindsight-api" },
@@ -1319,7 +1319,7 @@ provides-extras = ["test"]
 
 [[package]]
 name = "hindsight-api"
-version = "0.3.0"
+version = "0.4.0"
 source = { editable = "hindsight-api" }
 dependencies = [
     { name = "aiohttp" },
@@ -1447,7 +1447,7 @@ dev = [
 
 [[package]]
 name = "hindsight-client"
-version = "0.3.0"
+version = "0.4.0"
 source = { editable = "hindsight-clients/python" }
 dependencies = [
     { name = "aiohttp" },
@@ -1481,7 +1481,7 @@ provides-extras = ["test"]
 
 [[package]]
 name = "hindsight-dev"
-version = "0.3.0"
+version = "0.4.0"
 source = { editable = "hindsight-dev" }
 dependencies = [
     { name = "hindsight-api" },
@@ -1527,7 +1527,7 @@ dev = [
 
 [[package]]
 name = "hindsight-embed"
-version = "0.3.0"
+version = "0.4.0"
 source = { editable = "hindsight-embed" }
 dependencies = [
     { name = "httpx" },