Get desktop application:
View/edit binary Protocol Buffers messages
Client Telemetry Service
Client Heartbeat
Delete Client Command
Get Client Telemetry
Push Client Command
The serialized `schema.FieldSchema`
Row Level Security (RLS) APIs
User Tag Management
* Alter collection in milvus
Not useful for now
The unique collection name in milvus.(Required)
Not useful for now
The unique collection name in milvus.(Required)
Alter index
Not useful for now
role name
role description
Not useful for now
The collection name you want to describe, you can pass collection_name or collectionID
The collection ID you want to describe
Contain error_code and reason
work as hint for now
vectors on the left of operator
vectors on the right of operator
"metric":"L2"/"IP"/"HAMMIN"/"TANIMOTO"
num(op_left)*num(op_right) distance values, "HAMMIN" return integer distance
(message has no fields)
* Create collection in milvus
Not useful for now
The unique collection name in milvus.(Required)
The serialized `schema.CollectionSchema`(Required)
Once set, no modification is allowed (Optional) https://github.com/milvus-io/milvus/issues/6690
The consistency level that the collection used, modification is not supported now.
num of default physical partitions, only used in partition key mode and changes are not supported
https://wiki.lfaidata.foundation/display/MIL/MEP+27+--+Support+Basic+Authentication
Not useful for now
username
ciphertext password
create time
modify time
Create index for vector datas
Not useful for now
The particular collection name you want to create index.
The vector field name in this particular collection
Support keys: index_type,metric_type, params. Different index_type may has different params.
Version before 2.0.2 doesn't contain index_name, we use default index name.
Create partition in created collection.
Not useful for now
The collection name in milvus
The partition name you want to create.
Not useful for now
group name
CreateReplicateStream establishes a replication stream on the target Milvus cluster. Semantics: - Sets up a continuous data stream that receives replicated messages (DDL, insert, delete, etc.) from the source cluster via CDC. - Once established, the target cluster persists incoming messages into its WAL (Write-Ahead Log) ensuring durability and consistency.
https://wiki.lfaidata.foundation/display/MIL/MEP+29+--+Support+Role-Based+Access+Control
Not useful for now
role
Row Policy Management
actions this policy controls
role names that this policy applies to
expression for filtering data during queries
expression for validating data during mutations
optional description
snapshot related
Snapshot Management
user-defined snapshot name
user-defined snapshot description
db name
collection name
duration in seconds to protect referenced segments from compaction, 0 = no protection
Not useful for now
Not useful for now
Describe alias response
Response status
* Get collection meta datas like: schema, collectionID, shards number ...
Not useful for now
The collection name you want to describe, you can pass collection_name or collectionID
The collection ID you want to describe
If time_stamp is not zero, will describe collection success when time_stamp >= created collection timestamp, otherwise will throw error.
Get created index information. Current release of Milvus only supports showing latest built index.
Not useful for now
The particular collection name in Milvus
The vector field name in this particular collection
No need to set up for now @2021.06.30
Describe index response
Response status
All index informations, for now only return tha latest index you created for the collection.
snapshot name
db name
collection name
user-defined snapshot name
user-defined snapshot description
collection name
the partition names in the snapshot.
create timestamp, all stream data before this timestamp will be flushed to segment list.
s3 location of the snapshot meta file if snapshot is exported to s3.
* Drop collection in milvus, also will drop data in collection.
Not useful for now
The unique collection name in milvus.(Required)
must
must
Deprecated: not be used in the milvus
Drop partition in created collection.
Not useful for now
The collection name in milvus
The partition name you want to drop
Not useful for now
group name
Not useful for now
role name
force to drop the role even if there is permission binding
snapshot name
db name
collection name
DumpMessages streams messages from a WAL range for data salvage. Semantics: - Reads messages starting from start_message_id position (inclusive). - Use start_timetick/end_timetick to filter messages by timetick range. - Only returns non-system messages (filters out TimeTick, CreateSegment, Flush, RollbackTxn). - Streams messages until end_timetick is reached or context is cancelled. - Typically used after force failover to recover unsynchronized messages from the old primary cluster using the salvage checkpoint.
DumpMessagesRequest is used to dump messages from a WAL range for data salvage. Usage: After force failover, use GetReplicateInfo to get the salvage_checkpoint, then call DumpMessages with the checkpoint's message_id and timetick to retrieve unsynchronized messages from the old primary cluster.
Physical channel name to dump from (required)
Start position in WAL (required) - typically from salvage_checkpoint.message_id
Start timetick filter (optional) - only dump messages with timetick >= start_timetick, typically from salvage_checkpoint.time_tick
End timetick filter (optional) - only dump messages with timetick <= end_timetick, 0 means no limit (stream until cancelled)
If true, dump starts from start_message_id inclusively; otherwise starts after it.
DumpMessagesResponse streams messages from the WAL. Each response contains either an error status or a message (never both).
Error status (only sent when an error occurs)
A dumped message (filtered to exclude system messages)
export a snapshot and all referenced files to a self-contained external bundle
source snapshot name
source db name (where the snapshot lives)
source collection name (where the snapshot lives)
target root path for the exported snapshot bundle
Optional external storage spec JSON for the external snapshot storage. Snapshot APIs consume the extfs section for provider, endpoint, region, SSL, IAM, and credential settings.
physical time for backup tool
hybrid ts for geting flush tate
pchannel -> FlushAllMsg
* Get collection statistics like row_count.
Not useful for now
The collection name you want get statistics
* Will return collection statistics in stats field like [{key:"row_count",value:"1"}]
Contain error_code and reason
Collection statistics data
(message has no fields)
pchannel -> flush all ts
True if all data in the cluster are flushed
Detailed flush state for each database Provides per-database and per-collection flush status information
id of an import task
Deprecated: use DescribeIndex instead
Get index building progress
Not useful for now
The collection name in milvus
The vector field name in this collection
Not useful for now
Deprecated: use DescribeIndex instead
must
must
No need to set up for now @2021.06.30
Not useful for now
Not useful for now
The particular collection name in Milvus
The index name in this particular collection
Response status
All index information.
Not useful for now
Not useful for now
https://wiki.lfaidata.foundation/display/MIL/MEP+8+--+Add+metrics+for+proxy
request is of jsonic format
response is of jsonic format
metrics from which component
Get partition statistics like row_count.
Not useful for now
The collection name in milvus
The partition name you want to collect statistics
must
must
must
must
Job ID from RefreshExternalCollection response
GetReplicateConfiguration retrieves the current cross-cluster replication topology. Sensitive fields (like connection tokens) are redacted in the response.
(message has no fields)
GetReplicateInfo retrieves replication-related metadata of specified channel from a target Milvus cluster.
salvage_checkpoint is captured during force promote. It represents the last synced position from the old primary. Data salvage tools can use this to dump unsynchronized messages from the old primary.
restore job ID
(message has no fields)
* Check collection exist in milvus or not.
Not useful for now
The collection name you want to check.
If time_stamp is not zero, will return true when time_stamp >= created collection timestamp, otherwise will return false.
Check if partition exist in collection or not.
Not useful for now
The collection name in milvus
The partition name you want to check
must
must
must
must
guarantee_timestamp
Mutually exclusive with function_chains. Milvus returns invalid argument if both are set.
Function chains for top-level hybrid fusion/rerank. Mutually exclusive with function_score. Milvus returns invalid argument if both are set.
https://wiki.lfaidata.foundation/display/MIL/MEP+24+--+Support+bulk+load
target collection
target partition
channel names for the collection
the file is row-based or column-based
file paths to be imported
import options, bucket, etc.
target database
serialized `schema.ClusteringInfo`
id array of import tasks
List aliases response
Response status
Not useful for now
Contain error_code and reason
username array
target collection, list all tasks if the name is empty
maximum number of tasks returned, list all tasks if the value is 0
list of all import tasks
Not useful for now
Database name
Collection name (optional, if empty lists all)
db name
collection name, if collection_name is empty, it will list all snapshots.
return all snapshots for the given collection Note: list snapshots is not a privilege check operation
Do load balancing operation from src_nodeID to dst_nodeID.
* Load collection data into query nodes, then you can do vector search on this collection.
Not useful for now
The collection name you want to load
The replica number to load, default by 1
create replica used resource group
Whether to enable refresh mode.
Field Partial Load fields list
Field Partial load skip load dynamic fields
Additional parameters for load
Load specific partitions data of one collection into query nodes Then you can get these data as result when you do vector search on this collection.
Not useful for now
The collection name in milvus
The partition names you want to load
The replicas number you would load, 1 by default
create replica used resource group
Whether to enable refresh mode.
Field Partial Load fields list
Field Partial load skip load dynamic fields
Additional parameters for load
Not useful for now
grant
operation type
version
Not useful for now
group name
privileges
operation type
Not useful for now
role
privilege
operation type
db name
collection name
Not useful for now
username
role name
operation type
snapshot name
database name
collection name
optional: 0 = no expiry, >0 = auto-expire in seconds
allocated pin ID for Unpin
guarantee_timestamp
optional
for session-like operation like iterator
External Collection APIs
Database name
Collection name (required)
Optional: new external source path
Optional: new external spec configuration
Unique job identifier for tracking
TODO: remove
* Release collection data from query nodes, then you can't do vector search on this collection.
Not useful for now
The collection name you want to release
Release specific partitions data of one collection from query nodes. Then you can not get these data as result when you do vector search on this collection.
Not useful for now
The collection name in milvus
The partition names you want to release
Deprecated CDC API
restore an external snapshot by metadata URI to a new collection
target db name (where to restore to)
target collection name (must not exist)
external snapshot metadata URI
Optional external storage spec JSON for the external snapshot storage. Snapshot APIs consume the extfs section for provider, endpoint, region, SSL, IAM, and credential settings.
Not useful for now
restore a snapshot to a new collection
snapshot name
source db name (where the snapshot lives)
source collection name (where the snapshot lives)
if true, use import to rewrite data to collection
target db name (where to restore to)
target collection name (must not exist)
Not useful for now
grant
Not useful for now
grant info array
Not useful for now
role
include user info
Not useful for now
role result array
Not useful for now
user
include user info
Not useful for now
user result array
List collections
Not useful for now
Not useful for now
Decide return Loaded collections or All collections(Optional)
When type is InMemory, will return these collection's inMemory_percentages.(Optional) Deprecated: use GetLoadingProgress rpc instead
Return basic collection infos.
Contain error_code and reason
Collection name array
Collection Id array
Hybrid timestamps in milvus
The utc timestamp calculated by created_timestamp
Load percentage on querynode when type is InMemory Deprecated: use GetLoadingProgress rpc instead
Indicate whether query service is available
List all partitions for particular collection
Not useful for now
The collection name you want to describe, you can pass collection_name or collectionID
The collection id in milvus
When type is InMemory, will return these patitions's inMemory_percentages.(Optional)
Decide return Loaded partitions or All partitions(Optional) Deprecated: use GetLoadingProgress rpc instead
List all partitions for particular collection response. The returned datas are all rows, we can format to columns by therir index.
Contain error_code and reason
All partition names for this collection
All partition ids for this collection
All hybrid timestamps
All utc timestamps calculated by created_timestamps
Load percentage on querynode Deprecated: use GetLoadingProgress rpc instead
transfer `nodeNum` nodes from `source_resource_group` to `target_resource_group`
transfer `replicaNum` replicas in `collectionID` from `source_resource_group` to `target_resource_group`
Truncate collection in milvus
Not useful for now
The unique collection name in milvus.(Required)
pin ID from PinSnapshotData response
Not useful for now
username
old password
new password
create time
modify time
CDC v2 APIs UpdateReplicateConfiguration applies a full replacement of the current replication configuration across Milvus clusters. Semantics: - The provided ReplicateConfiguration completely replaces any existing configuration persisted in the metadata store. - Passing an empty ReplicateConfiguration is treated as a "clear" operation, effectively removing all replication configuration. - The RPC is expected to be idempotent: submitting the same configuration multiple times must not cause side effects.
If true, the current cluster will be promoted to primary forcefully. This is intended for failover scenarios.
field_ops carries per-field FieldPartialUpdateOp directives describing how each FieldData should be merged during a partial upsert. Ops are matched to fields_data entries by field_name. A non-REPLACE op implies partial_update=true regardless of the partial_update flag.
Action to perform on the collection schema
Used in:
Add fields and functions request
Used in:
Drop field or function request
Used in:
The serialized `schema.FieldSchema`
Used in:
Version before 2.0.2 doesn't contain index_name, we use default index name.
Support keys: index_type,metric_type, params. Different index_type may has different params.
Used in:
Used in:
Used in:
Used as response type in: MilvusService.HasCollection, MilvusService.HasPartition
Used in:
Used in:
Used in:
Used in:
* DescribeCollection Response
Used as response type in: MilvusService.DescribeCollection
Used as field type in:
Contain error_code and reason
The schema param when you created collection.
The collection id
System design related, users should not perceive
System design related, users should not perceive
Hybrid timestamp in milvus
The utc timestamp calculated by created_timestamp
The shards number you set.
The aliases of this collection
The message ID/posititon when collection is created
The consistency level that the collection used, modification is not supported now.
The collection name
Used in:
Used in:
Deprecated Flush result for a single flush target
Used in:
Database name containing the collections
List of collections that were actually processed for this target If the target had empty collection_names (flush all), this shows all collections in the database If the target specified collection_names, this shows only those collections
Deprecated
Used in:
Database name for this flush state report
Collection-level flush state mapping Key: collection name, Value: true if collection is fully flushed
Deprecated, FlushAll semantics changed to flushing the entire cluster. Specific collection to flush with database context This message allows targeting specific collections within a database for flush operations
Used in: ,
Database name to target for flush operation
Collection names within this database to flush If empty, flush all collections in this database
Used in:
Collection name
Segment IDs
Segment IDs that were actually flushed
Seal time (physical time for backup tools)
Flush timestamp (hybrid timestamp for getting flush state)
channel name to checkpoint position
Used as response type in: MilvusService.GetImportState
Used as field type in:
is this import task finished or not
if the task is finished, this value is how many rows are imported. if the task is not finished, this value is how many rows are parsed. return 0 if failed.
auto generated ids if the primary key is autoid
more information about the task, progress percent, file path, failed reason, etc.
id of an import task
collection ID of the import task.
a list of segment IDs created by the import task.
timestamp when the import task is created.
* Get statistics like row_count. WARNING: This API is experimental and not useful for now.
Not useful for now
The collection name you want get statistics
The partition names you want get statistics, empty for all partitions
Not useful for now, reserved for future
* Will return statistics in stats field like [{key:"row_count",value:"1"}] WARNING: This API is experimental and not useful for now.
Contain error_code and reason
Collection statistics data
Used in: , , ,
role
object
object name
privilege
db name
Used in: , ,
Index informations
Used in: ,
Index name
Index id
Will return index_type, metric_type, params(like nlist).
The vector field name
index build progress
index state
minimal index version for current index
maximum index version for current index
Used as response type in: MilvusService.Delete, MilvusService.Insert, MilvusService.Upsert
required for insert, delete, upsert
error indexes indicate
error indexes indicate
Used in:
Used in:
Used in: ,
Used in:
Used in:
Used in: , ,
Used in: ,
Used in:
deprecated, check node_ids(NodeIds) field
Used in:
Used in: ,
user
role
(role, object, previledge)
privilege group info
Used in: ,
Job identifier
Collection name
Current job state
Progress percentage (0-100)
Error message if failed
External source used for this job
Job start timestamp (Unix epoch ms)
Job end timestamp (0 if not completed)
External spec used for this job
External Collection Refresh APIs
Used in:
Job is queued, waiting to execute
Job is currently executing
Job completed successfully
Job failed with error
Used as request type in: MilvusService.RegisterLink, ProxyService.RegisterLink
(message has no fields)
Used as response type in: MilvusService.RegisterLink, ProxyService.RegisterLink
ReplicaGroup
Used in:
empty indicates to load collection
include leaders
outbound access rg -> node num
Used in:
the source time tick of the last confirmed message that persisted into target WAL.
Used in:
the cluster id of source cluster
replicate message to be sent
Used in:
collection name -> loaded replica num
collection name -> accessed other rg's node num
collection name -> be accessed node num by other rg
resource group configuration.
query node belong to this resource group now.
Used in: ,
progress percentage (0-100)
start timestamp in milliseconds
time cost in milliseconds
Used in:
Job created, waiting to start
Executing restore snapshot job
All done
Failed
https://wiki.lfaidata.foundation/display/MIL/MEP+29+--+Support+Role-Based+Access+Control
Used in: , , , , , , ,
Used in:
Used in:
Row Policy Action enum
Used in: ,
Used as request type in: MilvusService.Search
Used as field type in:
must
must
must
must
serialized `PlaceholderGroup` for vector search
primary keys for search by IDs
must
must
guarantee_timestamp
use search_input instead
Mutually exclusive with function_chains. Milvus returns invalid argument if both are set.
Mutually exclusive with function_score. Milvus returns invalid argument if both are set.
Used as response type in: MilvusService.HybridSearch, MilvusService.Search
for session-like operation like iterator
Used in:
IP:port
optional, DO NOT save it in meta, set it only for GetReplicas() if with_shard_nodes is true
Deprecated: use GetLoadingProgress rpc instead
Used in: ,
Will return all collections
Will return loaded collections with their inMemory_percentages
Used in:
must
serialized `PlaceholderGroup`
must
must
must
Used in: , , ,
Used in:
Used in:
Used in:
Used in:
vector ids
vectors data