lib/api/src/grpc/proto/collections.proto

syntax = "proto3";
package qdrant;
option csharp_namespace = "Qdrant.Client.Grpc";

enum Datatype {
  Default = 0;
  Float32 = 1;
  Uint8 = 2;
  Float16 = 3;
}

message VectorParams {
  uint64 size = 1; // Size of the vectors
  Distance distance = 2; // Distance function used for comparing vectors
  optional HnswConfigDiff hnsw_config = 3; // Configuration of vector HNSW graph. If omitted - the collection configuration will be used
  optional QuantizationConfig quantization_config = 4; // Configuration of vector quantization config. If omitted - the collection configuration will be used
  optional bool on_disk = 5; // If true - serve vectors from disk. If set to false, the vectors will be loaded in RAM.
  optional Datatype datatype = 6; // Data type of the vectors
  optional MultiVectorConfig multivector_config = 7; // Configuration for multi-vector search
}

message VectorParamsDiff {
  optional HnswConfigDiff hnsw_config = 1; // Update params for HNSW index. If empty object - it will be unset
  optional QuantizationConfigDiff quantization_config = 2; // Update quantization params. If none - it is left unchanged.
  optional bool on_disk = 3; // If true - serve vectors from disk. If set to false, the vectors will be loaded in RAM.
}

message VectorParamsMap {
  map<string, VectorParams> map = 1;
}

message VectorParamsDiffMap {
  map<string, VectorParamsDiff> map = 1;
}

message VectorsConfig {
  oneof config {
    VectorParams params = 1;
    VectorParamsMap params_map = 2;
  }
}

message VectorsConfigDiff {
  oneof config {
    VectorParamsDiff params = 1;
    VectorParamsDiffMap params_map = 2;
  }
}

enum Modifier {
    None = 0;
    Idf = 1; // Apply Inverse Document Frequency
}

message SparseVectorParams {
  optional SparseIndexConfig index = 1; // Configuration of sparse index
  optional Modifier modifier = 2; // If set - apply modifier to the vector values
}

message SparseVectorConfig {
  map<string, SparseVectorParams> map = 1;
}

enum MultiVectorComparator {
    MaxSim = 0;
}

message MultiVectorConfig {
    MultiVectorComparator comparator = 1; // Comparator for multi-vector search
}


message GetCollectionInfoRequest {
  string collection_name = 1; // Name of the collection
}

message CollectionExistsRequest {
  string collection_name = 1;
}

message CollectionExists {
  bool exists = 1;
}

message CollectionExistsResponse {
  CollectionExists result = 1;
  double time = 2; // Time spent to process
}

message ListCollectionsRequest {
}

message CollectionDescription {
  string name = 1; // Name of the collection
}

message GetCollectionInfoResponse {
  CollectionInfo result = 1;
  double time = 2; // Time spent to process
}

message ListCollectionsResponse {
  repeated CollectionDescription collections = 1;
  double time = 2; // Time spent to process
}

enum Distance {
  UnknownDistance = 0;
  Cosine = 1;
  Euclid = 2;
  Dot = 3;
  Manhattan = 4;
}

enum CollectionStatus {
  UnknownCollectionStatus = 0;
  Green = 1; // All segments are ready
  Yellow = 2; // Optimization in process
  Red = 3; // Something went wrong
  Grey = 4; // Optimization is pending
}

enum PayloadSchemaType {
  UnknownType = 0;
  Keyword = 1;
  Integer = 2;
  Float = 3;
  Geo = 4;
  Text = 5;
  Bool = 6;
  Datetime = 7;
  Uuid = 8;
}

enum QuantizationType {
  UnknownQuantization = 0;
  Int8 = 1;
}

enum CompressionRatio {
  x4 = 0;
  x8 = 1;
  x16 = 2;
  x32 = 3;
  x64 = 4;
}

message OptimizerStatus {
  bool ok = 1;
  string error = 2;
}

message HnswConfigDiff {
  /*
  Number of edges per node in the index graph. Larger the value - more accurate the search, more space required.
  */
  optional uint64 m = 1;
  /*
  Number of neighbours to consider during the index building. Larger the value - more accurate the search, more time required to build the index.
  */
  optional uint64 ef_construct = 2;
  /*
  Minimal size (in KiloBytes) of vectors for additional payload-based indexing.
  If the payload chunk is smaller than `full_scan_threshold` additional indexing won't be used -
  in this case full-scan search should be preferred by query planner and additional indexing is not required.
  Note: 1 Kb = 1 vector of size 256
  */
  optional uint64 full_scan_threshold = 3;
  /*
  Number of parallel threads used for background index building.
  If 0 - automatically select from 8 to 16.
  Best to keep between 8 and 16 to prevent likelihood of building broken/inefficient HNSW graphs.
  On small CPUs, less threads are used.
  */
  optional uint64 max_indexing_threads = 4;
  /*
  Store HNSW index on disk. If set to false, the index will be stored in RAM.
  */
  optional bool on_disk = 5;
  /*
  Number of additional payload-aware links per node in the index graph. If not set - regular M parameter will be used.
  */
  optional uint64 payload_m = 6;
}

message SparseIndexConfig {
  /*
    Prefer a full scan search upto (excluding) this number of vectors.
    Note: this is number of vectors, not KiloBytes.
   */
  optional uint64 full_scan_threshold = 1;
  /*
  Store inverted index on disk. If set to false, the index will be stored in RAM.
   */
  optional bool on_disk = 2;
  /*
  Datatype used to store weights in the index.
  */
  optional Datatype datatype = 3;
}

message WalConfigDiff {
  optional uint64 wal_capacity_mb = 1; // Size of a single WAL block file
  optional uint64 wal_segments_ahead = 2; // Number of segments to create in advance
}

message OptimizersConfigDiff {
  /*
  The minimal fraction of deleted vectors in a segment, required to perform segment optimization
  */
  optional double deleted_threshold = 1;
  /*
  The minimal number of vectors in a segment, required to perform segment optimization
  */
  optional uint64 vacuum_min_vector_number = 2;
  /*
  Target amount of segments the optimizer will try to keep.
  Real amount of segments may vary depending on multiple parameters:

   - Amount of stored points.
   - Current write RPS.

  It is recommended to select the default number of segments as a factor of the number of search threads,
  so that each segment would be handled evenly by one of the threads.
  */
  optional uint64 default_segment_number = 3;
  /*
  Do not create segments larger this size (in kilobytes).
  Large segments might require disproportionately long indexation times,
  therefore it makes sense to limit the size of segments.

  If indexing speed is more important - make this parameter lower.
  If search speed is more important - make this parameter higher.
  Note: 1Kb = 1 vector of size 256
  If not set, will be automatically selected considering the number of available CPUs.
  */
  optional uint64 max_segment_size = 4;
  /*
  Maximum size (in kilobytes) of vectors to store in-memory per segment.
  Segments larger than this threshold will be stored as read-only memmaped file.

  Memmap storage is disabled by default, to enable it, set this threshold to a reasonable value.

  To disable memmap storage, set this to `0`.

  Note: 1Kb = 1 vector of size 256
  */
  optional uint64 memmap_threshold = 5;
  /*
  Maximum size (in kilobytes) of vectors allowed for plain index, exceeding this threshold will enable vector indexing

  Default value is 20,000, based on <https://github.com/google-research/google-research/blob/master/scann/docs/algorithms.md>.

  To disable vector indexing, set to `0`.

  Note: 1kB = 1 vector of size 256.
  */
  optional uint64 indexing_threshold = 6;
  /*
  Interval between forced flushes.
  */
  optional uint64 flush_interval_sec = 7;
  /*
  Max number of threads (jobs) for running optimizations per shard.
  Note: each optimization job will also use `max_indexing_threads` threads by itself for index building.
  If null - have no limit and choose dynamically to saturate CPU.
  If 0 - no optimization threads, optimizations will be disabled.
  */
  optional uint64 max_optimization_threads = 8;
}

message ScalarQuantization {
  QuantizationType type = 1; // Type of quantization
  optional float quantile = 2; // Number of bits to use for quantization
  optional bool always_ram = 3; // If true - quantized vectors always will be stored in RAM, ignoring the config of main storage
}

message ProductQuantization {
  CompressionRatio compression = 1; // Compression ratio
  optional bool always_ram = 2; // If true - quantized vectors always will be stored in RAM, ignoring the config of main storage
}

message BinaryQuantization {
  optional bool always_ram = 1; // If true - quantized vectors always will be stored in RAM, ignoring the config of main storage
}

message QuantizationConfig {
  oneof quantization {
    ScalarQuantization scalar = 1;
    ProductQuantization product = 2;
    BinaryQuantization binary = 3;
  }
}

message Disabled {

}

message QuantizationConfigDiff {
  oneof quantization {
    ScalarQuantization scalar = 1;
    ProductQuantization product = 2;
    Disabled disabled = 3;
    BinaryQuantization binary = 4;
  }
}

enum ShardingMethod {
  Auto = 0; // Auto-sharding based on record ids
  Custom = 1; // Shard by user-defined key
}

message StrictModeConfig {
  optional bool enabled = 1;
  optional uint32 max_query_limit = 2;
  optional uint32 max_timeout = 3;
  optional bool unindexed_filtering_retrieve = 4;
  optional bool unindexed_filtering_update = 5;

  optional uint32 search_max_hnsw_ef = 6;
  optional bool search_allow_exact  = 7;
  optional float search_max_oversampling  = 8;
}

message CreateCollection {
  string collection_name = 1; // Name of the collection
  reserved 2; // Deprecated
  reserved 3; // Deprecated
  optional HnswConfigDiff hnsw_config = 4; // Configuration of vector index
  optional WalConfigDiff wal_config = 5; // Configuration of the Write-Ahead-Log
  optional OptimizersConfigDiff optimizers_config = 6; // Configuration of the optimizers
  optional uint32 shard_number = 7; // Number of shards in the collection, default is 1 for standalone, otherwise equal to the number of nodes. Minimum is 1
  optional bool on_disk_payload = 8; // If true - point's payload will not be stored in memory
  optional uint64 timeout = 9; // Wait timeout for operation commit in seconds, if not specified - default value will be supplied
  optional VectorsConfig vectors_config = 10; // Configuration for vectors
  optional uint32 replication_factor = 11; // Number of replicas of each shard that network tries to maintain, default = 1
  optional uint32 write_consistency_factor = 12; // How many replicas should apply the operation for us to consider it successful, default = 1
  optional string init_from_collection = 13; // Specify name of the other collection to copy data from
  optional QuantizationConfig quantization_config = 14; // Quantization configuration of vector
  optional ShardingMethod sharding_method = 15; // Sharding method
  optional SparseVectorConfig sparse_vectors_config = 16; // Configuration for sparse vectors
  optional StrictModeConfig strict_mode_config = 17; // Configuration for strict mode
}

message UpdateCollection {
  string collection_name = 1; // Name of the collection
  optional OptimizersConfigDiff optimizers_config = 2; // New configuration parameters for the collection. This operation is blocking, it will only proceed once all current optimizations are complete
  optional uint64 timeout = 3; // Wait timeout for operation commit in seconds if blocking, if not specified - default value will be supplied
  optional CollectionParamsDiff params = 4; // New configuration parameters for the collection
  optional HnswConfigDiff hnsw_config = 5; // New HNSW parameters for the collection index
  optional VectorsConfigDiff vectors_config = 6; // New vector parameters
  optional QuantizationConfigDiff quantization_config = 7; // Quantization configuration of vector
  optional SparseVectorConfig sparse_vectors_config = 8; // New sparse vector parameters
  optional StrictModeConfig strict_mode_config = 9; // New strict mode configuration
}

message DeleteCollection {
  string collection_name = 1; // Name of the collection
  optional uint64 timeout = 2; // Wait timeout for operation commit in seconds, if not specified - default value will be supplied
}

message CollectionOperationResponse {
  bool result = 1; // if operation made changes
  double time = 2; // Time spent to process
}

message CollectionParams {
  reserved 1; // Deprecated
  reserved 2; // Deprecated
  uint32 shard_number = 3; // Number of shards in collection
  bool on_disk_payload = 4; // If true - point's payload will not be stored in memory
  optional VectorsConfig vectors_config = 5; // Configuration for vectors
  optional uint32 replication_factor = 6; // Number of replicas of each shard that network tries to maintain
  optional uint32 write_consistency_factor = 7; // How many replicas should apply the operation for us to consider it successful
  optional uint32 read_fan_out_factor = 8; // Fan-out every read request to these many additional remote nodes (and return first available response)
  optional ShardingMethod sharding_method = 9; // Sharding method
  optional SparseVectorConfig sparse_vectors_config = 10; // Configuration for sparse vectors
}

message CollectionParamsDiff {
  optional uint32 replication_factor = 1; // Number of replicas of each shard that network tries to maintain
  optional uint32 write_consistency_factor = 2; // How many replicas should apply the operation for us to consider it successful
  optional bool on_disk_payload = 3; // If true - point's payload will not be stored in memory
  optional uint32 read_fan_out_factor = 4; // Fan-out every read request to these many additional remote nodes (and return first available response)
}

message CollectionConfig {
  CollectionParams params = 1; // Collection parameters
  HnswConfigDiff hnsw_config = 2; // Configuration of vector index
  OptimizersConfigDiff optimizer_config = 3; // Configuration of the optimizers
  WalConfigDiff wal_config = 4; // Configuration of the Write-Ahead-Log
  optional QuantizationConfig quantization_config = 5; // Configuration of the vector quantization
  optional StrictModeConfig strict_mode_config = 6; // Configuration of strict mode.
}

enum TokenizerType {
  Unknown = 0;
  Prefix = 1;
  Whitespace = 2;
  Word = 3;
  Multilingual = 4;
}

message KeywordIndexParams {
    optional bool is_tenant = 1; // If true - used for tenant optimization.
    optional bool on_disk = 2; // If true - store index on disk.
}

message IntegerIndexParams {
  optional bool lookup = 1; // If true - support direct lookups.
  optional bool range = 2; // If true - support ranges filters.
  optional bool is_principal = 3; // If true - use this key to organize storage of the collection data. This option assumes that this key will be used in majority of filtered requests.
  optional bool on_disk = 4; // If true - store index on disk.
}

message FloatIndexParams {
  optional bool on_disk = 1; // If true - store index on disk.
  optional bool is_principal = 2; // If true - use this key to organize storage of the collection data. This option assumes that this key will be used in majority of filtered requests.
}

message GeoIndexParams {
  optional bool on_disk = 1; // If true - store index on disk.
}

message TextIndexParams {
  TokenizerType tokenizer = 1; // Tokenizer type
  optional bool lowercase = 2; // If true - all tokens will be lowercase
  optional uint64 min_token_len = 3; // Minimal token length
  optional uint64 max_token_len = 4; // Maximal token length
  optional bool on_disk = 5; // If true - store index on disk.
}

message BoolIndexParams {
}

message DatetimeIndexParams {
  optional bool on_disk = 1; // If true - store index on disk.
  optional bool is_principal = 2; // If true - use this key to organize storage of the collection data. This option assumes that this key will be used in majority of filtered requests.
}

message UuidIndexParams {
  optional bool is_tenant = 1; // If true - used for tenant optimization.
  optional bool on_disk = 2; // If true - store index on disk.
}

message PayloadIndexParams {
  oneof index_params {
    KeywordIndexParams keyword_index_params = 3; // Parameters for keyword index
    IntegerIndexParams integer_index_params = 2; // Parameters for integer index
    FloatIndexParams float_index_params = 4; // Parameters for float index
    GeoIndexParams geo_index_params = 5; // Parameters for geo index
    TextIndexParams text_index_params = 1; // Parameters for text index
    BoolIndexParams bool_index_params = 6; // Parameters for bool index
    DatetimeIndexParams datetime_index_params = 7; // Parameters for datetime index
    UuidIndexParams uuid_index_params = 8; // Parameters for uuid index
  }
}

message PayloadSchemaInfo {
  PayloadSchemaType data_type = 1; // Field data type
  optional PayloadIndexParams params = 2; // Field index parameters
  optional uint64 points = 3; // Number of points indexed within this field indexed
}

message CollectionInfo {
  CollectionStatus status = 1; // operating condition of the collection
  OptimizerStatus optimizer_status = 2; // status of collection optimizers
  optional uint64 vectors_count = 3; // Approximate number of vectors in the collection
  uint64 segments_count = 4; // Number of independent segments
  reserved 5; // Deprecated
  reserved 6; // Deprecated
  CollectionConfig config = 7; // Configuration
  map<string, PayloadSchemaInfo> payload_schema = 8; // Collection data types
  optional uint64 points_count = 9; // Approximate number of points in the collection
  optional uint64 indexed_vectors_count = 10; // Approximate number of indexed vectors in the collection.
}

message ChangeAliases {
  repeated AliasOperations actions = 1; // List of actions
  optional uint64 timeout = 2; // Wait timeout for operation commit in seconds, if not specified - default value will be supplied
}

message AliasOperations {
  oneof action {
    CreateAlias create_alias = 1;
    RenameAlias rename_alias = 2;
    DeleteAlias delete_alias = 3;
  }
}

message CreateAlias {
  string collection_name = 1; // Name of the collection
  string alias_name = 2; // New name of the alias
}

message RenameAlias {
  string old_alias_name = 1; // Name of the alias to rename
  string new_alias_name = 2; // Name of the alias
}

message DeleteAlias {
  string alias_name = 1; // Name of the alias
}

message ListAliasesRequest {
}

message ListCollectionAliasesRequest {
  string collection_name = 1; // Name of the collection
}

message AliasDescription {
  string alias_name = 1; // Name of the alias
  string collection_name = 2; // Name of the collection
}

message ListAliasesResponse {
  repeated AliasDescription aliases = 1;
  double time = 2; // Time spent to process
}

message CollectionClusterInfoRequest {
  string collection_name = 1; // Name of the collection
}

enum ReplicaState {
  Active = 0; // Active and sound
  Dead = 1; // Failed for some reason
  Partial = 2; // The shard is partially loaded and is currently receiving data from other shards
  Initializing = 3; // Collection is being created
  Listener = 4; // A shard which receives data, but is not used for search; Useful for backup shards
  PartialSnapshot = 5; // Deprecated: snapshot shard transfer is in progress; Updates should not be sent to (and are ignored by) the shard
  Recovery = 6; // Shard is undergoing recovered by an external node; Normally rejects updates, accepts updates if force is true
  Resharding = 7; // Points are being migrated to this shard as part of resharding
}

message ShardKey {
  oneof key {
    string keyword = 1; // String key
    uint64 number = 2; // Number key
  }
}

message LocalShardInfo {
  uint32 shard_id = 1; // Local shard id
  uint64 points_count = 2; // Number of points in the shard
  ReplicaState state = 3;  // Is replica active
  optional ShardKey shard_key = 4; // User-defined shard key
}

message RemoteShardInfo {
  uint32 shard_id = 1; // Local shard id
  uint64 peer_id = 2; // Remote peer id
  ReplicaState state = 3; // Is replica active
  optional ShardKey shard_key = 4; // User-defined shard key
}

message ShardTransferInfo {
  uint32 shard_id = 1; // Local shard id
  optional uint32 to_shard_id = 5;
  uint64 from = 2;
  uint64 to = 3;
  bool sync = 4; // If `true` transfer is a synchronization of a replicas; If `false` transfer is a moving of a shard from one peer to another
}

message ReshardingInfo {
  uint32 shard_id = 1;
  uint64 peer_id = 2;
  optional ShardKey shard_key = 3;
}

message CollectionClusterInfoResponse {
  uint64 peer_id = 1;  // ID of this peer
  uint64 shard_count = 2; // Total number of shards
  repeated LocalShardInfo local_shards = 3; // Local shards
  repeated RemoteShardInfo remote_shards = 4; // Remote shards
  repeated ShardTransferInfo shard_transfers = 5; // Shard transfers
  // TODO(resharding): enable on release:
  // repeated ReshardingInfo resharding_operations = 6; // Resharding operations
}

message MoveShard {
  uint32 shard_id = 1; // Local shard id
  optional uint32 to_shard_id = 5;
  uint64 from_peer_id = 2;
  uint64 to_peer_id = 3;
  optional ShardTransferMethod method = 4;
}

message ReplicateShard {
  uint32 shard_id = 1; // Local shard id
  optional uint32 to_shard_id = 5;
  uint64 from_peer_id = 2;
  uint64 to_peer_id = 3;
  optional ShardTransferMethod method = 4;
}

message AbortShardTransfer {
  uint32 shard_id = 1; // Local shard id
  optional uint32 to_shard_id = 4;
  uint64 from_peer_id = 2;
  uint64 to_peer_id = 3;
}

message RestartTransfer {
  uint32 shard_id = 1; // Local shard id
  optional uint32 to_shard_id = 5;
  uint64 from_peer_id = 2;
  uint64 to_peer_id = 3;
  ShardTransferMethod method = 4;
}

enum ShardTransferMethod {
  StreamRecords = 0; // Stream shard records in batches
  Snapshot = 1; // Snapshot the shard and recover it on the target peer
  WalDelta = 2; // Resolve WAL delta between peers and transfer the difference
  ReshardingStreamRecords = 3; // Stream shard records in batches for resharding
}

message Replica {
  uint32 shard_id = 1;
  uint64 peer_id = 2;
}

message CreateShardKey {
    ShardKey shard_key = 1; // User-defined shard key
    optional uint32 shards_number = 2; // Number of shards to create per shard key
    optional uint32 replication_factor = 3; // Number of replicas of each shard to create
    repeated uint64 placement = 4; // List of peer ids, allowed to create shards. If empty - all peers are allowed
}

message DeleteShardKey {
    ShardKey shard_key = 1; // Shard key to delete
}

message UpdateCollectionClusterSetupRequest {
  string collection_name = 1; // Name of the collection
  oneof operation {
    MoveShard move_shard = 2;
    ReplicateShard replicate_shard = 3;
    AbortShardTransfer abort_transfer = 4;
    Replica drop_replica = 5;
    CreateShardKey create_shard_key = 7;
    DeleteShardKey delete_shard_key = 8;
    RestartTransfer restart_transfer = 9;
  }
  optional uint64 timeout = 6; // Wait timeout for operation commit in seconds, if not specified - default value will be supplied
}

message UpdateCollectionClusterSetupResponse {
  bool result = 1;
}

message CreateShardKeyRequest {
    string collection_name = 1; // Name of the collection
    CreateShardKey request = 2; // Request to create shard key
    optional uint64 timeout = 3; // Wait timeout for operation commit in seconds, if not specified - default value will be supplied
}

message DeleteShardKeyRequest {
    string collection_name = 1; // Name of the collection
    DeleteShardKey request = 2; // Request to delete shard key
    optional uint64 timeout = 3; // Wait timeout for operation commit in seconds, if not specified - default value will be supplied
}

message CreateShardKeyResponse {
    bool result = 1;
}

message DeleteShardKeyResponse {
    bool result = 1;
}