vector_amp 0.5.0 → 0.5.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 47c268daf0e7c0577fdff02c84f948bb2e394d607683464b6248d5039af98d39
-  data.tar.gz: c7a616a0458451afc28163c91c7c221f5b7cb5892203766cf59daa33c961d8d8
+  metadata.gz: 70c698bb53d97209e9eeac1751e5e1630f3f71b4bd5f06a1ba65e807ddc573ac
+  data.tar.gz: e651d9b66da88dff32f1bf913db8fa86fac51bde05497cacf65372086e6ca4fd
 SHA512:
-  metadata.gz: ecff0393b786f01d56c48376a2c5bc27cf718ddba580c3e4bf951e9e8adb35acea6e76309f1be2342dbee71e881cd2bfbbd70f663ae5cb95ae08d5cd0977e419
-  data.tar.gz: b89d436c1d51a0b6feca1819bb1996a335da6ab03b6417c866a4e7811e45067b22304a077f70edf1cf646f26c255a1ca0b3a831a98aa98781c626795c3e1f391
+  metadata.gz: 2b1c6a7837315f1c0b9d76c41dd354d17109a0175a695f79026db437424d090c41b535937d07658923049e4715900fda0080796f76ea4624049a215f0ad5fcfe
+  data.tar.gz: e49eae451bcd0e9070fb42dbd4fc05bc6bec9b56d347474a1b4dde07d2960ee176bdb696a1bcdebeeb3e6c58e9cb3a1759f30ba9ed7999ea496e29f91ff29395

data/README.md

@@ -306,10 +306,12 @@ Non-streaming:
 ```ruby
 answer = client.ask(
   "What are the key features?",
-  dataset_id: "all",
-  top_k: 5
+  top_k: 5           # omit dataset_ids to search every dataset you can see
 )
 puts answer["answer"]
+
+# Scope one question to any number of datasets:
+answer = client.ask("Which contracts renew in Q4?", dataset_ids: %w[ds_contracts ds_invoices])
 ```
 
 Multi-turn conversations: the Intelligence API is stateless, so send prior turns
@@ -323,7 +325,6 @@ history = [
 
 follow_up = client.intelligence.query(
   "Which of those help with relevance?",
-  dataset_id: "all",
   conversation_history: history.last(10) # include as many prior turns as you want
 )
 puts follow_up["answer"]
@@ -332,7 +333,7 @@ puts follow_up["answer"]
 Streaming SSE:
 
 ```ruby
-client.ask_stream("Summarize the docs", dataset_id: "dataset-uuid") do |event|
+client.ask_stream("Summarize the docs", dataset_ids: ["dataset-uuid"]) do |event|
   print event["content"] if event["chunk_type"] == "text"
 end
 
@@ -409,7 +410,7 @@ Dataset-object helpers: `search`, `insert`, `add_texts`, `embed`, `delete`, `sta
 
 | Method | Required | Optional (defaults) |
 |---|---|---|
-| `query(query)` / `ask` / `ask_stream` | `query` | `dataset_id` ("all" when unscoped), `top_k` (5 server-side), `conversation_history`, `include_sources`, `stream` |
+| `query(query)` / `ask` / `ask_stream` | `query` | `dataset_ids` (array; omit to search every visible dataset), `top_k` (5 server-side), `conversation_history`, `include_sources`, `stream` |
 | `create_session` | — | `title`, `dataset_id`, `workspace_id`, `metadata` |
 | `list_sessions` | — | `limit` (50) |
 | `get_session(session_id)` | `session_id` | — |

data/lib/vector_amp/dataset.rb

@@ -129,11 +129,11 @@ module VectorAmp
 
     # Ask an intelligence question constrained to this dataset.
     # @param query [String] natural-language question.
-    # @param options [Hash] forwarded to {Client#ask}; `dataset_id` is set to this dataset id.
+    # @param options [Hash] forwarded to {Client#ask}; `dataset_ids` is set to this dataset id.
     # @return [Hash] intelligence response.
     def ask(query, **options)
       require_client!("ask")
-      client.ask(query, **options.merge(dataset_id: id))
+      client.ask(query, **options.merge(dataset_ids: [id]))
     end
 
     # Upload local files by auto-creating a `file_upload` source, initializing presigned uploads, and completing the upload job.

data/lib/vector_amp/intelligence.rb

@@ -11,20 +11,24 @@ module VectorAmp
       @transport = transport
     end
 
-    # Ask an intelligence query, optionally scoped to a dataset and streamed.
+    # Ask an intelligence query, optionally scoped to datasets and streamed.
     # @param query [String] natural-language question.
-    # @param dataset_id [String, nil] optional dataset id scope.
+    # @param dataset_ids [Array<String>, String, nil] datasets to scope the question to; omit to
+    #   search every dataset the API key can see.
     # @param top_k [Integer, nil] optional retrieval result count.
     # @param conversation_history [Array<Hash>, nil] optional prior conversation messages.
     # @param include_sources [Boolean, nil] include source chunks/citations when supported.
     # @param stream [Boolean] stream chunks when true; defaults to false.
     # @yieldparam chunk [Object] streamed response chunk when stream is true.
     # @return [Hash, Enumerator, Object] response hash, enumerator without a stream block, or transport stream result.
-    def query(query, dataset_id: nil, top_k: nil, conversation_history: nil, include_sources: nil, stream: false, **unknown, &block)
+    def query(query, dataset_ids: nil, top_k: nil, conversation_history: nil, include_sources: nil, stream: false, **unknown, &block)
       Utils.ensure_no_unknown!(unknown, "query")
+      scope = self.class.normalize_dataset_ids(dataset_ids)
       body = Utils.compact_hash(
         query: query,
-        dataset_id: dataset_id,
+        # An absent dataset_ids is how the API says "every dataset you can see";
+        # an empty array would be a narrower, different request.
+        dataset_ids: scope.empty? ? nil : scope,
         top_k: top_k,
         conversation_history: conversation_history,
         include_sources: include_sources,
@@ -32,7 +36,7 @@ module VectorAmp
       )
 
       if stream
-        return enum_for(:query, query, dataset_id: dataset_id, top_k: top_k,
+        return enum_for(:query, query, dataset_ids: dataset_ids, top_k: top_k,
                         conversation_history: conversation_history, include_sources: include_sources,
                         stream: true) unless block
 
@@ -42,6 +46,18 @@ module VectorAmp
       end
     end
 
+    # Normalize a dataset scope for the wire.
+    #
+    # `POST /intelligence/query` scopes with `dataset_ids` and reads an absent field as "every
+    # dataset the caller can see". The singular `dataset_id` is retired and now draws a 400, and
+    # the `"all"` sentinel it carried says nothing an omitted field does not, so both are dropped
+    # here rather than transmitted.
+    # @param dataset_ids [Array<String>, String, nil] requested scope.
+    # @return [Array<String>] normalized dataset ids.
+    def self.normalize_dataset_ids(dataset_ids)
+      Array(dataset_ids).map { |id| id.to_s.strip }.reject { |id| id.empty? || id == "all" }
+    end
+
     # Create an intelligence conversation session.
     # @param title [String, nil] optional session title.
     # @param dataset_id [String, nil] optional dataset scope.

data/lib/vector_amp/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module VectorAmp
-  VERSION = "0.5.0"
+  VERSION = "0.5.1"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: vector_amp
 version: !ruby/object:Gem::Version
-  version: 0.5.0
+  version: 0.5.1
 platform: ruby
 authors:
 - VectorAmp