fix: coerce MCP integer arguments to native Python int

[AlexeySamosadov]

Summary

• MCP JSON-RPC transport can deliver integer-typed arguments as floats (3.0) or strings ("3") depending on the client
• ChromaDB's collection.query(n_results=...) requires a native Python int, causing mempalace_search to fail with: "Expected requested number of results to be a int, got 3 in query."
• This affects all tools with integer params: limit (search), max_hops (traverse), last_n (diary_read)

Fix

Auto-coerce tool arguments in handle_request() based on the declared input_schema types before calling handlers. This is generic — covers all current and future tools without per-handler changes.

schema_props = TOOLS[tool_name]["input_schema"].get("properties", {})
for key, value in list(tool_args.items()):
    declared_type = schema_props.get(key, {}).get("type")
    if declared_type == "integer" and not isinstance(value, int):
        tool_args[key] = int(value)
    elif declared_type == "number" and not isinstance(value, (int, float)):
        tool_args[key] = float(value)

Reproduction

from mempalace.searcher import search_memories
search_memories("test", palace_path, n_results=3.0)
# → "Search error: Expected requested number of results to be a int, got 3.0 in query."

Test plan

• limit=3 (int) — works before and after
• limit=3.0 (float from JSON) — fails before, works after
• limit="3" (string) — fails before, works after
• No regressions on tools without integer params

Discovered while using MemPalace via Claude Code MCP integration.

🤖 Generated with Claude Code