fix(captain): use metadata for openai_file_id and add pdf-reader gem

Gemfile.lock

@@ -25,6 +25,7 @@ GIT
 GEM
   remote: https://rubygems.org/
   specs:
+    Ascii85 (2.0.1)
     actioncable (7.1.5.2)
       actionpack (= 7.1.5.2)
       activesupport (= 7.1.5.2)
@@ -126,6 +127,7 @@ GEM
       jbuilder (~> 2)
       rails (>= 4.2, < 7.2)
       selectize-rails (~> 0.6)
+    afm (1.0.0)
     ai-agents (0.7.0)
       ruby_llm (~> 1.8.2)
     annotaterb (4.20.0)
@@ -429,6 +431,7 @@ GEM
     hana (1.3.7)
     hash_diff (1.1.1)
     hashdiff (1.1.0)
+    hashery (2.1.2)
     hashie (5.0.0)
     html2text (0.4.0)
       nokogiri (>= 1.0, < 2.0)
@@ -654,6 +657,12 @@ GEM
     parser (3.3.8.0)
       ast (~> 2.4.1)
       racc
+    pdf-reader (2.15.1)
+      Ascii85 (>= 1.0, < 3.0, != 2.0.0)
+      afm (>= 0.2.1, < 2)
+      hashery (~> 2.0)
+      ruby-rc4
+      ttfunk
     pg (1.5.3)
     pg_search (2.3.6)
       activerecord (>= 5.2)
@@ -813,6 +822,7 @@ GEM
       faraday (>= 1)
       faraday-multipart (>= 1)
     ruby-progressbar (1.13.0)
+    ruby-rc4 (0.1.5)
     ruby-saml (1.18.1)
       nokogiri (>= 1.13.10)
       rexml
@@ -945,6 +955,8 @@ GEM
       i18n
     timeout (0.4.3)
     trailblazer-option (0.1.2)
+    ttfunk (1.8.0)
+      bigdecimal (~> 3.1)
     twilio-ruby (7.6.0)
       faraday (>= 0.9, < 3.0)
       jwt (>= 1.5, < 3.0)
@@ -1097,6 +1109,7 @@ DEPENDENCIES
   opensearch-ruby
   opentelemetry-exporter-otlp
   opentelemetry-sdk
+  pdf-reader
   pg
   pg_search
   pgvector
@@ -1164,4 +1177,4 @@ RUBY VERSION
    ruby 3.4.4p34
 
 BUNDLED WITH
-  4.0.3
+   2.5.11

enterprise/app/services/captain/llm/pdf_processing_service.rb

@@ -29,8 +29,9 @@ class Captain::Llm::PdfProcessingService < Llm::LegacyBaseOpenAiService
 
     if content.present?
       Rails.logger.info "PDF extracted content for document #{document.id} (chars=#{content.length})"
-      # Update content and ensure openai_file_id is nil to force standard FAQ generation
-      document.update!(content: content, openai_file_id: nil)
+      # Update content and clear openai_file_id in metadata to force standard FAQ generation.
+      metadata = (document.metadata || {}).merge('openai_file_id' => nil)
+      document.update!(content: content, metadata: metadata)
     else
       Rails.logger.warn "PDF extracted content is empty for document #{document.id}"
     end