Vectorsearch providers should utilize the global Langchain.logger

.env.example

@@ -17,7 +17,7 @@ HUGGING_FACE_API_KEY=
 LLAMACPP_MODEL_PATH=
 LLAMACPP_N_THREADS=
 LLAMACPP_N_GPU_LAYERS=
-MILVUS_URL=
+MILVUS_URL=http://localhost:19530
 MISTRAL_AI_API_KEY=
 NEWS_API_KEY=
 OLLAMA_URL=http://localhost:11434

CHANGELOG.md

@@ -1,5 +1,8 @@
 ## [Unreleased]
+- [BREAKING] Langchain::Vectorsearch::Milvus was rewritten to work with newer milvus 0.10.0 gem
 - Assistant can now process image_urls in the messages (currently only for OpenAI and Mistral AI)
+- Vectorsearch providers utilize the global Langchain.logger
+- Update required milvus, qdrant and weaviate versions
 
 ## [0.16.1] - 2024-09-30
 - Deprecate Langchain::LLM::GooglePalm

Gemfile.lock

@@ -157,6 +157,7 @@ GEM
       faraday (~> 2.0)
       typhoeus (~> 1.4)
     ffi (1.16.3)
+    fiber-storage (1.0.0)
     google-cloud-env (2.1.1)
       faraday (>= 1.0, < 3.a)
     google_palm_api (0.1.3)
@@ -169,12 +170,13 @@ GEM
       multi_json (~> 1.11)
       os (>= 0.9, < 2.0)
       signet (>= 0.16, < 2.a)
-    graphlient (0.7.0)
+    graphlient (0.8.0)
       faraday (~> 2.0)
       graphql-client
-    graphql (2.3.4)
+    graphql (2.3.16)
       base64
-    graphql-client (0.22.0)
+      fiber-storage
+    graphql-client (0.23.0)
       activesupport (>= 3.0)
       graphql (>= 1.13.0)
     hashdiff (1.1.0)
@@ -211,7 +213,7 @@ GEM
       net-smtp
     matrix (0.4.2)
     method_source (1.1.0)
-    milvus (0.9.3)
+    milvus (0.10.3)
       faraday (>= 2.0.1, < 3)
     mini_mime (1.1.5)
     mini_portile2 (2.8.6)
@@ -282,7 +284,7 @@ GEM
     psych (5.1.2)
       stringio
     public_suffix (5.0.5)
-    qdrant-ruby (0.9.7)
+    qdrant-ruby (0.9.8)
       faraday (>= 2.0.1, < 3)
     racc (1.8.0)
     rack (3.0.11)
@@ -419,9 +421,9 @@ GEM
       parser (>= 3.3.0)
     uri (0.13.0)
     vcr (6.2.0)
-    weaviate-ruby (0.8.10)
+    weaviate-ruby (0.9.2)
       faraday (>= 2.0.1, < 3.0)
-      graphlient (~> 0.7.0)
+      graphlient (>= 0.7.0, < 0.9.0)
     webmock (3.23.1)
       addressable (>= 2.8.0)
       crack (>= 0.3.2)
@@ -461,7 +463,7 @@ DEPENDENCIES
   langchainrb!
   llama_cpp (~> 0.9.4)
   mail (~> 2.8)
-  milvus (~> 0.9.3)
+  milvus (~> 0.10.3)
   mistral-ai
   nokogiri (~> 1.13)
   pdf-reader (~> 2.0)
@@ -470,7 +472,7 @@ DEPENDENCIES
   pinecone (~> 0.1.6)
   power_point_pptx (~> 0.1.0)
   pry-byebug (~> 3.10.0)
-  qdrant-ruby (~> 0.9.4)
+  qdrant-ruby (~> 0.9.8)
   rake (~> 13.0)
   rdiscount (~> 2.2.7)
   replicate-ruby (~> 0.2.2)
@@ -483,7 +485,7 @@ DEPENDENCIES
   sequel (~> 5.68.0)
   standard (>= 1.35.1)
   vcr
-  weaviate-ruby (~> 0.8.10)
+  weaviate-ruby (~> 0.9.2)
   webmock
   wikipedia-client (~> 1.17.0)
   yard (~> 0.9.34)

langchain.gemspec

@@ -57,7 +57,7 @@ Gem::Specification.new do |spec|
   spec.add_development_dependency "google_search_results", "~> 2.0.0"
   spec.add_development_dependency "hnswlib", "~> 0.8.1"
   spec.add_development_dependency "hugging-face", "~> 0.3.4"
-  spec.add_development_dependency "milvus", "~> 0.9.3"
+  spec.add_development_dependency "milvus", "~> 0.10.3"
   spec.add_development_dependency "llama_cpp", "~> 0.9.4"
   spec.add_development_dependency "nokogiri", "~> 1.13"
   spec.add_development_dependency "mail", "~> 2.8"
@@ -67,13 +67,13 @@ Gem::Specification.new do |spec|
   spec.add_development_dependency "pdf-reader", "~> 2.0"
   spec.add_development_dependency "pinecone", "~> 0.1.6"
   spec.add_development_dependency "replicate-ruby", "~> 0.2.2"
-  spec.add_development_dependency "qdrant-ruby", "~> 0.9.4"
+  spec.add_development_dependency "qdrant-ruby", "~> 0.9.8"
   spec.add_development_dependency "roo", "~> 2.10.0"
   spec.add_development_dependency "roo-xls", "~> 1.2.0"
   spec.add_development_dependency "ruby-openai", "~> 7.1.0"
   spec.add_development_dependency "safe_ruby", "~> 1.0.4"
   spec.add_development_dependency "sequel", "~> 5.68.0"
-  spec.add_development_dependency "weaviate-ruby", "~> 0.8.10"
+  spec.add_development_dependency "weaviate-ruby", "~> 0.9.2"
   spec.add_development_dependency "wikipedia-client", "~> 1.17.0"
   spec.add_development_dependency "power_point_pptx", "~> 0.1.0"
 end

lib/langchain/vectorsearch/elasticsearch.rb

@@ -37,7 +37,7 @@ def initialize(url:, index_name:, llm:, api_key: nil, es_options: {})
       @options = {
         url: url,
         request_timeout: 20,
-        log: false
+        logger: Langchain.logger
       }.merge(es_options)
 
       @es_client = ::Elasticsearch::Client.new(**options)

lib/langchain/vectorsearch/milvus.rb

@@ -6,16 +6,18 @@ class Milvus < Base
     # Wrapper around Milvus REST APIs.
     #
     # Gem requirements:
-    #     gem "milvus", "~> 0.9.3"
+    #     gem "milvus", "~> 0.10.3"
     #
     # Usage:
-    # milvus = Langchain::Vectorsearch::Milvus.new(url:, index_name:, llm:, api_key:)
+    #     milvus = Langchain::Vectorsearch::Milvus.new(url:, index_name:, llm:, api_key:)
     #
-
     def initialize(url:, index_name:, llm:, api_key: nil)
       depends_on "milvus"
 
-      @client = ::Milvus::Client.new(url: url)
+      @client = ::Milvus::Client.new(
+        url: url,
+        logger: Langchain.logger
+      )
       @index_name = index_name
 
       super(llm: llm)
@@ -24,33 +26,24 @@ def initialize(url:, index_name:, llm:, api_key: nil)
     def add_texts(texts:)
       client.entities.insert(
         collection_name: index_name,
-        num_rows: Array(texts).size,
-        fields_data: [
-          {
-            field_name: "content",
-            type: ::Milvus::DATA_TYPES["varchar"],
-            field: Array(texts)
-          }, {
-            field_name: "vectors",
-            type: ::Milvus::DATA_TYPES["float_vector"],
-            field: Array(texts).map { |text| llm.embed(text: text).embedding }
-          }
-        ]
+        data: texts.map do |text|
+          {content: text, vector: llm.embed(text: text).embedding}
+        end
       )
     end
 
+    # TODO: Add update_texts method
+
     # Deletes a list of texts in the index
     #
     # @param ids [Array<Integer>] The ids of texts to delete
     # @return [Boolean] The response from the server
     def remove_texts(ids:)
       raise ArgumentError, "ids must be an array" unless ids.is_a?(Array)
-      # Convert ids to integers if strings are passed
-      ids = ids.map(&:to_i)
 
       client.entities.delete(
         collection_name: index_name,
-        expression: "id in #{ids}"
+        filter: "id in #{ids}"
       )
     end
 
@@ -62,33 +55,25 @@ def create_default_schema
       client.collections.create(
         auto_id: true,
         collection_name: index_name,
-        description: "Default schema created by langchain.rb",
         fields: [
           {
-            name: "id",
-            is_primary_key: true,
-            autoID: true,
-            data_type: ::Milvus::DATA_TYPES["int64"]
+            fieldName: "id",
+            isPrimary: true,
+            dataType: "Int64"
           }, {
-            name: "content",
-            is_primary_key: false,
-            data_type: ::Milvus::DATA_TYPES["varchar"],
-            type_params: [
-              {
-                key: "max_length",
-                value: "32768" # Largest allowed value
-              }
-            ]
+            fieldName: "content",
+            isPrimary: false,
+            dataType: "VarChar",
+            elementTypeParams: {
+              max_length: "32768" # Largest allowed value
+            }
           }, {
-            name: "vectors",
-            data_type: ::Milvus::DATA_TYPES["float_vector"],
-            is_primary_key: false,
-            type_params: [
-              {
-                key: "dim",
-                value: llm.default_dimensions.to_s
-              }
-            ]
+            fieldName: "vector",
+            isPrimary: false,
+            dataType: "FloatVector",
+            elementTypeParams: {
+              dim: llm.default_dimensions.to_s
+            }
           }
         ]
       )
@@ -97,27 +82,31 @@ def create_default_schema
     # Create the default index
     # @return [Boolean] The response from the server
     def create_default_index
-      client.indices.create(
+      client.indexes.create(
         collection_name: index_name,
-        field_name: "vectors",
-        extra_params: [
-          {key: "metric_type", value: "L2"},
-          {key: "index_type", value: "IVF_FLAT"},
-          {key: "params", value: "{\"nlist\":1024}"}
+        index_params: [
+          {
+            metricType: "L2",
+            fieldName: "vector",
+            indexName: "vector_idx",
+            indexConfig: {
+              index_type: "AUTOINDEX"
+            }
+          }
         ]
       )
     end
 
     # Get the default schema
     # @return [Hash] The response from the server
     def get_default_schema
-      client.collections.get(collection_name: index_name)
+      client.collections.describe(collection_name: index_name)
     end
 
     # Delete default schema
     # @return [Hash] The response from the server
     def destroy_default_schema
-      client.collections.delete(collection_name: index_name)
+      client.collections.drop(collection_name: index_name)
     end
 
     # Load default schema into memory
@@ -138,16 +127,12 @@ def similarity_search(query:, k: 4)
     def similarity_search_by_vector(embedding:, k: 4)
       load_default_schema
 
-      client.search(
+      client.entities.search(
         collection_name: index_name,
-        output_fields: ["id", "content"], # Add "vectors" if need to have full vectors returned.
-        top_k: k.to_s,
-        vectors: [embedding],
-        dsl_type: 1,
-        params: "{\"nprobe\": 10}",
-        anns_field: "vectors",
-        metric_type: "L2",
-        vector_type: ::Milvus::DATA_TYPES["float_vector"]
+        anns_field: "vector",
+        data: [embedding],
+        limit: k,
+        output_fields: ["content", "id", "vector"]
       )
     end
 
@@ -159,8 +144,7 @@ def similarity_search_by_vector(embedding:, k: 4)
     def ask(question:, k: 4, &block)
       search_results = similarity_search(query: question, k: k)
 
-      content_field = search_results.dig("results", "fields_data").select { |field| field.dig("field_name") == "content" }
-      content_data = content_field.first.dig("Field", "Scalars", "Data", "StringData", "data")
+      content_data = search_results.dig("data").map { |result| result.dig("content") }
 
       context = content_data.join("\n---\n")
 

lib/langchain/vectorsearch/qdrant.rb

@@ -6,7 +6,7 @@ class Qdrant < Base
     # Wrapper around Qdrant
     #
     # Gem requirements:
-    #     gem "qdrant-ruby", "~> 0.9.3"
+    #     gem "qdrant-ruby", "~> 0.9.8"
     #
     # Usage:
     #     qdrant = Langchain::Vectorsearch::Qdrant.new(url:, api_key:, index_name:, llm:)
@@ -22,7 +22,8 @@ def initialize(url:, api_key:, index_name:, llm:)
 
       @client = ::Qdrant::Client.new(
         url: url,
-        api_key: api_key
+        api_key: api_key,
+        logger: Langchain.logger
       )
       @index_name = index_name
 

lib/langchain/vectorsearch/weaviate.rb

@@ -6,7 +6,7 @@ class Weaviate < Base
     # Wrapper around Weaviate
     #
     # Gem requirements:
-    #     gem "weaviate-ruby", "~> 0.9.0"
+    #     gem "weaviate-ruby", "~> 0.9.2"
     #
     # Usage:
     #     weaviate = Langchain::Vectorsearch::Weaviate.new(url: ENV["WEAVIATE_URL"], api_key: ENV["WEAVIATE_API_KEY"], index_name: "Docs", llm: llm)
@@ -22,7 +22,8 @@ def initialize(url:, index_name:, llm:, api_key: nil)
 
       @client = ::Weaviate::Client.new(
         url: url,
-        api_key: api_key
+        api_key: api_key,
+        logger: Langchain.logger
       )
 
       # Weaviate requires the class name to be Capitalized: https://weaviate.io/developers/weaviate/configuration/schema-configuration#create-a-class