summarize-meeting 1.0.0 → 1.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 502d04b52ca4fb6aeba632d7d9959ef043db57cd7f70f9de278a5fc70f52a034
-  data.tar.gz: b1aae558f143f92a5d21e0f5268025768ef517b1ace5773c5f4b44a0b9fab10d
+  metadata.gz: 50b154d8f5d117b51fb3a377ad725088262543c05d5ee2614929511ad093a553
+  data.tar.gz: 66f0a4144e8b5bf6d36792bdb0432014b803e4eff020060fc9d5045c1631edef
 SHA512:
-  metadata.gz: 781f9a4108fce791c7355e73161922b1271d8c7aefadc878dedadc5dfeec5b7a408a454c0d93d20912dd897a16b8c017ea6a4fcda27d2859fd90e252538163ae
-  data.tar.gz: 7b01038c1cd4a0473d98338131866fb751b471fcdffa49e91619ce12d0e83cf8dddf68814dbaf2c822e29d3d43c9260a0dbcb44d304f6bd4f4aee9e262ca8210
+  metadata.gz: fd077cdfe8a177d145a5ab4672d14477cbd7a06124f0eeb03c834ef5eb632022d76920ab1ee6cba38e81204f91faa48cb1bdcc1683a86db1f7305e232cc3380a
+  data.tar.gz: fc2061143753fc8865b39702cfc215ae7732f8e697f8288170f07a8ba1f627c3078f4d37c40e725506413a2718cb5b7b3f93d427cd6d94a3d1ee7f0d30246192

data/lib/summarize-meeting/ai.rb

@@ -1,11 +1,17 @@
-require "openai"
-
 module SummarizeMeeting
   module Ai
+    class OpenAiError < StandardError; end
+    MAX_TOTAL_TOKENS = 4096
+    WORDS_PER_TOKEN = 0.75
+
     @@access_token = ENV["OPENAI_KEY"]
     @@organization_id = ENV["OPENAI_ORG"]
 
     def self.client
+      @client ||= new_client(access_token: access_token, organization_id: organization_id)
+    end
+
+    def self.new_client(access_token:, organization_id:)
       OpenAI::Client.new(access_token: access_token, organization_id: organization_id)
     end
 
@@ -24,5 +30,27 @@ module SummarizeMeeting
     def self.organization_id=(id)
       @@organization_id = id
     end
+
+    def self.calculate_token_word_count(token_count)
+      (token_count * WORDS_PER_TOKEN.to_f).ceil
+    end
+
+    def self.calculate_word_token_count(word_count)
+      (word_count / WORDS_PER_TOKEN.to_f).ceil
+    end
+
+    def self.chat(messages, client: self.client)
+      parameters = {
+        model: "gpt-3.5-turbo",
+        messages: messages,
+      }
+      response = client.chat(parameters: parameters)
+      content = response.dig("choices", 0, "message", "content")
+      if !content
+        raise OpenAiError, "No response from OpenAI"
+      else
+        content
+      end
+    end
   end
 end

data/lib/summarize-meeting/meeting.rb

@@ -4,6 +4,8 @@ require "openai"
 
 module SummarizeMeeting
   class Meeting
+    RESPONSE_RESERVE_TOKENS = 500
+
     LINE_SUMMARY_PROMPT_TEMPLATE = [
       {
         role: "system",
@@ -11,7 +13,7 @@ module SummarizeMeeting
       },
       {
         role: "system",
-        content: "The transcript of the meeting is split into {{chunkCount}} chunks. This is the {{chunkIndex}} chunk.",
+        content: "The transcript of the meeting is split into {{chunkCount}} chunks. This is the chunk number {{chunkIndex}} of {{chunkCount}}.",
       },
       {
         role: "assistant",
@@ -60,17 +62,14 @@ module SummarizeMeeting
     attr_reader :transcript
 
     def summarize
-
       # Step 1. Split the transcript into lines.
       lines = transcript.lines
 
       # Step 2. Calculate the maximum chunk size in words.
-      max_total_tokens = 4000
-      response_token_reserve = 500
-      template_tokens = LINE_SUMMARY_PROMPT_TEMPLATE.map { |line| line[:content].split.size }.sum
-      max_chunk_tokens = max_total_tokens - response_token_reserve - template_tokens
-      words_per_token = 0.7
-      max_chunk_word_count = max_chunk_tokens * words_per_token
+      template_word_count = LINE_SUMMARY_PROMPT_TEMPLATE.map { |line| line[:content].split.size }.sum
+      template_token_count = SummarizeMeeting::Ai.calculate_word_token_count(template_word_count)
+      max_chunk_token_count = SummarizeMeeting::Ai::MAX_TOTAL_TOKENS - RESPONSE_RESERVE_TOKENS - template_token_count
+      max_chunk_word_count = SummarizeMeeting::Ai.calculate_token_word_count(max_chunk_token_count)
 
       # Step 3. Split the transcript into equally sized chunks.
       chunks = split_lines_into_equal_size_chunks(lines, max_chunk_word_count)
@@ -86,27 +85,14 @@ module SummarizeMeeting
       consolidated_template = CONSOLIDATED_SUMMARY_PROMPT_TEMPLATE
       prompt = Mustache.render(consolidated_template.to_json, { notes: previous_chunks_summary.to_json })
       messages = JSON.parse(prompt)
-      response = SummarizeMeeting::Ai.client.chat(
-        parameters: {
-          model: "gpt-3.5-turbo",
-          messages: messages,
-        }
-      )
-      response.dig("choices", 0, "message", "content")
+      SummarizeMeeting::Ai.chat(messages)
     end
 
     def summarize_chunk(chunk, chunk_index, chunk_count, previous_chunks_summary)
       template = LINE_SUMMARY_PROMPT_TEMPLATE
       prompt = Mustache.render(template.to_json, { chunkCount: chunk_count, chunkIndex: chunk_index + 1, chunk: chunk.join("\n").to_json })
       messages = JSON.parse(prompt)
-
-      response = SummarizeMeeting::Ai.client.chat(
-        parameters: {
-          model: "gpt-3.5-turbo",
-          messages: messages,
-        }
-      )
-      response.dig("choices", 0, "message", "content")
+      SummarizeMeeting::Ai.chat(messages)
     end
 
     def split_lines_into_equal_size_chunks(lines, max_chunk_word_count)

data/lib/summarize-meeting/version.rb

@@ -1,3 +1,3 @@
 module SummarizeMeeting
-  VERSION = "1.0.0"
+  VERSION = "1.2.0"
 end

data/lib/summarize-meeting.rb

@@ -3,4 +3,4 @@ require_relative "./summarize-meeting/ai"
 require_relative "./summarize-meeting/meeting"
 
 module SummarizeMeeting
-end
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: summarize-meeting
 version: !ruby/object:Gem::Version
-  version: 1.0.0
+  version: 1.2.0
 platform: ruby
 authors:
 - Sean Devine
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2023-03-03 00:00:00.000000000 Z
+date: 2023-03-04 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: optparse