feat: add infra diagnostics tool to the Moose Dev MCP

[callicles]

Note

Adds a new MCP tool diagnose_infrastructure with multiple ClickHouse diagnostics and wires it into the server alongside tests and docs updates.

• MCP Server:
  • Registers new tool diagnose_infrastructure in list_tools and routes in call_tool.
  • Updates server instructions to mention diagnostics and source browsing.
  • Updates tests to account for 6 tools and validates tool names.
• New Tool: infra_issues (ClickHouse diagnostics):
  • Core module infra_issues/mod.rs with params parsing, severity filtering, Redis-backed infra map usage, and aggregated JSON output with summaries.
  • Diagnostic providers:
    • MutationDiagnostic (system.mutations): stuck/failed mutations.
    • PartsDiagnostic (system.parts): excessive parts per partition.
    • MergeDiagnostic (system.merges): long-running merges.
    • ErrorStatsDiagnostic (system.errors): top recurring errors (system-wide).
    • S3QueueDiagnostic (system.s3queue_log): ingestion failures (S3Queue tables).
    • ReplicationDiagnostic (system.replication_queue/system.replicas): lag and health issues (replicated tables).
    • MergeFailureDiagnostic (system.metrics): background merge failures (system-wide).
    • StoppedOperationsDiagnostic: merges/replication possibly stopped.
• Tool Registry:
  • Exposes infra_issues in tools/mod.rs.
• Misc:
  • get_source.rs: remove unused SourceLocationNotAvailable error variant.

^{Written by Cursor Bugbot for commit 7c40c36. This will update automatically on new commits. Configure here.}

[linear]

ENG-1132 Add general-purpose infrastructure diagnostics MCP tool (infra_issues)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
docs-v2	Ready	Preview	Comment	Nov 12, 2025 7:39pm
framework-docs	Ready	Preview	Comment	Nov 12, 2025 7:39pm
moosestack-framework-docs-v2	Ready	Preview	Comment	Nov 12, 2025 7:39pm

[cursor]

Bug: Test Inconsistency with New Tool Addition

The test cases test_list_tools_count() and test_tool_names_are_consistent() are incomplete after adding the diagnose_infrastructure tool. The tests only verify 5 tools, but the list_tools() method now returns 6 tools (including the new infra_issues::tool_definition() at line 86). The test at line 212 asserts all_tools.len() == 5 but should be 6, and the expected_tools array at line 226 is missing the "diagnose_infrastructure" tool that is now returned by list_tools().

apps/framework-cli/src/mcp/server.rs#L203-L243

moosestack/apps/framework-cli/src/mcp/server.rs

Lines 203 to 243 in d002009

	#[test]
	fn test_list_tools_count() {
	// Test that all expected tools are returned
	let logs_tool = logs::tool_definition();
	let infra_tool = infra_map::tool_definition();
	let olap_tool = query_olap::tool_definition();
	let stream_tool = sample_stream::tool_definition();
	let get_source_tool = get_source::tool_definition();
	// Ensure we have 5 tools
	let all_tools = vec![
	&logs_tool,
	&infra_tool,
	&olap_tool,
	&stream_tool,
	&get_source_tool,
	];
	assert_eq!(all_tools.len(), 5);
	// Verify each tool has required fields
	for tool in all_tools {
	assert!(!tool.name.is_empty());
	assert!(tool.description.is_some());
	assert!(!tool.input_schema.is_empty());
	}
	}
	#[test]
	fn test_tool_names_are_consistent() {
	// Ensure tool names match between routing and definitions
	let expected_tools = [
	"get_logs",
	"get_infra_map",
	"query_olap",
	"get_stream_sample",
	"get_source",
	];
	let logs_tool = logs::tool_definition();
	let infra_tool = infra_map::tool_definition();
	let olap_tool = query_olap::tool_definition();

 

---

[onelesd]

can you explain a bit the tools added and how they interact, perhaps with an example use case?

actually, just pasting the output of a session with tool calls would be illuminating

[callicles]

@onelesd Did you take a look at the issue? there is a bunch of context there

[cursor]

Bug: Test Fails to Validate New Tool Consistency

The test_tool_names_are_consistent test has an expected_tools array with only 5 tool names, missing "diagnose_infrastructure", and doesn't instantiate or check the infra_issues::tool_definition(). This causes the test to not validate the newly added tool's name consistency with the routing logic in call_tool.

apps/framework-cli/src/mcp/server.rs#L253-L272

moosestack/apps/framework-cli/src/mcp/server.rs

Lines 253 to 272 in 360a8eb

	#[test]
	fn test_tool_names_are_consistent() {
	// Ensure tool names match between routing and definitions
	let expected_tools = [
	"get_logs",
	"get_infra_map",
	"query_olap",
	"get_stream_sample",
	"get_source",
	];
	let logs_tool = logs::tool_definition();
	let infra_tool = infra_map::tool_definition();
	let olap_tool = query_olap::tool_definition();
	let stream_tool = sample_stream::tool_definition();
	let get_source_tool = get_source::tool_definition();
	assert_eq!(logs_tool.name, expected_tools[0]);
	assert_eq!(infra_tool.name, expected_tools[1]);
	assert_eq!(olap_tool.name, expected_tools[2]);

 

---

[cursor]

Bug: Test fails to validate all tools.

The test test_list_tools_count only validates 5 tool definitions but list_tools() now returns 6 tools including the newly added infra_issues tool. The test's comment claims to verify "all expected tools" but omits infra_issues_tool, leaving that tool's definition untested.

apps/framework-cli/src/mcp/server.rs#L232-L250

moosestack/apps/framework-cli/src/mcp/server.rs

Lines 232 to 250 in 205dbc8

	let stream_tool = sample_stream::tool_definition();
	let get_source_tool = get_source::tool_definition();
	// Ensure we have 5 tools
	let all_tools = vec![
	&logs_tool,
	&infra_tool,
	&olap_tool,
	&stream_tool,
	&get_source_tool,
	];
	assert_eq!(all_tools.len(), 5);
	// Verify each tool has required fields
	for tool in all_tools {
	assert!(!tool.name.is_empty());
	assert!(tool.description.is_some());
	assert!(!tool.input_schema.is_empty());
	}

 

---

[cursor]

Bug: Tool Count Test: Expectation vs. Reality

The test_list_tools_count test checks only 5 tools but list_tools() now returns 6 tools including the new infra_issues::tool_definition(). The test doesn't include infra_issues_tool in the count verification, causing the test to assert an incorrect tool count that doesn't match the actual implementation.

apps/framework-cli/src/mcp/server.rs#L226-L248

moosestack/apps/framework-cli/src/mcp/server.rs

Lines 226 to 248 in ebe61ae

	#[test]
	fn test_list_tools_count() {
	// Test that all expected tools are returned
	let logs_tool = logs::tool_definition();
	let infra_tool = infra_map::tool_definition();
	let olap_tool = query_olap::tool_definition();
	let stream_tool = sample_stream::tool_definition();
	let get_source_tool = get_source::tool_definition();
	// Ensure we have 5 tools
	let all_tools = vec![
	&logs_tool,
	&infra_tool,
	&olap_tool,
	&stream_tool,
	&get_source_tool,
	];
	assert_eq!(all_tools.len(), 5);
	// Verify each tool has required fields
	for tool in all_tools {
	assert!(!tool.name.is_empty());
	assert!(tool.description.is_some());

 

---

[LucioFranco]

LGTM, looking forward to giving this a try and improving on this! This is a really great start and I am quite happy because I think what is in here aligns really well with what we've been talking about!