Persist input history across runs

Load and save readline history in .cache/history (last 1000 entries).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

.gitignore

@@ -1,2 +1,3 @@
 .venv/
+__pycache__/
 .cache/

embedlab.py

@@ -2,6 +2,7 @@
 
 print("Loading torch/transformers, please wait...", flush=True)
 
+import atexit
 import os
 import readline
 
@@ -22,6 +23,7 @@ MODELS = {
 PARAM_KEYS = ["pooling", "max_length", "batch_size"]
 DEFAULT_MODEL = "BAAI/bge-small-zh-v1.5"
 LAST_MODEL_FILE = os.path.join(os.path.dirname(os.path.abspath(__file__)), ".cache", "last_model")
+HISTORY_FILE = os.path.join(os.path.dirname(os.path.abspath(__file__)), ".cache", "history")
 COMMANDS = ["/select", "/sim", "/tokens", "/params", "/set", "/reset", "/lsmod", "/help", "/quit", "/exit"]
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -202,6 +204,20 @@ def setup_readline():
         readline.parse_and_bind("bind ^I rl_complete")
     else:
         readline.parse_and_bind("tab: complete")
+    try:
+        readline.read_history_file(HISTORY_FILE)
+    except OSError:
+        pass
+    readline.set_history_length(1000)
+    atexit.register(save_history)
+
+
+def save_history():
+    try:
+        os.makedirs(os.path.dirname(HISTORY_FILE), exist_ok=True)
+        readline.write_history_file(HISTORY_FILE)
+    except OSError:
+        pass
 
 
 HELP = """Commands: