feat(cross-repo): search, flow tracing with call chains, and impact analysis

Adds three major cross-repo capabilities building on the _cross_repo.db
infrastructure from the previous commit:

1. Cross-repo search (search_graph with project='*'):
   Dispatches to cbm_cross_repo_search() which runs BM25+vector+RRF
   against the unified cross-repo index (134K nodes, 134K embeddings).
   Returns results with both short project name and full project_id
   for follow-up per-repo queries.

2. Enhanced trace_cross_repo with call chains:
   When a channel filter is provided, traces upstream callers of the
   emitter and downstream callees of the listener, depth 2 per hop.
   New cbm_cross_repo_trace_in_project() helper opens project DBs
   ad-hoc, resolves Class→Method and (file-level) listeners, runs
   cbm_store_bfs, returns structured step arrays.
   Without channel filter: returns channel matches only (unchanged).

3. Cross-repo impact analysis (get_impact with cross_repo=true):
   After per-repo BFS, checks if any d=1 impacted symbols emit
   channels in cross_channels. For each affected channel, opens
   consumer project DB, traces downstream from the listener function,
   returns cross_repo_impacts array with consumer repo, listener
   function, downstream affected count, and trace steps.

New functions in cross_repo.c:
- cbm_cross_repo_trace_in_project: open DB, resolve function,
  BFS, return steps. Handles Class→Method resolution and file-level
  listener fallback via channels table node_id lookup.
- cbm_cross_trace_free: cleanup for trace step arrays.

Changes in mcp.c:
- handle_search_graph: early-exit to handle_cross_repo_search when
  project='*'
- handle_cross_repo_search: new handler with hybrid search + JSON output
- handle_trace_cross_repo: enhanced with call chain tracing per match
- handle_get_impact: cross_repo=true triggers channel + consumer trace

No new dependencies. Uses existing cbm_store_bfs and cbm_store_open_path_query.

Author: Your Name

src/mcp/mcp.c

@@ -1000,8 +1000,21 @@ static char *handle_get_graph_schema(cbm_mcp_server_t *srv, const char *args) {
     return result;
 }
 
+/* Forward declarations — defined in cross-repo handler section */
+static char *handle_cross_repo_search(cbm_mcp_server_t *srv, const char *args);
+static char *derive_short_project(const char *full_project);
+static void add_trace_steps(yyjson_mut_doc *doc, yyjson_mut_val *parent,
+                            const char *key, cbm_cross_trace_step_t *steps, int count);
+
 static char *handle_search_graph(cbm_mcp_server_t *srv, const char *args) {
     char *project = cbm_mcp_get_string_arg(args, "project");
+
+    /* Cross-repo search: project="*" dispatches to unified _cross_repo.db */
+    if (project && strcmp(project, "*") == 0) {
+        free(project);
+        return handle_cross_repo_search(srv, args);
+    }
+
     cbm_store_t *store = resolve_store(srv, project);
     REQUIRE_STORE(store, project);
 
@@ -1422,6 +1435,7 @@ static char *handle_get_impact(cbm_mcp_server_t *srv, const char *args) {
     char *target = cbm_mcp_get_string_arg(args, "target");
     char *direction = cbm_mcp_get_string_arg(args, "direction");
     int max_depth = cbm_mcp_get_int_arg(args, "max_depth", 3);
+    bool cross_repo = cbm_mcp_get_bool_arg(args, "cross_repo");
     cbm_store_t *store = resolve_store(srv, project);
     REQUIRE_STORE(store, project);
 
@@ -1661,6 +1675,83 @@ static char *handle_get_impact(cbm_mcp_server_t *srv, const char *args) {
         cbm_store_free_processes(procs, pcount);
     }
 
+    /* ── Cross-repo impact: check if d=1 nodes emit channels to other repos ── */
+    if (cross_repo && tr.visited_count > 0) {
+        cbm_cross_repo_t *cr = cbm_cross_repo_open();
+        if (cr) {
+            yyjson_mut_val *xr_arr = yyjson_mut_arr(doc);
+            int xr_count = 0;
+
+            /* For each d=1 visited node, check cross_channels for emitters */
+            for (int vi = 0; vi < tr.visited_count && xr_count < 10; vi++) {
+                if (tr.visited[vi].hop != 1) continue; /* only d=1 */
+                const char *vname = tr.visited[vi].node.name;
+                if (!vname) continue;
+
+                /* Query cross_channels for this function as an emitter */
+                cbm_cross_channel_match_t *xmatches = NULL;
+                int xmatch_count = 0;
+                /* Use the function name to filter — imprecise but functional */
+                cbm_cross_repo_match_channels(cr, NULL, &xmatches, &xmatch_count);
+
+                for (int xi = 0; xi < xmatch_count && xr_count < 10; xi++) {
+                    /* Match: emitter function name matches d=1 visited node */
+                    if (!xmatches[xi].emit_function) continue;
+                    if (strcmp(xmatches[xi].emit_function, vname) != 0) continue;
+                    /* And the emitter project matches our project */
+                    if (!xmatches[xi].emit_project || !project) continue;
+                    if (strcmp(xmatches[xi].emit_project, project) != 0) continue;
+
+                    /* Found a cross-repo channel triggered by this d=1 node */
+                    yyjson_mut_val *xr_item = yyjson_mut_obj(doc);
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "channel",
+                        xmatches[xi].channel_name ? xmatches[xi].channel_name : "");
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "transport",
+                        xmatches[xi].transport ? xmatches[xi].transport : "");
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "triggered_by", vname);
+                    yyjson_mut_obj_add_int(doc, xr_item, "triggered_by_depth", 1);
+
+                    char *lp_short = derive_short_project(
+                        xmatches[xi].listen_project ? xmatches[xi].listen_project : "");
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "consumer_repo",
+                                             lp_short ? lp_short : "");
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "consumer_project_id",
+                        xmatches[xi].listen_project ? xmatches[xi].listen_project : "");
+                    free(lp_short);
+
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "listener_function",
+                        xmatches[xi].listen_function ? xmatches[xi].listen_function : "");
+                    yyjson_mut_obj_add_strcpy(doc, xr_item, "listener_file",
+                        xmatches[xi].listen_file ? xmatches[xi].listen_file : "");
+
+                    /* Trace downstream in consumer repo */
+                    char db_path_buf[2048];
+                    project_db_path(xmatches[xi].listen_project, db_path_buf, sizeof(db_path_buf));
+                    cbm_cross_trace_step_t *ds_steps = NULL;
+                    int ds_count = 0;
+                    cbm_cross_repo_trace_in_project(db_path_buf,
+                        xmatches[xi].listen_function, xmatches[xi].listen_file,
+                        xmatches[xi].channel_name, "outbound", 2, &ds_steps, &ds_count);
+                    yyjson_mut_obj_add_int(doc, xr_item, "downstream_affected", ds_count);
+                    if (ds_count > 0) {
+                        add_trace_steps(doc, xr_item, "downstream", ds_steps, ds_count);
+                    }
+                    cbm_cross_trace_free(ds_steps, ds_count);
+
+                    yyjson_mut_arr_add_val(xr_arr, xr_item);
+                    xr_count++;
+                }
+                cbm_cross_channel_free(xmatches, xmatch_count);
+            }
+
+            if (xr_count > 0) {
+                yyjson_mut_obj_add_val(doc, root, "cross_repo_impacts", xr_arr);
+                yyjson_mut_obj_add_int(doc, root, "cross_repo_impact_count", xr_count);
+            }
+            cbm_cross_repo_close(cr);
+        }
+    }
+
     char *json = yy_doc_to_str(doc);
     yyjson_mut_doc_free(doc);
     cbm_store_traverse_free(&tr);
@@ -3890,6 +3981,95 @@ static char *handle_generate_embeddings(cbm_mcp_server_t *srv, const char *args)
 
 /* ── build_cross_repo_index ──────────────────────────────────── */
 
+/* ── Cross-repo search (search_graph with project="*") ───────── */
+
+static char *derive_short_project(const char *full_project) {
+    /* "mnt-c-Users-Name-Projects-repo-name" → "repo-name" */
+    const char *marker = strstr(full_project, "Projects-");
+    if (marker) return heap_strdup(marker + 9);
+    return heap_strdup(full_project);
+}
+
+static char *handle_cross_repo_search(cbm_mcp_server_t *srv, const char *args) {
+    (void)srv;
+
+    cbm_cross_repo_t *cr = cbm_cross_repo_open();
+    if (!cr) {
+        return cbm_mcp_text_result(
+            "{\"error\":\"Cross-repo index not built. Run build_cross_repo_index first.\"}", true);
+    }
+
+    char *query = cbm_mcp_get_string_arg(args, "query");
+    int limit = cbm_mcp_get_int_arg(args, "limit", 30);
+    if (limit <= 0) limit = 30;
+
+    if (!query || !query[0]) {
+        cbm_cross_repo_close(cr);
+        free(query);
+        return cbm_mcp_text_result(
+            "{\"error\":\"query parameter required for cross-repo search\"}", true);
+    }
+
+    /* Embed query for hybrid search if configured */
+    float *query_vec = NULL;
+    int dims = 0;
+    if (cbm_embedding_is_configured()) {
+        cbm_embedding_config_t cfg = cbm_embedding_get_config();
+        query_vec = cbm_embedding_embed_text(&cfg, query);
+        dims = cfg.dims;
+    }
+
+    cbm_cross_search_output_t out = {0};
+    cbm_cross_repo_search(cr, query, query_vec, dims, limit, &out);
+
+    yyjson_mut_doc *doc = yyjson_mut_doc_new(NULL);
+    yyjson_mut_val *root = yyjson_mut_obj(doc);
+    yyjson_mut_doc_set_root(doc, root);
+
+    yyjson_mut_obj_add_int(doc, root, "total", out.total);
+    yyjson_mut_obj_add_str(doc, root, "search_mode",
+                           out.used_vector ? "hybrid_bm25_vector" : "bm25");
+    yyjson_mut_obj_add_bool(doc, root, "cross_repo", true);
+
+    yyjson_mut_val *results = yyjson_mut_arr(doc);
+    for (int i = 0; i < out.count; i++) {
+        cbm_cross_search_result_t *r = &out.results[i];
+        yyjson_mut_val *item = yyjson_mut_obj(doc);
+
+        /* Short project name for display */
+        char *short_proj = derive_short_project(r->project ? r->project : "");
+        yyjson_mut_obj_add_strcpy(doc, item, "project", short_proj ? short_proj : "");
+        /* Full project ID for follow-up calls */
+        yyjson_mut_obj_add_strcpy(doc, item, "project_id", r->project ? r->project : "");
+        free(short_proj);
+
+        yyjson_mut_obj_add_strcpy(doc, item, "name", r->name ? r->name : "");
+        yyjson_mut_obj_add_strcpy(doc, item, "qualified_name",
+                                  r->qualified_name ? r->qualified_name : "");
+        yyjson_mut_obj_add_strcpy(doc, item, "label", r->label ? r->label : "");
+        yyjson_mut_obj_add_strcpy(doc, item, "file_path", r->file_path ? r->file_path : "");
+        yyjson_mut_obj_add_real(doc, item, "score", r->score);
+        if (r->similarity > 0)
+            yyjson_mut_obj_add_real(doc, item, "similarity", r->similarity);
+
+        yyjson_mut_arr_add_val(results, item);
+    }
+    yyjson_mut_obj_add_val(doc, root, "results", results);
+
+    char *json = yy_doc_to_str(doc);
+    yyjson_mut_doc_free(doc);
+    cbm_cross_search_free(&out);
+    cbm_cross_repo_close(cr);
+    free(query);
+    free(query_vec);
+
+    char *result = cbm_mcp_text_result(json, false);
+    free(json);
+    return result;
+}
+
+/* ── build_cross_repo_index ──────────────────────────────────── */
+
 static char *handle_build_cross_repo_index(cbm_mcp_server_t *srv, const char *args) {
     (void)srv; (void)args;
 
@@ -3918,9 +4098,26 @@ static char *handle_build_cross_repo_index(cbm_mcp_server_t *srv, const char *ar
 
 /* ── trace_cross_repo ────────────────────────────────────────── */
 
+/* Add trace steps as a JSON array to a parent object */
+static void add_trace_steps(yyjson_mut_doc *doc, yyjson_mut_val *parent,
+                            const char *key, cbm_cross_trace_step_t *steps, int count) {
+    yyjson_mut_val *arr = yyjson_mut_arr(doc);
+    for (int i = 0; i < count; i++) {
+        yyjson_mut_val *step = yyjson_mut_obj(doc);
+        yyjson_mut_obj_add_strcpy(doc, step, "name", steps[i].name ? steps[i].name : "");
+        yyjson_mut_obj_add_strcpy(doc, step, "label", steps[i].label ? steps[i].label : "");
+        yyjson_mut_obj_add_strcpy(doc, step, "file_path",
+                                  steps[i].file_path ? steps[i].file_path : "");
+        yyjson_mut_obj_add_int(doc, step, "depth", steps[i].depth);
+        yyjson_mut_arr_add_val(arr, step);
+    }
+    yyjson_mut_obj_add_val(doc, parent, key, arr);
+}
+
 static char *handle_trace_cross_repo(cbm_mcp_server_t *srv, const char *args) {
     (void)srv;
     char *channel = cbm_mcp_get_string_arg(args, "channel");
+    bool trace_calls = (channel && channel[0]); /* only trace call chains when channel filter given */
 
     cbm_cross_repo_t *cr = cbm_cross_repo_open();
     if (!cr) {
@@ -3945,6 +4142,7 @@ static char *handle_trace_cross_repo(cbm_mcp_server_t *srv, const char *args) {
     yyjson_mut_obj_add_int(doc, root, "total_repos", info.total_repos);
     yyjson_mut_obj_add_int(doc, root, "total_cross_repo_channels", info.cross_repo_channel_count);
     yyjson_mut_obj_add_int(doc, root, "matches", match_count);
+    yyjson_mut_obj_add_bool(doc, root, "call_chains_included", trace_calls);
     if (info.built_at)
         yyjson_mut_obj_add_strcpy(doc, root, "built_at", info.built_at);
 
@@ -3955,16 +4153,52 @@ static char *handle_trace_cross_repo(cbm_mcp_server_t *srv, const char *args) {
         yyjson_mut_obj_add_strcpy(doc, item, "channel", m->channel_name ? m->channel_name : "");
         yyjson_mut_obj_add_strcpy(doc, item, "transport", m->transport ? m->transport : "");
 
+        /* Emitter side */
         yyjson_mut_val *emit = yyjson_mut_obj(doc);
-        yyjson_mut_obj_add_strcpy(doc, emit, "project", m->emit_project ? m->emit_project : "");
+        char *ep_short = derive_short_project(m->emit_project ? m->emit_project : "");
+        yyjson_mut_obj_add_strcpy(doc, emit, "project", ep_short ? ep_short : "");
+        free(ep_short);
         yyjson_mut_obj_add_strcpy(doc, emit, "file", m->emit_file ? m->emit_file : "");
         yyjson_mut_obj_add_strcpy(doc, emit, "function", m->emit_function ? m->emit_function : "");
+
+        /* Trace upstream callers of the emitter (what triggers the emit) */
+        if (trace_calls && m->emit_project) {
+            char db_path[2048];
+            project_db_path(m->emit_project, db_path, sizeof(db_path));
+            cbm_cross_trace_step_t *steps = NULL;
+            int step_count = 0;
+            cbm_cross_repo_trace_in_project(db_path, m->emit_function,
+                                            m->emit_file, m->channel_name,
+                                            "inbound", 2, &steps, &step_count);
+            if (step_count > 0) {
+                add_trace_steps(doc, emit, "upstream", steps, step_count);
+            }
+            cbm_cross_trace_free(steps, step_count);
+        }
         yyjson_mut_obj_add_val(doc, item, "emitter", emit);
 
+        /* Listener side */
         yyjson_mut_val *listen = yyjson_mut_obj(doc);
-        yyjson_mut_obj_add_strcpy(doc, listen, "project", m->listen_project ? m->listen_project : "");
+        char *lp_short = derive_short_project(m->listen_project ? m->listen_project : "");
+        yyjson_mut_obj_add_strcpy(doc, listen, "project", lp_short ? lp_short : "");
+        free(lp_short);
         yyjson_mut_obj_add_strcpy(doc, listen, "file", m->listen_file ? m->listen_file : "");
         yyjson_mut_obj_add_strcpy(doc, listen, "function", m->listen_function ? m->listen_function : "");
+
+        /* Trace downstream callees of the listener (what the listener triggers) */
+        if (trace_calls && m->listen_project) {
+            char db_path[2048];
+            project_db_path(m->listen_project, db_path, sizeof(db_path));
+            cbm_cross_trace_step_t *steps = NULL;
+            int step_count = 0;
+            cbm_cross_repo_trace_in_project(db_path, m->listen_function,
+                                            m->listen_file, m->channel_name,
+                                            "outbound", 2, &steps, &step_count);
+            if (step_count > 0) {
+                add_trace_steps(doc, listen, "downstream", steps, step_count);
+            }
+            cbm_cross_trace_free(steps, step_count);
+        }
         yyjson_mut_obj_add_val(doc, item, "listener", listen);
 
         yyjson_mut_arr_add_val(arr, item);