Cross-service communication discovery + RAM-first incremental indexing

AST-based detection of HTTP calls, async dispatch (Pub/Sub, Cloud Tasks,
Kafka, SQS, etc.), and config accesses via resolved qualified names.
Route nodes as cross-service rendezvous points with infra→handler matching.
Constant propagation for module-level string assignments. YAML infrastructure
URL extraction from Cloud Scheduler configs.

RAM-first incremental pipeline: load DB into graph buffer, purge changed
file nodes, extract directly into existing buffer (resolver sees all nodes),
dump back to disk. Zero edge gap on kubernetes/django/meilisearch/neovim.

- service_patterns.c: ~170 library patterns (90 HTTP, 50 async, 30 config)
- pass_route_nodes.c: Route node creation + infra URL matching
- extract_unified.c: string constant collection + string ref classification
- extract_calls.c: first_string_arg + keyword argument extraction
- pipeline_incremental.c: RAM-first load→purge→extract→resolve→dump
- graph_buffer.c: load_from_db, delete_by_file, foreach visitors
- C++ LSP crash fix: NULL guard in cbm_type_substitute

Author: DeusData

CONTRIBUTING.md

@@ -118,8 +118,32 @@ Examples: `fix(store): set busy_timeout before WAL`, `feat(cli): add --progress
 
 ## Pull Request Guidelines
 
-- **One issue per PR.** Each PR must address exactly one bug, one feature, or one refactor. Do not bundle multiple fixes or feature additions into a single PR. If your change touches multiple areas, split it into separate PRs.
-- **Open an issue first.** Every PR should reference a tracking issue (`Fixes #N` or `Closes #N`). This ensures the change is discussed before code is written.
+### Before You Write Code
+
+- **Open an issue first — always.** Every PR must reference a tracking issue (`Fixes #N` or `Closes #N`). Describe what you want to change and why. Wait for maintainer feedback before implementing. PRs without a prior issue discussion will be closed.
+- **Bug fixes and test additions** are the exception — these are welcome without prior discussion, as long as they're focused.
+
+### What Requires Explicit Maintainer Approval
+
+The following changes will not be merged without prior design discussion in an issue:
+
+- **API surface changes** — adding, removing, renaming, or changing defaults of MCP tools
+- **New pipeline passes or indexing algorithms** — anything that changes what gets extracted or how
+- **Build system / Makefile changes** — beyond trivial fixes
+- **Project configuration** — CLAUDE.md, skill files, .mcp.json, CI workflows
+- **New dependencies** — vendored or otherwise
+- **Breaking changes** of any kind
+
+If in doubt, open an issue and ask.
+
+### PR Scope and Size
+
+- **One issue per PR.** Each PR must address exactly one bug, one feature, or one refactor. Do not bundle multiple fixes or feature additions into a single PR. Kitchen-sink PRs will be closed with a request to split.
+- **Keep PRs small.** A good PR is under 500 lines. If your change is larger, split it into reviewable increments that each stand on their own.
+- **Don't mix features with fixes.** If you find a bug while implementing a feature, submit the bug fix as a separate PR.
+
+### Code Requirements
+
 - **C code only** — this project was rewritten from Go to pure C in v0.5.0. Go PRs will be acknowledged and potentially ported, but cannot be merged directly.
 - Include tests for new functionality
 - Run `scripts/test.sh` and `scripts/lint.sh` before submitting

Makefile.cbm

@@ -118,7 +118,8 @@ EXTRACTION_SRCS = \
     $(CBM_DIR)/extract_env_accesses.c \
     $(CBM_DIR)/extract_k8s.c \
     $(CBM_DIR)/helpers.c \
-    $(CBM_DIR)/lang_specs.c
+    $(CBM_DIR)/lang_specs.c \
+    $(CBM_DIR)/service_patterns.c
 
 # LSP resolvers (compiled as one unit via lsp_all.c)
 LSP_SRCS = $(CBM_DIR)/lsp_all.c
@@ -175,6 +176,7 @@ PIPELINE_SRCS = \
     src/pipeline/pass_gitdiff.c \
     src/pipeline/pass_configures.c \
     src/pipeline/pass_configlink.c \
+    src/pipeline/pass_route_nodes.c \
     src/pipeline/pass_enrichment.c \
     src/pipeline/pass_envscan.c \
     src/pipeline/pass_compile_commands.c \

internal/cbm/cbm.c

@@ -124,6 +124,11 @@ void cbm_typeassign_push(CBMTypeAssignArray *arr, CBMArena *a, CBMTypeAssign ta)
     arr->items[arr->count++] = ta;
 }
 
+void cbm_stringref_push(CBMStringRefArray *arr, CBMArena *a, CBMStringRef sr) {
+    GROW_ARRAY(arr, a);
+    arr->items[arr->count++] = sr;
+}
+
 void cbm_impltrait_push(CBMImplTraitArray *arr, CBMArena *a, CBMImplTrait it) {
     GROW_ARRAY(arr, a);
     arr->items[arr->count++] = it;

internal/cbm/cbm.h

@@ -110,6 +110,7 @@ typedef struct {
 typedef struct {
     const char *callee_name;       // raw callee text ("pkg.Func", "foo")
     const char *enclosing_func_qn; // QN of enclosing function (or module QN)
+    const char *first_string_arg;  // first string literal argument (URL, topic, key) or NULL
 } CBMCall;
 
 typedef struct {
@@ -149,6 +150,19 @@ typedef struct {
     const char *enclosing_func_qn; // QN of enclosing function
 } CBMTypeAssign;
 
+// String reference: URL, config key, or async target found in source.
+// Extracted from string literals during AST walk.
+typedef enum {
+    CBM_STRREF_URL = 0,    // REST path or full URL
+    CBM_STRREF_CONFIG = 1, // config file path or env var key
+} CBMStringRefKind;
+
+typedef struct {
+    const char *value;             // the string literal content
+    const char *enclosing_func_qn; // QN of enclosing function
+    CBMStringRefKind kind;         // URL, CONFIG
+} CBMStringRef;
+
 // Rust: impl Trait for Struct
 typedef struct {
     const char *trait_name;  // trait name (raw text)
@@ -225,6 +239,12 @@ typedef struct {
     int cap;
 } CBMTypeAssignArray;
 
+typedef struct {
+    CBMStringRef *items;
+    int count;
+    int cap;
+} CBMStringRefArray;
+
 typedef struct {
     CBMImplTrait *items;
     int count;
@@ -246,6 +266,7 @@ typedef struct {
     CBMTypeAssignArray type_assigns;
     CBMImplTraitArray impl_traits;       // Rust: impl Trait for Struct pairs
     CBMResolvedCallArray resolved_calls; // LSP-resolved calls (high confidence)
+    CBMStringRefArray string_refs;       // URL/config string literals from AST
 
     const char *module_qn;    // module qualified name
     const char **exports;     // NULL-terminated (NULL if none)
@@ -279,6 +300,14 @@ typedef struct {
 
 // --- Extraction context passed to sub-extractors ---
 
+// Module-level string constant map (for constant propagation)
+#define CBM_MAX_STRING_CONSTANTS 256
+typedef struct {
+    const char *names[CBM_MAX_STRING_CONSTANTS];
+    const char *values[CBM_MAX_STRING_CONSTANTS];
+    int count;
+} CBMStringConstantMap;
+
 typedef struct {
     CBMArena *arena;
     CBMFileResult *result;
@@ -289,8 +318,9 @@ typedef struct {
     const char *rel_path;
     const char *module_qn;
     TSNode root;
-    EFCache ef_cache;               // enclosing function cache
-    const char *enclosing_class_qn; // for nested class QN computation
+    EFCache ef_cache;                      // enclosing function cache
+    const char *enclosing_class_qn;        // for nested class QN computation
+    CBMStringConstantMap string_constants; // module-level NAME = "value" pairs
 } CBMExtractCtx;
 
 // --- Public API ---
@@ -346,6 +376,7 @@ void cbm_rw_push(CBMRWArray *arr, CBMArena *a, CBMReadWrite rw);
 void cbm_typerefs_push(CBMTypeRefArray *arr, CBMArena *a, CBMTypeRef tr);
 void cbm_envaccess_push(CBMEnvAccessArray *arr, CBMArena *a, CBMEnvAccess ea);
 void cbm_typeassign_push(CBMTypeAssignArray *arr, CBMArena *a, CBMTypeAssign ta);
+void cbm_stringref_push(CBMStringRefArray *arr, CBMArena *a, CBMStringRef sr);
 void cbm_impltrait_push(CBMImplTraitArray *arr, CBMArena *a, CBMImplTrait it);
 void cbm_resolvedcall_push(CBMResolvedCallArray *arr, CBMArena *a, CBMResolvedCall rc);
 

internal/cbm/extract_calls.c

@@ -8,6 +8,20 @@
 #include <string.h>
 #include <ctype.h>
 
+/* Look up a module-level string constant by name. */
+static const char *lookup_string_constant(const CBMExtractCtx *ctx, const char *name) {
+    if (!name || !name[0]) {
+        return NULL;
+    }
+    const CBMStringConstantMap *map = &ctx->string_constants;
+    for (int i = 0; i < map->count; i++) {
+        if (strcmp(map->names[i], name) == 0) {
+            return map->values[i];
+        }
+    }
+    return NULL;
+}
+
 // Forward declarations
 static void walk_calls(CBMExtractCtx *ctx, TSNode node, const CBMLangSpec *spec);
 static char *extract_callee_name(CBMArena *a, TSNode node, const char *source, CBMLanguage lang);
@@ -256,6 +270,46 @@ static void walk_calls(CBMExtractCtx *ctx, TSNode node, const CBMLangSpec *spec)
                 CBMCall call;
                 call.callee_name = callee;
                 call.enclosing_func_qn = cbm_enclosing_func_qn_cached(ctx, node);
+                call.first_string_arg = NULL;
+
+                /* Extract first string literal argument (URL, topic, key) */
+                TSNode args = ts_node_child_by_field_name(node, "arguments", 9);
+                if (!ts_node_is_null(args)) {
+                    uint32_t nc = ts_node_named_child_count(args);
+                    for (uint32_t ai = 0; ai < nc && ai < 3; ai++) {
+                        TSNode arg = ts_node_named_child(args, ai);
+                        const char *ak = ts_node_type(arg);
+                        if (strcmp(ak, "string") == 0 || strcmp(ak, "string_literal") == 0 ||
+                            strcmp(ak, "interpreted_string_literal") == 0 ||
+                            strcmp(ak, "raw_string_literal") == 0 ||
+                            strcmp(ak, "string_content") == 0) {
+                            char *text = cbm_node_text(ctx->arena, arg, ctx->source);
+                            if (text && text[0]) {
+                                /* Strip quotes */
+                                int len = (int)strlen(text);
+                                if (len >= 2 && (text[0] == '"' || text[0] == '\'')) {
+                                    text =
+                                        cbm_arena_strndup(ctx->arena, text + 1, (size_t)(len - 2));
+                                    len -= 2;
+                                }
+                                /* Validate: must be printable ASCII, no control chars */
+                                // NOLINTNEXTLINE(readability-implicit-bool-conversion)
+                                bool valid = (text != NULL && len > 0 && len < 512);
+                                for (int vi = 0; vi < len && valid; vi++) {
+                                    unsigned char ch = (unsigned char)text[vi];
+                                    if (ch < 0x20 && ch != '\t') {
+                                        valid = false;
+                                    }
+                                }
+                                if (valid) {
+                                    call.first_string_arg = text;
+                                }
+                            }
+                            break;
+                        }
+                    }
+                }
+
                 cbm_calls_push(&ctx->result->calls, ctx->arena, call);
             }
         }
@@ -294,7 +348,7 @@ static void extract_jsx_refs(CBMExtractCtx *ctx, TSNode node) {
         return;
     }
 
-    CBMCall call;
+    CBMCall call = {0};
     call.callee_name = name;
     call.enclosing_func_qn = cbm_enclosing_func_qn_cached(ctx, node);
     cbm_calls_push(&ctx->result->calls, ctx->arena, call);
@@ -321,9 +375,135 @@ void handle_calls(CBMExtractCtx *ctx, TSNode node, const CBMLangSpec *spec, Walk
     if (cbm_kind_in_set(node, spec->call_node_types)) {
         char *callee = extract_callee_name(ctx->arena, node, ctx->source, ctx->language);
         if (callee && callee[0] && !cbm_is_keyword(callee, ctx->language)) {
-            CBMCall call;
+            CBMCall call = {0};
             call.callee_name = callee;
             call.enclosing_func_qn = state->enclosing_func_qn;
+
+            /* Extract URL/topic/key from call arguments.
+             * Strategy: check keyword args first (url=, topic_id=, queue=),
+             * then first positional string, then resolve constant references. */
+            TSNode args = ts_node_child_by_field_name(node, "arguments", 9);
+            if (!ts_node_is_null(args)) {
+                /* Keyword patterns that indicate URL/topic/key */
+                static const char *url_keywords[] = {"url",        "endpoint", "path", "uri",
+                                                     "target_url", "base_url", NULL};
+                static const char *topic_keywords[] = {"topic",   "topic_id",   "topic_name",
+                                                       "queue",   "queue_name", "queue_id",
+                                                       "subject", "channel",    NULL};
+
+                uint32_t nc = ts_node_named_child_count(args);
+                for (uint32_t ai = 0; ai < nc && !call.first_string_arg; ai++) {
+                    TSNode arg = ts_node_named_child(args, ai);
+                    const char *ak = ts_node_type(arg);
+
+                    /* Check keyword_argument nodes: url="...", topic_id="..." */
+                    if (strcmp(ak, "keyword_argument") == 0 || strcmp(ak, "pair") == 0) {
+                        TSNode key_node = ts_node_child_by_field_name(arg, "name", 4);
+                        TSNode val_node = ts_node_child_by_field_name(arg, "value", 5);
+                        if (ts_node_is_null(key_node)) {
+                            key_node = ts_node_child_by_field_name(arg, "key", 3);
+                        }
+                        if (ts_node_is_null(key_node) || ts_node_is_null(val_node)) {
+                            continue;
+                        }
+
+                        char *key = cbm_node_text(ctx->arena, key_node, ctx->source);
+                        if (!key) {
+                            continue;
+                        }
+
+                        /* Check if key matches URL or topic patterns */
+                        bool is_url_kw = false;
+                        bool is_topic_kw = false;
+                        for (int ki = 0; url_keywords[ki]; ki++) {
+                            if (strcmp(key, url_keywords[ki]) == 0) {
+                                is_url_kw = true;
+                                break;
+                            }
+                        }
+                        if (!is_url_kw) {
+                            for (int ki = 0; topic_keywords[ki]; ki++) {
+                                if (strcmp(key, topic_keywords[ki]) == 0) {
+                                    is_topic_kw = true;
+                                    break;
+                                }
+                            }
+                        }
+                        if (!is_url_kw && !is_topic_kw) {
+                            continue;
+                        }
+
+                        /* Extract value — string literal or constant reference */
+                        const char *vk = ts_node_type(val_node);
+                        if (strcmp(vk, "string") == 0 || strcmp(vk, "string_literal") == 0 ||
+                            strcmp(vk, "interpreted_string_literal") == 0 ||
+                            strcmp(vk, "raw_string_literal") == 0) {
+                            char *text = cbm_node_text(ctx->arena, val_node, ctx->source);
+                            if (text && text[0]) {
+                                int len = (int)strlen(text);
+                                if (len >= 2 && (text[0] == '"' || text[0] == '\'')) {
+                                    text =
+                                        cbm_arena_strndup(ctx->arena, text + 1, (size_t)(len - 2));
+                                }
+                                if (text && text[0]) {
+                                    call.first_string_arg = text;
+                                }
+                            }
+                        } else if (strcmp(vk, "identifier") == 0) {
+                            /* Constant reference: url=MY_URL_CONSTANT */
+                            char *const_name = cbm_node_text(ctx->arena, val_node, ctx->source);
+                            if (const_name) {
+                                const char *resolved = lookup_string_constant(ctx, const_name);
+                                if (resolved) {
+                                    call.first_string_arg = resolved;
+                                }
+                            }
+                        }
+                        continue;
+                    }
+
+                    /* First positional string argument (fallback) */
+                    if (ai < 3) {
+                        if (strcmp(ak, "string") == 0 || strcmp(ak, "string_literal") == 0 ||
+                            strcmp(ak, "interpreted_string_literal") == 0 ||
+                            strcmp(ak, "raw_string_literal") == 0) {
+                            char *text = cbm_node_text(ctx->arena, arg, ctx->source);
+                            if (text && text[0]) {
+                                int len = (int)strlen(text);
+                                if (len >= 2 && (text[0] == '"' || text[0] == '\'')) {
+                                    text =
+                                        cbm_arena_strndup(ctx->arena, text + 1, (size_t)(len - 2));
+                                    len -= 2;
+                                }
+                                /* Validate printable */
+                                // NOLINTNEXTLINE(readability-implicit-bool-conversion)
+                                bool valid = (text != NULL && len > 0 && len < 512);
+                                for (int vi = 0; vi < len && valid; vi++) {
+                                    if ((unsigned char)text[vi] < 0x20 && text[vi] != '\t') {
+                                        valid = false;
+                                    }
+                                }
+                                if (valid) {
+                                    call.first_string_arg = text;
+                                }
+                            }
+                            break;
+                        }
+                        /* Positional constant reference: create_task(MY_URL) */
+                        if (strcmp(ak, "identifier") == 0) {
+                            char *const_name = cbm_node_text(ctx->arena, arg, ctx->source);
+                            if (const_name) {
+                                const char *resolved = lookup_string_constant(ctx, const_name);
+                                if (resolved) {
+                                    call.first_string_arg = resolved;
+                                    break;
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+
             cbm_calls_push(&ctx->result->calls, ctx->arena, call);
         }
     }
@@ -336,7 +516,7 @@ void handle_calls(CBMExtractCtx *ctx, TSNode node, const CBMLangSpec *spec, Walk
             if (!ts_node_is_null(name_node)) {
                 char *name = cbm_node_text(ctx->arena, name_node, ctx->source);
                 if (name && name[0] >= 'A' && name[0] <= 'Z') {
-                    CBMCall call;
+                    CBMCall call = {0};
                     call.callee_name = name;
                     call.enclosing_func_qn = state->enclosing_func_qn;
                     cbm_calls_push(&ctx->result->calls, ctx->arena, call);