trending/codebase-memory-mcp
1
0
Fork 0
mirror of https://github.com/DeusData/codebase-memory-mcp.git synced 2026-11-06 08:35:21 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add cross-service query guidance to tool descriptions

Browse Source 
Clarify that search_graph relationship filter counts node degrees
(not edges), direct users to query_graph Cypher for listing
HTTP_CALLS/ASYNC_CALLS edges with properties, and recommend
direction=both in trace_call_path for full cross-service context.
This commit is contained in:
Martin Vogel
2026-02-25 14:20:48 +01:00
parent 55850fa888
commit e4839e758d
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/tools/tools.go
+3 -3
View File
@@ -66,7 +66,7 @@ func (s *Server) registerIndexAndTraceTool() {
s.mcp.AddTool(&mcp.Tool{
Name: "trace_call_path",
Description: "Trace call paths from/to a function. Requires EXACT function name — if unsure, call search_graph(name_pattern=...) first to discover the correct name. Returns hop-by-hop callees/callers with edge types (CALLS, HTTP_CALLS, ASYNC_CALLS). Use depth=1 first, increase only if needed. IMPORTANT: If the function is not found, use search_graph with a name_pattern regex to find similar names before retrying.",
Description: "Trace call paths from/to a function. Requires EXACT function name — if unsure, call search_graph(name_pattern=...) first to discover the correct name. Returns hop-by-hop callees/callers with edge types (CALLS, HTTP_CALLS, ASYNC_CALLS). Use depth=1 first, increase only if needed. IMPORTANT: Use direction='both' for full cross-service context — HTTP_CALLS edges from other services appear as inbound edges on backend functions, so direction='outbound' alone misses cross-service callers. For async dispatch (Cloud Tasks, Pub/Sub), find dispatch functions via search_graph(name_pattern='.*CreateTask.*') then trace via CALLS edges. If the function is not found, use search_graph with a name_pattern regex to find similar names before retrying.",
InputSchema: json.RawMessage(`{
"type": "object",
"properties": {
@@ -115,7 +115,7 @@ func (s *Server) registerSchemaAndSnippetTools() {
func (s *Server) registerSearchTools() {
s.mcp.AddTool(&mcp.Tool{
Name: "search_graph",
Description: "Search the code graph. Returns 10 results per page (use offset to paginate, has_more indicates more pages). For dead code: use relationship='CALLS', direction='inbound', max_degree=0, exclude_entry_points=true. For fan-out: use relationship='CALLS', direction='outbound', min_degree=N. Route nodes: properties.handler contains the actual handler function name. Prefer this over query_graph for counting — no row cap. Relationship types: CALLS, HTTP_CALLS, ASYNC_CALLS, IMPORTS, DEFINES, DEFINES_METHOD, HANDLES, CONTAINS_FILE, CONTAINS_FOLDER, CONTAINS_PACKAGE, IMPLEMENTS.",
Description: "Search the code graph for nodes. Returns 10 results per page (use offset to paginate, has_more indicates more pages). For dead code: use relationship='CALLS', direction='inbound', max_degree=0, exclude_entry_points=true. For fan-out: use relationship='CALLS', direction='outbound', min_degree=N. Route nodes: properties.handler contains the actual handler function name. Prefer this over query_graph for counting — no row cap. IMPORTANT: The 'relationship' filter counts how many edges of that type each node has (degree filtering) — it does NOT return the actual edges. To list cross-service HTTP_CALLS or ASYNC_CALLS edges with their properties (url_path, confidence), use query_graph with Cypher instead: MATCH (a)-[r:HTTP_CALLS]->(b) RETURN a.name, b.name, r.url_path, r.confidence. Relationship types: CALLS, HTTP_CALLS, ASYNC_CALLS, IMPORTS, DEFINES, DEFINES_METHOD, HANDLES, CONTAINS_FILE, CONTAINS_FOLDER, CONTAINS_PACKAGE, IMPLEMENTS.",
InputSchema: json.RawMessage(`{
"type": "object",
"properties": {
@@ -207,7 +207,7 @@ func (s *Server) registerSearchTools() {
func (s *Server) registerQueryTool() {
s.mcp.AddTool(&mcp.Tool{
Name: "query_graph",
Description: "Execute a Cypher-like graph query. WARNING: 200-row cap applies BEFORE aggregation — COUNT queries on large codebases silently undercount. For fan-out/fan-in counting, use search_graph with min_degree/max_degree instead. Best for: relationship patterns, filtered joins, path queries. Always use LIMIT. Edge types: CALLS, HTTP_CALLS, ASYNC_CALLS, IMPORTS, DEFINES, DEFINES_METHOD, HANDLES, IMPLEMENTS.",
Description: "Execute a Cypher-like graph query. WARNING: 200-row cap applies BEFORE aggregation — COUNT queries on large codebases silently undercount. For fan-out/fan-in counting, use search_graph with min_degree/max_degree instead. Best for: relationship patterns, filtered joins, path queries. This is the correct tool for listing cross-service edges — use MATCH (a)-[r:HTTP_CALLS]->(b) RETURN a.name, b.name, r.url_path, r.confidence to see HTTP links with URLs and confidence scores, or MATCH (a)-[r:ASYNC_CALLS]->(b) for async dispatch edges. Always use LIMIT. Edge types: CALLS, HTTP_CALLS, ASYNC_CALLS, IMPORTS, DEFINES, DEFINES_METHOD, HANDLES, IMPLEMENTS.",
InputSchema: json.RawMessage(`{
"type": "object",
"properties": {