Elasticsearch Python Client release notes

Review the changes, fixes, and more in each version of Elasticsearch Python Client.

To check for security updates, go to Security announcements for the Elastic stack.

9.4.1 (2026-05-26)

Enhancements

• Optional backoff delay strategy for retries (Fixes #197) ( #3430)
• Sanitize index names in ES|QL query builder (Fixes #3422) ( #3425)

DSL

• Added flat_index_threshold attribute to DenseVectorIndexOptions type
• Added field attribute to InnerHits type
• Added embedding attribute to QueryVectorBuilder type

9.4.0 (2026-05-06)

Enhancements

• ES|QL updates for 9.4 and Serverless ( #3384)
• Added server_mode parameter to the client’s constructor

API

• Added inference.embedding API
• Added inference.put_fireworksai API
• Added project.create_many_routing, project.create_routing, project.delete_routing, project.get_many_routing and project.get_routing APIs
• Added security.clone_api_key API
• Added group_by parameter to reindex_rethrottle API
• Added return_intermediate_results parameter to async_search.get API
• Added project_routing and time_zone parameters to esql.async_query and esql.query APIs
• Removed master_timeout parameter from searchable_snapshots.cache_stats API
• Added name parameter to streams.logs_enable and streams.logs_disable APIs
• Added should_parse_recursively parameter to text_structure.find_field_structure, text_structure.find_message_structure and text_structure.find_structure APIs

DSL

• Data stream support ( #3413)
• Handle document attributes that have a name collision with a method (#3364)
• Added es_name parameter to mapped_field function
• Added _es_name parameter to all subclasses of Field
• Added script attribute to MultiTermLookup type
• Added lookup attribute to QueryVectorBuilder type

9.3.0 (2026-02-03)

Enhancements

• Add pack_dense_vector helper function to pack dense vectors for efficient uploading ( #3219)
• New and updated ES|QL functions in the ES|QL query builder for 9.3 and Serverless ( #3266)

API

• Added cat.circuit_breaker API
• Added inference.put_groq, inference.put_openshift_ai, inference.put_nvidia APIs
• Added downsampling_method argument to indices.put_data_lifecycle API
• Added return_documents and top_n arguments to inference.rerank API
• Added close_job argument to ml.stop_datafeed API
• Added certificate_identity to security.create_cross_cluster_api_key and security.update_cross_cluster_api_key APIs

Serverless-specific

• Added project_routing argument to project.tags API

DSL

• Added NumpyDenseVector field, with support for dense vectors based on numpy arrays ( #3218)
• Added ExponentialHistogram field
• Added time_series_metric argument to Histogram field
• Added on_disk_rescore argument to DenseVectorIndexOptions type
• Added slices argument to UpdateByQueryResponse type

9.2.1 (2025-12-23)

Enhancements

• Instrument ping with OTel ( #3160)
• Make positional arguments in DSL generated classes explicit ( #3233)
• Add warnings for private APIs ( #3212)

Bug fixes

• ES|QL query builder: fix missing assignment ( #3151)
• Use relative imports to fix elasticsearch9 package imports ( #3232)

API

• Added transform.get_node_stats API
• Added requests_per_second argument to delete_by_query_rethrottle, reindex_rethrottle and update_by_query_rethrottle APIs
• Added allow_closed, allow_no_indices, expand_wildcards, ignore_throttled and ignore_available arguments to cat.segments API
• Added chunking_settings argument to inference.put_watsonxai API
• Added id argument to ml.stop_trained_model_deployment API
• Removed chunking_settings argument from inference.put_anthropic, inference.put_contextualai, inference.put_deepseek APIs

9.2.0 (2025-10-28)

Enhancements

• Support Trio when using the HTTPX async client ( #3089)
• Pydantic integration for the DSL module ( #3086)
• Add flush_after_seconds option to streaming_bulk() ( #3064)
• Add TS, FUSE and INLINE STATS commands to the ES|QL query builder ( #3096)

Bug Fixes

• DSL: support passing inner documents as AttrDict instances ( #3080)
• DSL: add some recently added field classes as top-level exports for the package ( #3078)

API

• Add source_exclude_vectors to Field Caps, Scroll and Search APIs
• Add require_data_stream to Index API
• Add settings_filter to Cluster Get Component Template API
• Add cause to Cluster Put Component Template API
• Add master_timeout to Cluster State API
• Add ccs_minimize_roundtrips to EQL Search API
• Add keep_alive and keep_on_completion to ES|QL Async Query API
• Add format to ES|QL Async Query Get API
• Add ES|QL Get Query and List Queries APIs
• Add Indices Delete Data Stream Options API
• Add Indices Get Data Stream Mappings and Put Data Stream Mappings APIs
• Add Indices Get Data Stream Options and Put Data Stream Options APIS
• Add Indices Get Data Stream Settings and Put Data Stream Settings APIs
• Add allow_no_indices, expand_wildcards and ignore_available to Indices Recovery API
• Add Indices Remove Block API
• Add input_type to Inference API
• Add timeout to all Inference Put APIs
• Add Inference Put Custom API
• Add Inference Put DeepSeek API
• Add task_settings to Put HuggingFace API
• Add streams namespace with streams.logs_disable, streams.logs_enable, streams.status APIs
• Add inference.contextualai API
• Add security.get_stats API
• Add bytes and time parameters to various APIs in the cat namespace.
• Add include_execution_metadata parameter to esql.async_query and esql.query APIs
• Add index_template parameter to indices.simulate_index_template API
• Add input_type parameter to inference.text_embedding API
• Add field_access_pattern parameter to ingest.put_pipeline API
• Add refresh to Security Grant API Key API
• Add wait_for_completion to the Snapshot Delete API
• Add state to Snapshot Get API
• Add refresh to Synonyms Put Synonym, Put Synonym Rule and Delete Synonym Rule APIs
• Remove unsupported size parameter from reindex API
• Remove deprecated if_primary_term, if_seq_no and op_type from Create API
• Remove deprecated master_timeout from Ingest Get Ip Location Database API
• Remove deprecated application, priviledge and username from the Security Get User API
• Rename type_query_string to type in License Post Start Trial API

Serverless-specific

• Add project namespace with project.tags API
• Add project_routing parameter to count, field_caps, msearch, msearch_template, open_point_in_time, search, search_mvt, search_template, async_search.submit, cat.count, eql.search, indices.resolve_index, sql.query APIs

DSL

• New CartesianBounds, CartesianCentroid, ChangePoint aggregations
• Add p_value parameter to SignificantTerms aggregation
• Add fields parameter to SemanticText field
• Add visit_percentage parameter to Knn query
• Add sparse_vector field to SemanticTextIndexOptions type
• Add index_options to SparseVector type
• Add separator_group and separators to ChunkingSettings type
• Add SparseVectorIndexOptions type
• Add key to FiltersBucket type

Other

• Add 3.14 to CI builds ( #3103)
• Drop Python 3.9 support ( #3114)

9.1.2 (2025-10-28)

Enhancements

• Add flush_after_seconds option to streaming_bulk() ( #3064)

Bug Fixes

API

• Add streams namespace with streams.logs_disable, streams.logs_enable, streams.status APIs
• Add bytes and time parameters to various APIs in the cat namespace.
• Add index_template parameter to indices.simulate_index_template API
• Add input_type parameter to inference.text_embedding API

DSL

• New CartesianBounds, CartesianCentroid, ChangePoint aggregations
• Add p_value parameter to SignificantTerms aggregation
• Add index_options and fields parameters to SemanticText field
• Add visit_percentage parameter to Knn query

Other

• Add 3.14 to CI builds ( #3103)

9.1.1 (2025-09-11)

Enhancements

• ES|QL query builder integration with the DSL module ( #3058)
• ES|QL query builder robustness fixes ( #3017)
• Fix ES|QL multi_match() signature ( #3052)

API

• Add support for ES|QL query builder objects to ES|QL Query and Async Query APIs
• Add Transform Set Upgrade Mode API
• Fix type of fields parameter of Term Vectors API to array of strings
• Fix type of params parameter of SQL Query API to array

DSL

• Preserve the skip_empty setting in to_dict() recursive serializations ( #3041)
• Add separator_group and separators attributes to ChunkingSettings type
• Add primary attribute to ShardFailure type
• Fix type of key attribute of ArrayPercentilesItem to float

9.1.0 (2025-07-30)

Enhancements

• ES|QL query builder (technical preview) ( #2997)
• Update OpenTelemetry conventions ( #2999)
• Add option to disable accurate reporting of file and line location in warnings (Fixes #3003) ( #3006)

APIs

• Remove if_primary_term, if_seq_no and op_type from Create API
• Remove master_timeout from Ingest Get Ip Location Database API
• Remove application, priviledge and username from the Security Get User API
• Rename type_query_string to type in License Post Start Trial API
• Add require_data_stream to Index API
• Add settings_filter to Cluster Get Component Template API
• Add cause to Cluster Put Component Template API
• Add master_timeout to Cluster State API
• Add ccs_minimize_roundtrips to EQL Search API
• Add keep_alive and keep_on_completion to ES|QL Async Query API
• Add format to ES|QL Async Query Get API
• Add ES|QL Get Query and List Queries APIs
• Add Indices Delete Data Stream Options API
• Add Indices Get Data Stream Options and Put Data Stream Options APIS
• Add Indices Get Data Stream Settings and Put Data Stream Settings APIs
• Add allow_no_indices, expand_wildcards and ignore_available to Indices Recovery API
• Add Indices Remove Block API
• Add Amazon Sagemaker to Inference API
• Add input_type to Inference API
• Add timeout to all Inference Put APIs
• Add Inference Put Custom API
• Add Inference Put DeepSeek API
• Add refresh to Security Grant API Key API
• Add wait_for_completion to the Snapshot Delete API
• Add state to Snapshot Get API
• Add refresh to Synonyms Put Synonym, Put Synonym Rule and Delete Synonym Rule APIs

DSL

• Handle lists in copy_to option in DSL field declarations correctly (Fixes #2992) ( #2993)
• Add index_options to SparseVector type
• Add SparseVectorIndexOptions type
• Add key to FiltersBucket type

Other changes

• Drop support for Python 3.8 ( #3001)

9.0.4 (2025-09-11)

Enhancements

API

DSL

• Preserve the skip_empty setting in to_dict() recursive serializations ( #3041)
• Add primary attribute to ShardFailure type
• Fix type of key attribute of ArrayPercentilesItem to float

9.0.3 (2025-07-30)

Enhancements

• ES|QL query builder (technical preview) ( #2997)
• Add option to disable accurate reporting of file and line location in warnings (Fixes #3003) ( #3006)

APIs

• Remove if_primary_term, if_seq_no and op_type from Create API
• Remove stored_fields from Get Source API
• Remove master_timeout from Ingest Get Ip Location Database API
• Remove application, priviledge and username from the Security Get User API
• Rename type_query_string to type in License Post Start Trial API
• Add require_data_stream to Index API
• Add settings_filter to Cluster Get Component Template API
• Add cause to Cluster Put Component Template API
• Add ccs_minimize_roundtrips to EQL Search API
• Add keep_alive and keep_on_completion to ES|QL Async Query API
• Add format to ES|QL Async Query Get API
• Add allow_no_indices, expand_wildcards and ignore_available to Indices Recovery API
• Add timeout to all Inference Put APIs
• Add refresh to Security Get User Profile API
• Add wait_for_completion to the Snapshot Delete API

DSL

• Handle lists in copy_to field option correctly (Fixes #2992) ( #2993)
• Add key to FiltersBucket type

9.0.2 (2025-06-05)

DSL

• Add rescore_vector to DenseVectorIndexOptions

9.0.1 (2025-04-28)

Features and enhancements

• Surface caused_by in ApiError ( #2932)
• Clarify Elasticsearch 9.x compatibility ( #2928)
• Reorganize Sphinx docs to only include reference pages ( #2776)

9.0.0 (2025-04-15)

Tip

Upgrade to Elasticsearch 9 before using elasticsearch-py 9.0.0 or later. Using elasticsearch-py 9.0.0 on an Elasticsearch 8 server will fail. Since language clients are forward-compatible, you should first upgrade Elasticsearch, then the Elasticsearch client. See the compatibility documentation for more details.

Breaking changes

• Remove deprecated Elasticsearch() options ( #2840)
• Remove deprecated url_prefix and use_ssl options ( #2797)

Enhancements

• Merge elasticsearch-dsl-py _ package ( #2736)
• Add Python DSL documentation ( #2761)
• Autogenerate DSL field classes from schema ( #2780)
• Improve DSL documentation examples with class-based queries and type hints ( #2857)
• Document the use of param() in Python DSL methods ( #2861)
• Migrate documentation from AsciiDoc to Markdown format ( #2806)
• Document use of sub-clients ( #2798)
• Document how to make API calls ( #2843)
• Fix simulate sub-client documentation ( #2749)

APIs

• Remove deprecated /_knn_search API
• Remove Unfreeze an index API
• Remove min_compatible_shard_node from Search and Async Search Submit APIs
• Remove local parameter from cat alias, Alias exists, and Get alias APIs
• Remove verbose from Index segments API
• Remove include_model_definition from Get trained model configuration info API
• Remove wait_for_active_shards from experimental Get field usage stats API
• Support soft-deletes in connectors:
  • Add hard to Delete connector API
  • Add include_deleted to Get and List Connector APIs
• Add master_timeout to Migrate to data tiers routing APIs
• Add master_timeout to the Alias exists and Get alias APIs.
• Add expand_wildcards to Create snapshot API
• Rename incorrect access_token to token in Logout of OpenID Connect API
• Add inference APIs: Alibaba Cloud AI Search, Amazon Bedrock, Anthropic, Azure AI Studio, Azure OpenAI, Cohere, Elastic Inference Service (EIS), Elasticsearch, ELSER, Google AI Studio, Google Vertex AI, Hugging Face, Jina AI, Mistral, OpenAI, and Voyage AI
• Add Elastic Inference Service (EIS) chat completion API
• Add Reindex legacy backing indices APIs
• Add Create an index from a source index API
• Add include_source_on_error to Create, Index, Update and Bulk APIs
• Add Stop async ES|QL query API
• Add timeout to Resolve Cluster API
• Add adaptive_allocations body field to Start and Update a trained model deployment API
• Rename index_template_subtitutions to index_template_substitutions in Simulate data ingestion API
• Add if_primary_term, if_seq_no, op_type, require_alias and require_data_stream to Create API
• Add max_concurrent_shard_requests to Open point in time API
• Add local and flat_settings to Check index templates API
• Add reopen to Update index settings API
• Add resource to Reload search analyzer API
• Add lazy to Roll over to a new index API
• Add cause and create to Simulate index template APIs
• Add inference APIs: Alibaba Cloud AI Search, Amazon Bedrock, Anthropic, Azure AI Studio, Azure OpenAI, Cohere, Elasticsearch, ELSER, Google AI Studio, Google Vertex AI, Hugging Face, Jina AI, Mistral, OpenAI, and Voyage AI

DSL

• Add ignore_malformed, script, on_script_error and time_series_dimension to Boolean field
• Add index to GeoShape field
• Add search_inference_id to SemanticText field