Diagnostics across all open files in one call, grouped by file path.
Includes cross-file E0400 warnings for undefined imports, with optional
severity filter and file glob filter. 12/12 tests pass.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
getInScopeSymbols gains crossFile parameter to include exported symbols from
all open buffers. Import graph tracks file dependencies, auto-maintained on
openFile/closeFile. Text-based fallback for parsers without Import AST nodes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds multi-buffer management via openFile/closeFile/listBuffers/setActiveBuffer
RPC methods, plus indexWorkspace for scanning workspace files without opening
them. Language auto-detected from extension. 27 MCP tools total (was 22).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Systematic validation of all Phase 9c token-saving features: compact AST
(6-7% of full size), lean scope (4% of detailed), diagnostic deltas, budget
pagination across 3 pages, batch queries, and combined 95% savings. Benchmark
confirms sub-millisecond parse->mutate->diagnose cycles. Also adds Environment
Layer feature request to docs/annotations.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
batchQuery accepts array of sub-queries, returns all results in one
response. Errors in individual sub-queries don't affect others.
22 MCP tools total. 12/12 tests pass.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
getInScopeSymbols, getCallHierarchy, getDependencyGraph now default
to lean "symbols" mode returning names/IDs only. Pass detailed:true
for full node JSON. 12/12 tests pass.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Optional budget parameter on getAST and getDiagnostics. Responses exceeding
the budget are truncated with continuation tokens for paginated follow-up.
Binary search finds max array elements fitting the budget. Diagnostics sorted
by severity (errors first) before truncation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
DiagnosticVersionTracker records snapshots and computes added/removed deltas
between versions. Agents can efficiently check if their fix resolved an error
without re-fetching all diagnostics. 21 MCP tools total.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Concrete mutation objects for each diagnostic fix. getQuickFixes returns
reviewable fixes with categories; applyQuickFix applies the fix and reports
whether the diagnostic cleared. Heuristic fixes for missing return and
unused variables. 20 MCP tools total.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds token-efficient AST queries: compact mode (<30% of full size),
subtree extraction by nodeId, version-tracked AST diff after mutations,
and tokenEstimate on all responses. 12/12 tests pass, 17 tools total.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds 5 workspace-aware file I/O tools to the MCP server so agents can
manage files without a GUI. All paths are resolved against --workspace
root with escape-rejection security. 12/12 tests pass, tool count now 15.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wires HeadlessEditorState + MCPBridge into a standalone whetstone_mcp
binary that any MCP client (Claude Code, Cursor, etc.) can launch over
stdio. No SDL, no ImGui, no OpenGL — headless only. 12/12 tests pass.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds the headless agent architecture for Sprint 9 Phase 9a:
- ASTUtils.h: pure AST utilities extracted from EditorUtils.h
- HeadlessEditorState.h: GUI-free state with buffer management
- HeadlessAgentRPCHandler.h: full RPC dispatch (20+ methods)
- step245_test.cpp: 20 tests all passing
Also fixes test compilation errors (NotificationSystem API changes,
AgentRole permissions, DependencyPanel missing include) and adds
SDL2 system library detection fix in CMakeLists.txt.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
TraceGenerator with 6 scenario templates, built-in code corpus,
and deterministic batch generation. TraceExporter supports Anthropic
Messages, OpenAI Chat, JSONL, and Markdown formats with filtering
and statistics. 294 tests passing.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
