Semantic Search MCP Server

{"id": "django_request_response_lifecycle", "description": "Full request/response lifecycle through middleware and views", "repo_path": "django", "query": "My custom middleware adds a header but sometimes it's missing in the final response. I suspect something removes it after my middleware runs. I need to understand the complete sequence of what touches the response object from start to finish.", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "django/core/handlers/base.py", "must_include_substrings": ["class BaseHandler", "def get_response(", "_middleware_chain", "def _get_response(", "process_template_response", "process_exception_by_middleware", "load_middleware"]}, {"file_path": "django/core/handlers/wsgi.py", "must_include_substrings": ["class WSGIHandler", "def __call__(self, environ", "response_headers", "start_response"]}, {"file_path": "django/core/handlers/exception.py", "must_include_substrings": ["convert_exception_to_response"]}, {"file_path": "django/utils/deprecation.py", "must_include_substrings": ["class MiddlewareMixin", "process_request", "process_response", "get_response"]}]} {"id": "django_queryset_lazy_evaluation", "description": "How QuerySet defers database execution until iteration", "repo_path": "django", "query": "I have a view that builds up a complex query with many chained methods, but the page loads instantly even though the query should be slow. Then when I loop over the results it suddenly takes 5 seconds. Where is this delayed execution implemented?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "django/db/models/query.py", "must_include_substrings": ["class QuerySet", "def __iter__(self)", "def __repr__(", "def _fetch_all(", "_result_cache", "def __bool__", "def __len__"]}, {"file_path": "django/db/models/query.py", "must_include_substrings": ["class ModelIterable", "compiler.execute_sql"]}]} {"id": "gin_route_tree_matching", "description": "How Gin's radix tree matches URL paths to handlers", "repo_path": "gin", "query": "I registered /users/new before /users/:id but requests to /users/new are hitting the :id handler. Why does registration order matter and how does the framework decide between overlapping patterns?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "tree.go", "must_include_substrings": ["type node struct", "func (n *node) getValue(", "func (n *node) addRoute", "indices", "wildChild", "priority", "type nodeType", "incrementChildPrio", "skippedNodes", "type methodTree struct"]}, {"file_path": "gin.go", "must_include_substrings": ["methodTrees", "trees", "handleHTTPRequest", "ServeHTTP"]}]} {"id": "gin_context_pool_reuse", "description": "How Gin reuses Context objects via sync.Pool for performance", "repo_path": "gin", "query": "I spawned a goroutine from my handler and passed it the request context, but later the data in the context was corrupted or belonged to a different request. What causes this and how do I safely use the context in background tasks?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "gin.go", "must_include_substrings": ["pool", "pool.Get()", "pool.Put(", "func (engine *Engine) ServeHTTP(", "sync.Pool"]}, {"file_path": "context.go", "must_include_substrings": ["func (c *Context) reset()", "func (c *Context) Copy()", "type Context struct", "Keys map", "This has to be used when the context has to be passed to a goroutine"]}]} {"id": "codeql_dataflow_configuration", "description": "How CodeQL data flow configurations define sources, sinks and flow", "repo_path": "codeql", "query": "I wrote a function that validates and escapes user input but CodeQL still reports it as vulnerable. How do I mark my validation function as something that stops dangerous data from flowing through?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "javascript/ql/lib/semmle/javascript/dataflow/Configuration.qll", "must_include_substrings": ["class Configuration", "isSource", "isSink", "isBarrier", "isBarrierEdge"]}, {"file_path": "javascript/ql/lib/semmle/javascript/dataflow/TaintTracking.qll", "must_include_substrings": ["isSanitizer", "isSanitizerEdge", "isSanitizerGuard", "isBarrier"]}, {"file_path": "docs/codeql/codeql-language-guides/analyzing-data-flow-in-javascript-and-typescript.rst", "must_include_substrings": ["isBarrier", "sanitizer", "MakeBarrierGuard"]}]} {"id": "codeql_security_query_structure", "description": "How security queries are structured to find vulnerabilities", "repo_path": "codeql", "query": "I want to find places where data from HTTP request parameters ends up being written to the page without encoding. What's the standard approach for writing this kind of query? I need to trace data from input to dangerous output.", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "javascript/ql/lib/semmle/javascript/security/dataflow/DomBasedXssQuery.qll", "must_include_substrings": ["DomBasedXssConfig", "isSource", "isSink", "isBarrier", "TaintTracking"]}, {"file_path": "javascript/ql/lib/semmle/javascript/security/dataflow/Xss.qll", "must_include_substrings": ["module Shared", "class Source", "class Sink", "class Sanitizer"]}, {"file_path": "javascript/ql/lib/semmle/javascript/security/dataflow/ReflectedXssCustomizations.qll", "must_include_substrings": ["class Source", "class Sink", "Http::RequestInputAccess", "Http::ResponseSendArgument"]}]} {"id": "infinigen_procedural_material_nodes", "description": "How materials are procedurally generated using shader nodes", "repo_path": "infinigen", "query": "I want to create materials that look different each time based on a seed value - like varied wood grain or random rust patches. How does the system construct these randomized appearances and connect them to Blender's rendering?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "infinigen/core/nodes/node_wrangler.py", "must_include_substrings": ["class NodeWrangler", "def new_node(", "self.links", "self.nodes"]}, {"file_path": "infinigen/core/nodes/node_utils.py", "must_include_substrings": ["resample_node_group", "NoiseTexture", "random", "np.random"]}, {"file_path": "infinigen/core/surface.py", "must_include_substrings": ["def add_material(", "shaderfunc_to_material", "material.use_nodes"]}]} {"id": "infinigen_asset_placement_pipeline", "description": "How assets are placed and populated in scenes", "repo_path": "infinigen", "query": "When generating a nature scene, objects need to be distributed realistically - grass under trees, rocks on slopes, no floating objects. How does the pipeline handle these spatial constraints and object density?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "infinigen/core/placement/factory.py", "must_include_substrings": ["class AssetFactory", "def spawn_asset", "spawn_placeholder"]}, {"file_path": "infinigen/core/placement/placement.py", "must_include_substrings": ["placeholder_locs", "DistributePointsOnFaces", "populate_all"]}, {"file_path": "infinigen/core/placement/density.py", "must_include_substrings": ["placement_mask", "normal_thresh", "altitude_range", "tag_mask"]}, {"file_path": "infinigen/core/placement/instance_scatter.py", "must_include_substrings": ["geo_instance_scatter", "scatter_instances", "DistributePointsOnFaces", "min_spacing"]}]} {"id": "qgis_map_rendering_pipeline", "description": "How map layers are rendered in parallel to canvas", "repo_path": "QGIS", "query": "When I have 10 heavy layers and pan the map, the UI stays responsive and layers appear progressively. How does the application split up the drawing work so it doesn't freeze? What coordinates the background drawing?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "src/core/maprenderer/qgsmaprendererparalleljob.cpp", "must_include_substrings": ["QgsMapRendererParallelJob", "startPrivate", "renderLayerStatic", "QtConcurrent", "mFutureWatcher", "renderLayersFinished"]}, {"file_path": "src/core/maprenderer/qgsmaprendererjob.cpp", "must_include_substrings": ["QgsMapRendererJob", "prepareJobs", "LayerRenderJob", "composeImage"]}, {"file_path": "src/gui/qgsmapcanvas.cpp", "must_include_substrings": ["QgsMapCanvas", "mJob", "refresh(", "rendererJobFinished", "mUseParallelRendering", "QgsMapRendererParallelJob"]}]} {"id": "qgis_expression_evaluation", "description": "How QGIS evaluates expressions in symbology and labeling", "repo_path": "QGIS", "query": "I have a complex formula in my label settings that calculates text from multiple fields and geometry properties. Rendering is slow on large datasets. Where does this formula calculation happen and is there any way to optimize it?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "src/core/expression/qgsexpression.cpp", "must_include_substrings": ["QgsExpression", "evaluate(", "prepare(", "setExpression"]}, {"file_path": "src/core/qgsexpressioncontext.h", "must_include_substrings": ["QgsExpressionContext", "QgsExpressionContextScope", "setFeature"]}, {"file_path": "src/core/labeling/qgspallabeling.cpp", "must_include_substrings": ["registerFeatureWithDetails", "getLabelExpression", "expressionContext"]}]} {"id": "claude_flow_agent_coordination", "description": "How multiple agents coordinate and communicate in swarm", "repo_path": "claude-flow", "query": "I started multiple workers to process different parts of a project but two of them tried to modify the same file simultaneously and created a mess. How do I make them aware of each other and avoid conflicts?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "src/task/coordination.ts", "must_include_substrings": ["class TaskCoordinator", "agentCoordination", "storeInMemory", "memoryStore"]}, {"file_path": "src/cli/simple-commands/inject-memory-protocol.js", "must_include_substrings": ["MEMORY_PROTOCOL", "swarm/", "coordination", "status", "waiting_for"]}, {"file_path": "src/resources/resource-manager.ts", "must_include_substrings": ["class ResourceManager", "requestResources", "releaseResources", "ResourceReservation"]}]} {"id": "claude_flow_memory_system", "description": "How agent memory persists and is retrieved across sessions", "repo_path": "claude-flow", "query": "My AI assistant learned my coding preferences during our session but after restarting it forgot everything. How can I make it remember things permanently? Where would that information be saved?", "difficulty": "hard", "grep_resistant": true, "expected_items": [{"file_path": "src/swarm/memory.ts", "must_include_substrings": ["MemoryPersistence", "saveState", "loadState", "persistencePath"]}, {"file_path": "src/core/DatabaseManager.ts", "must_include_substrings": ["IDatabaseProvider", "JSONProvider", "store", "initialize"]}, {"file_path": "src/cli/simple-commands/init/templates/memory-bank-md.js", "must_include_substrings": ["claude-flow-data.json", "memory/sessions", "retentionPolicy", "permanent"]}]}