prospector_engine 0.1.2 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 8e754e44af2ba6797161e333fc790c36f3a5e8b79e30deb4dc5eef78ddb820c1
-  data.tar.gz: cf8338a50a7dfb29f2814b250dea974d69306c1e5c3ae7d69272663b4cd8cd43
+  metadata.gz: f6ff4fed78e9e8c81ec50623a7e2cbe0076b0f890f3c3fec3561551585de0513
+  data.tar.gz: 584620ecbee56ffdf94520638a5b7ab409e13c9fc264dd934a44d8e798bf4275
 SHA512:
-  metadata.gz: dc08f57c755577e9589a729e671d21b3729dad5a81a6e0b8129ffe432eb07edaaaba8a0ad755827c45247616f2a11517a0750ed295d5bdb498c22461401b6648
-  data.tar.gz: 22026bf0d973173b0c22d4b3c4648a0ac9957b83ae8dd164dab0cee6e3256bb3aa29f64a101207ed3f8084d1416337713df738835d6e1416abba16d16dfe8113
+  metadata.gz: 7bbc93fdaca17a89d0ba229fbd23bba95eef79cd723b288ec872162eab5801bc9bc4fa0f93506245b1e7a796ded93523c0d237d61d5463c60c603644cb7450e6
+  data.tar.gz: ac7828b1928c11db81a56079fa6b65442a8b85404a214d147283e02a69c4f9109b3fed4568da47548e1e9b758dc415206339ef3e19abeecde10d4e9588d7ec09

data/README.md

@@ -94,8 +94,9 @@ Prospector.configure do |config|
   # Optional: default source adapter (default: :google_places)
   config.default_source = :google_places
 
-  # Optional: default AI model for classification and keyword generation
-  config.default_classifier_model = "anthropic:claude-sonnet-4-20250514"
+  # Optional: model override for classification and keyword generation (a bare ruby_llm model id).
+  # Defaults to nil: classification uses the classifier's `model`, keyword generation uses RubyLLM's default.
+  # config.default_classifier_model = "claude-opus-4-6"
 
   # Optional: job queue name (default: :default)
   config.queue_name = :prospector
@@ -146,9 +147,8 @@ class MotorcycleClassifier < LlmClassifier::Classifier
              :detailing, :training, :towing, :insurance, :touring,
              :clubs, :parts
 
-  model "anthropic:claude-sonnet-4-20250514"
+  model "claude-opus-4-6"
   multi_label true
-  require_categories true
 
   system_prompt <<~PROMPT
     You are classifying businesses for the motorcycle services domain.
@@ -169,6 +169,10 @@ end
 
 The classifier receives a hash with `name`, `address`, `website`, `description`, and `source_types` (comma-joined string) keys.
 
+Use a bare model id that ruby_llm recognizes (`claude-opus-4-6`, not `anthropic:claude-opus-4-6`) and that supports structured outputs, since llm_classifier 0.3 sends every request with a JSON schema.
+
+Candidates the classifier assigns no categories are auto-rejected with reason `no_relevant_categories`. A failed classification (API error, unknown model, unparseable response) is recorded in `metadata["llm_error"]`, and the candidate stays pending so the next reclassification retries it. Leave `require_categories` off: with it on, llm_classifier reports "no categories" as a failure, so irrelevant businesses are counted as errors instead of being rejected.
+
 ### Programmatic Usage
 
 ```ruby
@@ -291,7 +295,8 @@ Prospector::Run (state machine)
   |     '-- Classification::Runner
   |           |-- Calls LLM via llm_classifier
   |           |-- Stores categories, confidence, reasoning in metadata
-  |           '-- Auto-rejects candidates with no relevant categories
+  |           |-- Auto-rejects candidates with no relevant categories
+  |           '-- Records failed classifications as llm_error, leaves them pending
   |
   '-- Admin UI
         |-- Runs: create, monitor, retry, restart, cancel

data/app/controllers/prospector/run_reclassifications_controller.rb

@@ -12,7 +12,7 @@ module Prospector
 
     def build_classification_run(run)
       run.classification_runs.new(
-        ai_model: params[:ai_model] || Prospector.config.default_classifier_model,
+        ai_model: params[:ai_model].presence || Prospector.config.resolved_classifier_model,
         include_approved: params[:include_approved] == "1",
         actor_id: respond_to?(:current_user) ? current_user&.id : nil
       )

data/lib/generators/prospector/install/templates/prospector.rb

@@ -24,8 +24,9 @@ Prospector.configure do |config|
   # Optional: default source adapter (default: :google_places)
   # config.default_source = :google_places
 
-  # Optional: default AI model for classification (passed to classifier)
-  # config.default_classifier_model = "anthropic:claude-sonnet-4-20250514"
+  # Optional: model override for classification and keyword generation (a bare ruby_llm model id)
+  # Defaults to nil: classification uses the classifier's `model`, keyword generation uses RubyLLM's default
+  # config.default_classifier_model = "claude-opus-4-6"
 
   # Optional: duplicate check hook
   # config.duplicate_check do |source_uid:, name:, **|

data/lib/prospector/CLAUDE.md

@@ -39,7 +39,7 @@ Key methods: `to_query_string` (for text search), `to_coordinate_restriction` (f
 
 ## Classification (`classification/`)
 
-- `Runner` - Iterates unprocessed candidates, calls LLM via ruby_llm, stores results in candidate metadata. Auto-rejects candidates with empty categories.
+- `Runner` - Iterates unprocessed candidates, calls the configured `LlmClassifier::Classifier`, stores results in candidate metadata. Auto-rejects candidates with empty categories. A failed `Result` (or raised error) is stored as `llm_error` and counted as an error, never a rejection; the candidate stays `ai_unprocessed` so it is retried.
 
 ## Keywords (`keywords/`)
 
@@ -47,6 +47,6 @@ Key methods: `to_query_string` (for text search), `to_coordinate_restriction` (f
 
 ## Other Files
 
-- `configuration.rb` - `Prospector.configure` DSL. Required keys: domain, on_approve, authenticate_admin_with.
+- `configuration.rb` - `Prospector.configure` DSL. Required keys: domain, on_approve, authenticate_admin_with. `default_classifier_model` defaults to nil so the classifier's own `model` applies; `resolved_classifier_model` reports the model that will actually run.
 - `engine.rb` - Rails::Engine with isolate_namespace.
 - `error.rb` - Error hierarchy: Error, ConfigurationError, AdapterError, MissingApiKeyError, ClassificationError.

data/lib/prospector/classification/runner.rb

@@ -34,25 +34,24 @@ module Prospector
 
       def classify_candidate(candidate)
         result = invoke_classifier(candidate)
+        return record_error(candidate, result.error) if result.failure?
+
         store_result(candidate, result)
         auto_reject_if_empty(candidate, result)
         @processed += 1
       rescue => e
-        candidate.update!(metadata: candidate.metadata.merge("llm_error" => e.message))
-        @errors += 1
+        record_error(candidate, e.message)
       end
 
       def invoke_classifier(candidate)
-        classifier = resolve_classifier
+        classifier = Prospector.config.classifier_class
         input = build_input(candidate)
         classifier.classify(input, model: @model)
       end
 
-      def resolve_classifier
-        raw = Prospector.config.classifier
-        raise ConfigurationError, "Prospector.config.classifier must be set" unless raw
-
-        raw.is_a?(String) ? raw.constantize : raw
+      def record_error(candidate, message)
+        candidate.update!(metadata: candidate.metadata.merge("llm_error" => message))
+        @errors += 1
       end
 
       def build_input(candidate)
@@ -66,7 +65,7 @@ module Prospector
       end
 
       def store_result(candidate, result)
-        candidate.update!(metadata: candidate.metadata.merge(
+        candidate.update!(metadata: candidate.metadata.except("llm_error").merge(
           "llm_categories" => result.categories,
           "llm_confidence" => result.confidence,
           "llm_reasoning" => result.reasoning,

data/lib/prospector/configuration.rb

@@ -12,7 +12,7 @@ module Prospector
     def initialize
       @domain = nil
       @default_source = :google_places
-      @default_classifier_model = "anthropic:claude-sonnet-4-20250514"
+      @default_classifier_model = nil
       @classifier = nil
       @turbo_enabled = defined?(Turbo::Broadcastable)
       @queue_name = :default
@@ -48,6 +48,18 @@ module Prospector
       adapter.is_a?(String) ? adapter.constantize : adapter
     end
 
+    def classifier_class
+      raise ConfigurationError, "Prospector.config.classifier must be set" unless classifier
+
+      classifier.is_a?(String) ? classifier.constantize : classifier
+    end
+
+    def resolved_classifier_model
+      require "ruby_llm"
+
+      default_classifier_model || classifier_class.model || RubyLLM.config.default_model
+    end
+
     def validate!
       raise ConfigurationError, "Prospector.config.domain must be set" if domain.blank?
       raise ConfigurationError, "Prospector.config.authenticate_admin_with must be set" unless @authenticate_admin_block

data/lib/prospector/keywords/generator.rb

@@ -30,6 +30,7 @@ module Prospector
 
         chat = RubyLLM.chat(model: Prospector.config.default_classifier_model)
         response = chat.ask(prompt_for(category))
+        @generated_by = chat.model.id
         parse_keywords(response.content)
       end
 
@@ -63,7 +64,7 @@ module Prospector
         ) do |record|
           record.source = "llm"
           record.metadata = {
-            "generated_by" => Prospector.config.default_classifier_model,
+            "generated_by" => @generated_by,
             "generated_at" => Time.current.iso8601
           }
         end

data/lib/prospector/version.rb

@@ -1,3 +1,3 @@
 module Prospector
-  VERSION = "0.1.2"
+  VERSION = "0.2.0"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: prospector_engine
 version: !ruby/object:Gem::Version
-  version: 0.1.2
+  version: 0.2.0
 platform: ruby
 authors:
 - AxiumFoundry
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-04-06 00:00:00.000000000 Z
+date: 2026-09-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rails
@@ -42,16 +42,16 @@ dependencies:
   name: llm_classifier
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.2.0
+        version: 0.3.0
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.2.0
+        version: 0.3.0
 - !ruby/object:Gem::Dependency
   name: minitest
   requirement: !ruby/object:Gem::Requirement
@@ -178,7 +178,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.4.20
+rubygems_version: 3.5.22
 signing_key:
 specification_version: 4
 summary: Business discovery from multiple sources with AI classification