When trying to use an .env file for arguments, the retrieval mode isn't respected.
I haven't looked at the source code thoroughly, but it seems that config.py has MemoryEngineConfig which reads MEMORY_RETRIEVAL_MODE from the env file, but main.py uses args.retrieval_mode from argparse instead.
I can manually change the hard-coded argument in main.py, but that kind of defeats the point of an argument:
parser.add_argument(
"--retrieval-mode",
default="smart_vector",
choices=["claude", "smart_vector", "hybrid"],
help="Memory retrieval strategy (default: smart_vector)\n"
"- claude: Use Claude for every retrieval (high quality, high cost)\n"
"- smart_vector: Intelligent vector search with metadata (fast, smart)\n"
"- hybrid: Start with vector, escalate to Claude for complex queries"
)
Also, start_server.py doesn't forward arguments, so to use CLI arguments, one would have to invoke main.py directly.
Unless I'm just missing something?
When trying to use an .env file for arguments, the retrieval mode isn't respected.
I haven't looked at the source code thoroughly, but it seems that config.py has MemoryEngineConfig which reads MEMORY_RETRIEVAL_MODE from the env file, but main.py uses args.retrieval_mode from argparse instead.
I can manually change the hard-coded argument in main.py, but that kind of defeats the point of an argument:
Also, start_server.py doesn't forward arguments, so to use CLI arguments, one would have to invoke main.py directly.
Unless I'm just missing something?