fix(llama-index,client,build): prevent param collision, port conflict, and sed idempotency

- base.py: use indexed params (p0, p1, …) in get() to prevent
  collision when different property keys sanitize to the same name
- client.py: raise ValueError when explicit port conflicts with port
  embedded in host string
- Makefile: add PYTHON ?= python3; use $(PYTHON) -m grpc_tools.protoc
- Makefile: tighten sed to match only v1. prefix so make proto is
  idempotent (repeated runs no longer double-rewrite imports)
- _types.py: use X | Y union syntax in isinstance() (UP038)
- test_sdk.py: drop bare f-string prefix (F541)

Author: polaz

Makefile

@@ -2,12 +2,13 @@
 
 PROTO_SRC  := proto
 PROTO_OUT  := coordinode/_proto
+PYTHON     ?= python3
 
 # Generate gRPC stubs from proto submodule into coordinode/_proto/
 proto:
 	@echo "==> Generating proto stubs..."
 	@mkdir -p $(PROTO_OUT)
-	python3 -m grpc_tools.protoc \
+	$(PYTHON) -m grpc_tools.protoc \
 		-I$(PROTO_SRC) \
 		--python_out=$(PROTO_OUT) \
 		--grpc_python_out=$(PROTO_OUT) \
@@ -19,7 +20,7 @@ proto:
 	@# sed -i.bak is portable: macOS needs empty-string backup arg, GNU sed uses -i alone;
 	@# using .bak suffix works on both, then we clean up the backup files.
 	@find $(PROTO_OUT) -name '*.py' -exec sed -i.bak \
-		's/from coordinode\./from coordinode._proto.coordinode./g' {} \;
+		's/from coordinode\.v1\./from coordinode._proto.coordinode.v1./g' {} \;
 	@find $(PROTO_OUT) -name '*.py.bak' -delete
 	@echo "==> Proto generation complete: $(PROTO_OUT)/"
 

coordinode/_types.py

@@ -39,7 +39,7 @@ def to_property_value(py_val: PyValue) -> Any:
         # This package requires Python >=3.11, so no tuple-of-types workaround needed.
         # bool is a subclass of int, so exclude it explicitly — [True, False] must
         # not be serialised as a Vector of 1.0/0.0 but as a PropertyList.
-        if py_val and all(isinstance(v, (int, float)) and not isinstance(v, bool) for v in py_val):
+        if py_val and all(isinstance(v, int | float) and not isinstance(v, bool) for v in py_val):
             vec = Vector(values=[float(v) for v in py_val])
             pv.vector_value.CopyFrom(vec)
         else:

coordinode/client.py

@@ -116,7 +116,14 @@ def __init__(
         # IPv6 addresses, avoiding the ambiguity of rsplit(":", 1) on "::1".
         m = _HOST_PORT_RE.match(host)
         if m:
-            host, port = m.group(1), int(m.group(2))
+            parsed_port = int(m.group(2))
+            if port != 7080 and port != parsed_port:
+                raise ValueError(
+                    f"Conflicting ports: port={port!r} (argument) vs {parsed_port!r} "
+                    f"(embedded in host={host!r}). Specify the port in the host string "
+                    "only, or use the port argument only."
+                )
+            host, port = m.group(1), parsed_port
         self._host = host
         self._port = port
         self._tls = tls

llama-index-coordinode/llama_index/graph_stores/coordinode/base.py

@@ -91,11 +91,16 @@ def get(
                 node_id = str(row.get("_nid", ""))
                 nodes.append(_node_result_to_labelled(node_id, node_data))
         elif properties:
-            # Build param mapping: safe Cypher param name → original value.
-            # Property keys may contain hyphens or other chars invalid in
-            # Cypher parameter names; _cypher_param_name() sanitises them.
-            param_map = {_cypher_param_name(k): v for k, v in properties.items()}
-            where_clauses = " AND ".join(f"n.{_cypher_ident(k)} = ${_cypher_param_name(k)}" for k in properties)
+            # Use indexed parameter names (p0, p1, …) to avoid collisions when
+            # different property keys sanitize to the same Cypher identifier
+            # (e.g. "a-b" and "a_b" both become "a_b").
+            param_map: dict[str, Any] = {}
+            clauses: list[str] = []
+            for idx, (k, v) in enumerate(properties.items()):
+                pname = f"p{idx}"
+                clauses.append(f"n.{_cypher_ident(k)} = ${pname}")
+                param_map[pname] = v
+            where_clauses = " AND ".join(clauses)
             cypher = f"MATCH (n) WHERE {where_clauses} RETURN n, n.id AS _nid LIMIT 1000"
             result = self._client.cypher(cypher, params=param_map)
             for row in result:

tests/integration/test_sdk.py

@@ -226,9 +226,7 @@ def test_hybrid_search_returns_results(client):
     vec = [float(i) / 8 for i in range(8)]
     # Create two nodes connected by RELATED edge, both have embeddings.
     client.cypher(
-        "CREATE (a:HybridTest {tag: $tag, embedding: $vec})"
-        "-[:RELATED]->"
-        "(b:HybridTest {tag: $tag, embedding: $vec})",
+        "CREATE (a:HybridTest {tag: $tag, embedding: $vec})-[:RELATED]->(b:HybridTest {tag: $tag, embedding: $vec})",
         params={"tag": tag, "vec": vec},
     )
     # Retrieve the start node id.
@@ -247,7 +245,7 @@ def test_hybrid_search_returns_results(client):
             top_k=1,
             vector_property="embedding",
         )
-        assert len(results) >= 1, f"hybrid_search returned no results"
+        assert len(results) >= 1, "hybrid_search returned no results"
         assert hasattr(results[0], "distance")
         assert hasattr(results[0], "node")
     finally: