FalkorDB · shahar-biron · Jul 12, 2026 · Jul 9, 2026 · Jul 12, 2026 · Jul 12, 2026
diff --git a/QUERY_EXPLANATIONS_AND_SAMPLES.md b/QUERY_EXPLANATIONS_AND_SAMPLES.md
@@ -1,11 +1,16 @@
 # Query Explanations & Samples
-This document covers the full query catalog currently generated by `UsersQueriesRepository` in `src/queries_repository.rs`, including core queries, optional algorithm queries, and phase-1 additions.
+This document covers the full query catalog currently generated by `UsersQueriesRepository` in `src/queries_repository.rs`, including core queries, optional algorithm queries, phase-1 additions, and profile-gated extended/fixture-dependent additions.
 
 ## Dataset assumptions
 - Primary label: `:User`
 - Primary relationship: `:Friend`
 - Common properties used by queries: `id`, `age`, and `bench_capacity`
 
+## Coverage profiles
+- `baseline`: core + phase-1 query set.
+- `extended-core`: baseline + `temporal_spatial_roundtrip` (where vendor support exists).
+- `fixture-dependent`: extended-core + text/vector smoke queries that require fixture seeding and additional indexes.
+
 ## Complete query catalog (alphabetical by query id)
 - `aggregate_age`: average age over all users.
 - `aggregate_age_distinct`: count distinct age values.
@@ -33,6 +38,8 @@ This document covers the full query catalog currently generated by `UsersQueries
 - `exact_5_hop_traverse_count`: fixed 5-hop traversal.
 - `exact_6_hop_traverse_count`: fixed 6-hop traversal.
 - `foreach_loop_mutation`: `FOREACH` mutation-loop coverage.
+- `fulltext_query_nodes_smoke`: fixture-dependent fulltext query on nodes.
+- `fulltext_query_relationships_smoke`: fixture-dependent fulltext query on relationships.
 - `id_range_scan`: internal-id range scan.
 - `id_seek`: internal-id point lookup.
 - `indexed_in_list_predicate`: IN-list index utilization shape.
@@ -57,13 +64,15 @@ This document covers the full query catalog currently generated by `UsersQueries
 - `single_vertex_read`: point lookup by user id.
 - `single_vertex_update`: update `rpc_social_credit` on a user.
 - `single_vertex_write`: create one `User` node.
+- `temporal_spatial_roundtrip`: extended-core temporal/spatial scalar-function roundtrip query.
 - `union_all_ids`: `UNION ALL` composition.
 - `union_distinct_ids`: `UNION` (distinct) composition.
 - `unwind_rows`: row-level list fan-out with `UNWIND`.
 - `value_join`: value-based join on `age`.
 - `value_join_cnt`: count variant of the value join.
 - `var_len_friends`: variable-length traversal (`*1..2`).
 - `var_len_with_edge_where_filter`: variable-length traversal with edge filtering.
+- `vector_query_nodes_smoke`: fixture-dependent vector-search query on nodes.
 - `vertex_on_label_property`: lookup by label + property.
 - `vertex_on_label_property_index`: same shape for index-path benchmarking.
 - `vertex_on_property`: lookup by property without label predicate.
@@ -92,9 +101,22 @@ RETURN labels(a), type(r), properties(a), nodes(p), relationships(p), length(p)
 LIMIT 1
 ```
 
+```cypher
+// Neo4j
+CALL db.index.vector.queryNodes('bench_user_embedding_idx', 10, [0.1, 0.2, 0.3])
+YIELD node, score
+RETURN id(node), score
+LIMIT 10
+```
+
 ## Vendor-specific notes
 - `shortest_path`, `shortest_path_with_filter`, and `all_shortest_paths_len` have vendor-specific query text.
 - `aggregate_count_users` uses a Falkor-optimized `db.meta.stats()` path.
+- `temporal_spatial_roundtrip` is currently generated only for FalkorDB and Neo4j.
+- Fixture-dependent text/vector smoke queries (`vector_query_nodes_smoke`, `fulltext_query_nodes_smoke`, `fulltext_query_relationships_smoke`) use vendor-specific procedures/index names:
+  - FalkorDB: `db.idx.vector.*`, `db.idx.fulltext.*`
+  - Neo4j: `db.index.vector.*`, `db.index.fulltext.*`
+  - Memgraph: `vector_search.*`, `text_search.*`
 - `algo_max_flow_single_pair` in Falkor uses one relationship type in config:
   - gets a single value from `db.relationshipTypes()`
   - passes `relationshipTypes: [relationshipType]`
diff --git a/readme.md b/readme.md
@@ -193,6 +193,11 @@ NOTE: It is possible to use the load command with externally run vendor endpoint
 - `cargo run --release --bin benchmark -- load --vendor neo4j -s small --endpoint neo4j://neo4j:benchmark123@127.0.0.1:7687`
 - `cargo run --release --bin benchmark -- load --vendor memgraph -s small --endpoint bolt://127.0.0.1:7687`
 
+Profile-aware loading (runs additional fixture/index setup when required):
+- `cargo run --release --bin benchmark -- load --vendor neo4j -s small --query-profile fixture-dependent`
+- `cargo run --release --bin benchmark -- load --vendor memgraph -s small --query-profile fixture-dependent`
+- `cargo run --release --bin benchmark -- load --vendor falkor -s small --query-profile fixture-dependent`
+
 ##### create a set of queries to be used with the run command
 
 -
@@ -203,6 +208,11 @@ NOTE: preparing a smaller run of 1,000,000 queries:
 
 `cargo run --release --bin benchmark -- generate-queries  -s1000000 --dataset small --name=small-readonly --write-ratio 0.0`
 
+Generate with a broader coverage profile:
+
+- `cargo run --release --bin benchmark -- generate-queries -s1000000 --dataset small --name=small-extended --write-ratio 0.0 --vendor neo4j --query-profile extended-core`
+- `cargo run --release --bin benchmark -- generate-queries -s1000000 --dataset small --name=small-fixtures --write-ratio 0.0 --vendor memgraph --query-profile fixture-dependent`
+
 ##### run the benchmarks
 
 - `cargo run --release --bin benchmark run --vendor falkor --name small-readonly -p40 --mps 4000`

diff --git a/scripts/run_large_benchmark.sh b/scripts/run_large_benchmark.sh
@@ -35,6 +35,11 @@ SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
 #  ENABLE_ALGO_MSF (default: 0)
 #  ENABLE_ALGO_HARMONIC (default: 0)
 #  FREE_PRIMARY_FALKOR_BEFORE_SECOND_RUN (default: 1)
+#  IN_SCRIPT_QUERY_PROFILE (default: empty; set in this file to force profile)
+#    baseline          - baseline query set
+#    extended-core     - baseline + extended core queries
+#    fixture-dependent - extended-core + fixture/index-dependent queries
+#  QUERY_PROFILE (env fallback when IN_SCRIPT_QUERY_PROFILE is empty; default: baseline)
 #
 # Results:
 #  RESULTS_DIR (default: Results-YYMMDD-HH:MM)
@@ -79,6 +84,23 @@ ENABLE_ALGO_MAX_FLOW=${ENABLE_ALGO_MAX_FLOW:-0}
 ENABLE_ALGO_MSF=${ENABLE_ALGO_MSF:-0}
 ENABLE_ALGO_HARMONIC=${ENABLE_ALGO_HARMONIC:-0}
 FREE_PRIMARY_FALKOR_BEFORE_SECOND_RUN=${FREE_PRIMARY_FALKOR_BEFORE_SECOND_RUN:-1}
+# Optional in-script query profile override.
+# Set this value directly in the script to force a profile for every run.
+# Leave empty ("") to keep env-based behavior (QUERY_PROFILE env var, else baseline).
+IN_SCRIPT_QUERY_PROFILE=""
+if [[ -n "$IN_SCRIPT_QUERY_PROFILE" ]]; then
+  QUERY_PROFILE="$IN_SCRIPT_QUERY_PROFILE"
+else
+  QUERY_PROFILE=${QUERY_PROFILE:-baseline}
+fi
+
+case "$QUERY_PROFILE" in
+  baseline|extended-core|fixture-dependent) ;;
+  *)
+    echo "Invalid QUERY_PROFILE '$QUERY_PROFILE'. Valid options: baseline, extended-core, fixture-dependent." >&2
+    exit 1
+    ;;
+esac
 
 # Derive per-vendor query file names so each engine can use vendor-optimized queries.
 QUERIES_FILE_BASE="${QUERIES_FILE}"
@@ -376,7 +398,7 @@ if [[ "${RUN_FALKOR}" == "1" ]]; then
   PYTHONPATH="../falkordb-bulk-loader" python3 "../falkordb-bulk-loader/falkordb_bulk_loader/bulk_insert.py" falkor \
     -u "$REDIS_URL" \
     -n "$CSV_DIR/User.csv" \
-    -r "$CSV_DIR/FRIEND.csv" \
+    -R Friend "$CSV_DIR/FRIEND.csv" \
     -j INTEGER -s -i User:id -i User:age \
     -c 128 -b 16 -t 16
 fi
@@ -461,7 +483,7 @@ fi
 
 if [[ "${RUN_NEO4J}" == "1" ]]; then
   echo "==> Loading large dataset into Neo4j"
-  cargo run --release --bin benchmark -- load --vendor neo4j --size large --endpoint "$NEO4J_ENDPOINT" -b "$BATCH_SIZE"
+  cargo run --release --bin benchmark -- load --vendor neo4j --size large --endpoint "$NEO4J_ENDPOINT" -b "$BATCH_SIZE" --query-profile "$QUERY_PROFILE"
 fi
 
 if [[ "${RUN_FALKOR}" == "1" ]]; then
@@ -536,17 +558,17 @@ if [[ "${RUN_MEMGRAPH}" == "1" ]]; then
   fi
 fi
 
-echo "==> Generating vendor-specific query files (base=${QUERIES_FILE_BASE}, dataset=large, count=${QUERIES_COUNT}, write_ratio=${WRITE_RATIO})"
+echo "==> Generating vendor-specific query files (base=${QUERIES_FILE_BASE}, dataset=large, count=${QUERIES_COUNT}, write_ratio=${WRITE_RATIO}, profile=${QUERY_PROFILE})"
 echo "==> Algorithm query toggles (pagerank=${ENABLE_ALGO_PAGERANK_BOOL}, max_flow=${ENABLE_ALGO_MAX_FLOW_BOOL}, msf=${ENABLE_ALGO_MSF_BOOL}, harmonic=${ENABLE_ALGO_HARMONIC_BOOL})"
 # Always regenerate so each selected vendor gets the latest query catalog + stable q_id fields.
 if [[ "${RUN_FALKOR}" == "1" || "${RUN_FALKOR_2}" == "1" ]]; then
-  cargo run --release --bin benchmark -- generate-queries --vendor falkor   --dataset large --size "$QUERIES_COUNT" --name "$FALKOR_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" "${ALGO_QUERY_ARGS[@]}"
+  cargo run --release --bin benchmark -- generate-queries --vendor falkor   --dataset large --size "$QUERIES_COUNT" --name "$FALKOR_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" --query-profile "$QUERY_PROFILE" "${ALGO_QUERY_ARGS[@]}"
 fi
 if [[ "${RUN_NEO4J}" == "1" ]]; then
-  cargo run --release --bin benchmark -- generate-queries --vendor neo4j   --dataset large --size "$QUERIES_COUNT" --name "$NEO4J_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" "${ALGO_QUERY_ARGS[@]}"
+  cargo run --release --bin benchmark -- generate-queries --vendor neo4j   --dataset large --size "$QUERIES_COUNT" --name "$NEO4J_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" --query-profile "$QUERY_PROFILE" "${ALGO_QUERY_ARGS[@]}"
 fi
 if [[ "${RUN_MEMGRAPH}" == "1" ]]; then
-  cargo run --release --bin benchmark -- generate-queries --vendor memgraph --dataset large --size "$QUERIES_COUNT" --name "$MEMGRAPH_QUERIES_FILE" --write-ratio "$WRITE_RATIO" "${ALGO_QUERY_ARGS[@]}"
+  cargo run --release --bin benchmark -- generate-queries --vendor memgraph --dataset large --size "$QUERIES_COUNT" --name "$MEMGRAPH_QUERIES_FILE" --write-ratio "$WRITE_RATIO" --query-profile "$QUERY_PROFILE" "${ALGO_QUERY_ARGS[@]}"
 fi
 
 echo "==> Running ${QUERIES_FILE} workload (parallel=${PARALLEL}, mps=${MPS})"
@@ -580,7 +602,7 @@ if [[ "${RUN_FALKOR_2}" == "1" ]]; then
   PYTHONPATH="../falkordb-bulk-loader" python3 "../falkordb-bulk-loader/falkordb_bulk_loader/bulk_insert.py" falkor \
     -u "$REDIS_URL_2" \
     -n "$CSV_DIR/User.csv" \
-    -r "$CSV_DIR/FRIEND.csv" \
+    -R Friend "$CSV_DIR/FRIEND.csv" \
     -j INTEGER -s -i User:id -i User:age \
     -c 128 -b 16 -t 16
 

diff --git a/scripts/run_medium_benchmark.sh b/scripts/run_medium_benchmark.sh
@@ -34,6 +34,11 @@ SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
 #  ENABLE_ALGO_MAX_FLOW (default: 1)
 #  ENABLE_ALGO_MSF (default: 1)
 #  ENABLE_ALGO_HARMONIC (default: 1)
+#  IN_SCRIPT_QUERY_PROFILE (default: empty; set in this file to force profile)
+#    baseline          - baseline query set
+#    extended-core     - baseline + extended core queries
+#    fixture-dependent - extended-core + fixture/index-dependent queries
+#  QUERY_PROFILE (env fallback when IN_SCRIPT_QUERY_PROFILE is empty; default: baseline)
 #
 # Results:
 #  RESULTS_DIR (default: Results-YYMMDD-HH:MM)
@@ -63,13 +68,30 @@ BATCH_SIZE=${BATCH_SIZE:-5000}
 PARALLEL=${PARALLEL:-10}
 MPS=${MPS:-3000}
 QUERIES_FILE=${QUERIES_FILE:-"medium-readonly"}
-QUERIES_COUNT=${QUERIES_COUNT:-25000}
+QUERIES_COUNT=${QUERIES_COUNT:-2500}
 WRITE_RATIO=${WRITE_RATIO:-0.05}
 FALKOR_QUERY_TIMEOUT_MS=${FALKOR_QUERY_TIMEOUT_MS:-900000}
-ENABLE_ALGO_PAGERANK=${ENABLE_ALGO_PAGERANK:-1}
-ENABLE_ALGO_MAX_FLOW=${ENABLE_ALGO_MAX_FLOW:-1}
-ENABLE_ALGO_MSF=${ENABLE_ALGO_MSF:-1}
-ENABLE_ALGO_HARMONIC=${ENABLE_ALGO_HARMONIC:-1}
+ENABLE_ALGO_PAGERANK=${ENABLE_ALGO_PAGERANK:-0}
+ENABLE_ALGO_MAX_FLOW=${ENABLE_ALGO_MAX_FLOW:-0}
+ENABLE_ALGO_MSF=${ENABLE_ALGO_MSF:-0}
+ENABLE_ALGO_HARMONIC=${ENABLE_ALGO_HARMONIC:-0}
+# Optional in-script query profile override.
+# Set this value directly in the script to force a profile for every run.
+# Leave empty ("") to keep env-based behavior (QUERY_PROFILE env var, else baseline).
+IN_SCRIPT_QUERY_PROFILE="fixture-dependent"
+if [[ -n "$IN_SCRIPT_QUERY_PROFILE" ]]; then
+  QUERY_PROFILE="$IN_SCRIPT_QUERY_PROFILE"
+else
+  QUERY_PROFILE=${QUERY_PROFILE:-baseline}
+fi
+
+case "$QUERY_PROFILE" in
+  baseline|extended-core|fixture-dependent) ;;
+  *)
+    echo "Invalid QUERY_PROFILE '$QUERY_PROFILE'. Valid options: baseline, extended-core, fixture-dependent." >&2
+    exit 1
+    ;;
+esac
 
 # Derive per-vendor query file names so each engine can use vendor-optimized queries.
 QUERIES_FILE_BASE="${QUERIES_FILE}"
@@ -246,7 +268,7 @@ if [[ "${RUN_FALKOR}" == "1" ]]; then
   PYTHONPATH="../falkordb-bulk-loader" python3 ../falkordb-bulk-loader/falkordb_bulk_loader/bulk_insert.py falkor \
     -u "$REDIS_URL" \
     -n "$CSV_DIR/User.csv" \
-    -r "$CSV_DIR/FRIEND.csv" \
+    -R Friend "$CSV_DIR/FRIEND.csv" \
     -j INTEGER -s -i User:id -i User:age \
     -c 128 -b 16 -t 16
 fi
@@ -262,20 +284,20 @@ if [[ "${RUN_FALKOR_2}" == "1" ]]; then
   PYTHONPATH="../falkordb-bulk-loader" python3 ../falkordb-bulk-loader/falkordb_bulk_loader/bulk_insert.py falkor \
     -u "$REDIS_URL_2" \
     -n "$CSV_DIR/User.csv" \
-    -r "$CSV_DIR/FRIEND.csv" \
+    -R Friend "$CSV_DIR/FRIEND.csv" \
     -j INTEGER -s -i User:id -i User:age \
     -c 128 -b 16 -t 16
 fi
 
 if [[ "${RUN_NEO4J}" == "1" ]]; then
   echo "==> Loading medium dataset into Neo4j"
-  cargo run --release --bin benchmark -- load --vendor neo4j --size medium --endpoint "$NEO4J_ENDPOINT" -b "$BATCH_SIZE"
+  cargo run --release --bin benchmark -- load --vendor neo4j --size medium --endpoint "$NEO4J_ENDPOINT" -b "$BATCH_SIZE" --query-profile "$QUERY_PROFILE"
 fi
 
 if [[ "${RUN_MEMGRAPH}" == "1" ]]; then
   echo "==> Loading medium dataset into Memgraph (UNWIND loader)"
   # --force clears the external Memgraph instance before loading
-  cargo run --release --bin benchmark -- load --vendor memgraph --size medium --endpoint "$MEMGRAPH_ENDPOINT" -b "$BATCH_SIZE" --force
+  cargo run --release --bin benchmark -- load --vendor memgraph --size medium --endpoint "$MEMGRAPH_ENDPOINT" -b "$BATCH_SIZE" --force --query-profile "$QUERY_PROFILE"
 fi
 if [[ "${RUN_FALKOR}" == "1" || "${RUN_FALKOR_2}" == "1" ]]; then
   echo "==> Configuring FalkorDB query timeout (${FALKOR_QUERY_TIMEOUT_MS}ms)"
@@ -287,17 +309,17 @@ if [[ "${RUN_FALKOR_2}" == "1" ]]; then
   set_falkor_query_timeout "FalkorDB (secondary)" "$FALKOR_2_HOST" "$FALKOR_2_PORT"
 fi
 
-echo "==> Generating vendor-specific query files (base=${QUERIES_FILE_BASE}, dataset=medium, count=${QUERIES_COUNT}, write_ratio=${WRITE_RATIO})"
+echo "==> Generating vendor-specific query files (base=${QUERIES_FILE_BASE}, dataset=medium, count=${QUERIES_COUNT}, write_ratio=${WRITE_RATIO}, profile=${QUERY_PROFILE})"
 echo "==> Algorithm query toggles (pagerank=${ENABLE_ALGO_PAGERANK_BOOL}, max_flow=${ENABLE_ALGO_MAX_FLOW_BOOL}, msf=${ENABLE_ALGO_MSF_BOOL}, harmonic=${ENABLE_ALGO_HARMONIC_BOOL})"
 # Always regenerate so each vendor gets the latest query catalog + stable q_id fields.
 if [[ "${RUN_FALKOR}" == "1" || "${RUN_FALKOR_2}" == "1" ]]; then
-  cargo run --release --bin benchmark -- generate-queries --vendor falkor   --dataset medium --size "$QUERIES_COUNT" --name "$FALKOR_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" "${ALGO_QUERY_ARGS[@]}"
+  cargo run --release --bin benchmark -- generate-queries --vendor falkor   --dataset medium --size "$QUERIES_COUNT" --name "$FALKOR_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" --query-profile "$QUERY_PROFILE" "${ALGO_QUERY_ARGS[@]}"
 fi
 if [[ "${RUN_NEO4J}" == "1" ]]; then
-  cargo run --release --bin benchmark -- generate-queries --vendor neo4j   --dataset medium --size "$QUERIES_COUNT" --name "$NEO4J_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" "${ALGO_QUERY_ARGS[@]}"
+  cargo run --release --bin benchmark -- generate-queries --vendor neo4j   --dataset medium --size "$QUERIES_COUNT" --name "$NEO4J_QUERIES_FILE"   --write-ratio "$WRITE_RATIO" --query-profile "$QUERY_PROFILE" "${ALGO_QUERY_ARGS[@]}"
 fi
 if [[ "${RUN_MEMGRAPH}" == "1" ]]; then
-  cargo run --release --bin benchmark -- generate-queries --vendor memgraph --dataset medium --size "$QUERIES_COUNT" --name "$MEMGRAPH_QUERIES_FILE" --write-ratio "$WRITE_RATIO" "${ALGO_QUERY_ARGS[@]}"
+  cargo run --release --bin benchmark -- generate-queries --vendor memgraph --dataset medium --size "$QUERIES_COUNT" --name "$MEMGRAPH_QUERIES_FILE" --write-ratio "$WRITE_RATIO" --query-profile "$QUERY_PROFILE" "${ALGO_QUERY_ARGS[@]}"
 fi
 
 echo "==> Running ${QUERIES_FILE} workload (parallel=${PARALLEL}, mps=${MPS})"