summarize-meeting 0.1.2 → 0.2.0

Sign up to get free protection for your applications and to get access to all the features.
checksums.yaml CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: 70d706fa7630923e2599b27fc3916a16523785a915a659a3b81d9ff47853b6b2
4
- data.tar.gz: 98bc7a33992281e1476702fb927f3f41f2b81a5f8b8ebd1c264d37ac236af87e
3
+ metadata.gz: 3c14fc93afabfd456a2c3778dffccf9d2a66e77f2d47ab98b864986ad0e10cd2
4
+ data.tar.gz: 68ba17378b48f84ad029ef1d3d6e1575110739bb4c8f5c40e25a759350722cba
5
5
  SHA512:
6
- metadata.gz: 31e4d53b5969ba8cc4341045f05888b68d6812ff4480d8b2d4f850533bd76f30665a4a92c29872c2731a040e0d1d0b25aa9554fe9f72046cde0d22b5a7fbb902
7
- data.tar.gz: 75035c790d2448947a98a51b2fb19f23c6e2d394095a9090bc5242009443db0b7ffd4ac23327d5fc9b24999a16fe6ad2367007be5096693106a8cf04701e41b7
6
+ metadata.gz: 40a944d8e2289d02368c1f05c98ef3b26612f40c01c0aa761c110e7f46664b8411f11c7881152b32cc4aa3d1c4db20d726906fdc00ef69c5d362779cd5300e40
7
+ data.tar.gz: 6477d716ebf27e2b6b7a8a5e15c4ce2ec205e7f089b277002fb5f1ab5273d1e486f2248625571038475fac55e91d81fa77378094615727d0bccae970251b1f74
@@ -1,8 +1,7 @@
1
1
  #!/usr/bin/env ruby
2
2
 
3
3
  require "optparse"
4
-
5
- require_relative "../lib/meeting"
4
+ require "summarize-meeting"
6
5
 
7
6
  def main
8
7
  options = {}
@@ -19,8 +18,8 @@ def main
19
18
  options[:output_file] = file
20
19
  end
21
20
 
22
- if ENV["OPENAI_KEY"]
23
- options[:openai_key] = ENV["OPENAI_KEY"]
21
+ if ENV["OPENAI_API_KEY"]
22
+ options[:openai_key] = ENV["OPENAI_API_KEY"]
24
23
  end
25
24
 
26
25
  opts.on("-k", "--openai-key KEY", "The OpenAI API key to use") do |key|
@@ -36,8 +35,8 @@ def main
36
35
  end
37
36
  end.parse!
38
37
 
39
- Ai.access_token = options[:openai_key] if options[:openai_key]
40
- Ai.organization_id = options[:openai_org] if options[:openai_org]
38
+ SummarizeMeeting::Ai.access_token = options[:openai_key] if options[:openai_key]
39
+ SummarizeMeeting::Ai.organization_id = options[:openai_org] if options[:openai_org]
41
40
 
42
41
  if ARGV.length != 1
43
42
  puts "Error: You must specify a transcript file to summarize."
@@ -47,7 +46,7 @@ def main
47
46
  transcript_file = ARGV[0]
48
47
  transcript = File.read(transcript_file)
49
48
 
50
- meeting = Meeting.new(transcript)
49
+ meeting = SummarizeMeeting::Meeting.new(transcript)
51
50
  summary = meeting.summarize
52
51
  summary_file_name = if options[:output_file]
53
52
  options[:output_file]
@@ -59,6 +58,6 @@ def main
59
58
  File.write(summary_file_name, summary)
60
59
  end
61
60
 
62
- if __FILE__ == $0
61
+ if __FILE__.to_s.end_with?("summarize-meeting") && $0.to_s.end_with?("summarize-meeting")
63
62
  main
64
63
  end
@@ -0,0 +1,28 @@
1
+ require "openai"
2
+
3
+ module SummarizeMeeting
4
+ module Ai
5
+ @@access_token = ENV["OPENAI_KEY"]
6
+ @@organization_id = ENV["OPENAI_ORG"]
7
+
8
+ def self.client
9
+ OpenAI::Client.new(access_token: access_token, organization_id: organization_id)
10
+ end
11
+
12
+ def self.access_token
13
+ @@access_token
14
+ end
15
+
16
+ def self.organization_id
17
+ @@organization_id
18
+ end
19
+
20
+ def self.access_token=(token)
21
+ @@access_token = token
22
+ end
23
+
24
+ def self.organization_id=(id)
25
+ @@organization_id = id
26
+ end
27
+ end
28
+ end
@@ -0,0 +1,130 @@
1
+ require "json"
2
+ require "mustache"
3
+ require "openai"
4
+
5
+ module SummarizeMeeting
6
+ class Meeting
7
+ LINE_SUMMARY_PROMPT_TEMPLATE = [
8
+ {
9
+ role: "system",
10
+ content: "You are an assistant summarizing a meeting.",
11
+ },
12
+ {
13
+ role: "system",
14
+ content: "The transcript of the meeting is split into {{chunkCount}} chunks. This is the {{chunkIndex}} chunk.",
15
+ },
16
+ {
17
+ role: "assistant",
18
+ content: "Please provide me with the next chunk of the transcript.",
19
+ },
20
+ {
21
+ role: "user",
22
+ content: "{{chunk}}",
23
+ }
24
+ ]
25
+
26
+ CONSOLIDATED_SUMMARY_PROMPT_TEMPLATE = [
27
+ {
28
+ role: "system",
29
+ content: "You are an assistant summarizing a meeting.",
30
+ },
31
+ {
32
+ role: "system",
33
+ content: "Notes about the meeting have been compiled.",
34
+ },
35
+ {
36
+ role: "system",
37
+ content: <<~CONTENT
38
+ Your job is to write a thorough summary of the meeting.
39
+ The summary should start with a brief overview of the meeting.
40
+ The summary should be detailed and should extract any action items that were discussed.
41
+ The summary should be organized into sections with headings and bullet points.
42
+ The summary should include a list of attendees.
43
+ The order of the sections should be overview, attendees, action items, and detailed notes by topic.
44
+ CONTENT
45
+ },
46
+ {
47
+ role: "assistant",
48
+ content: "Please provide me with notes from the meeting.",
49
+ },
50
+ {
51
+ role: "user",
52
+ content: "{{notes}}",
53
+ }
54
+ ]
55
+
56
+ def initialize(transcript)
57
+ @transcript = transcript
58
+ end
59
+
60
+ attr_reader :transcript
61
+
62
+ def summarize
63
+
64
+ # Step 1. Split the transcript into lines.
65
+ lines = transcript.split("\n")
66
+
67
+ # Step 2. Calculate the maximum chunk size in words.
68
+ max_total_tokens = 4000
69
+ response_token_reserve = 500
70
+ template_tokens = LINE_SUMMARY_PROMPT_TEMPLATE.map { |line| line[:content].split.size }.sum
71
+ max_chunk_tokens = max_total_tokens - response_token_reserve - template_tokens
72
+ words_per_token = 0.7
73
+ max_chunk_word_count = max_chunk_tokens * words_per_token
74
+
75
+ # Step 3. Split the transcript into equally sized chunks.
76
+ chunks = split_lines_into_equal_size_chunks(lines, max_chunk_word_count)
77
+
78
+ # Step 4. Summarize each chunk.
79
+ previous_chunks_summary = ""
80
+ chunks.each_with_index do |chunk, chunk_index|
81
+ chunk_summary = summarize_chunk(chunk, chunk_index, chunks.size, previous_chunks_summary)
82
+ previous_chunks_summary += chunk_summary
83
+ end
84
+
85
+ # Step 5. Write a consolidated summary.
86
+ consolidated_template = CONSOLIDATED_SUMMARY_PROMPT_TEMPLATE
87
+ prompt = Mustache.render(consolidated_template.to_json, { notes: previous_chunks_summary.to_json })
88
+ messages = JSON.parse(prompt)
89
+ response = SummarizeMeeting::Ai.client.chat(
90
+ parameters: {
91
+ model: "gpt-3.5-turbo",
92
+ messages: messages,
93
+ }
94
+ )
95
+ response.dig("choices", 0, "message", "content")
96
+ end
97
+
98
+ def summarize_chunk(chunk, chunk_index, chunk_count, previous_chunks_summary)
99
+ template = LINE_SUMMARY_PROMPT_TEMPLATE
100
+ prompt = Mustache.render(template.to_json, { chunkCount: chunk_count, chunkIndex: chunk_index + 1, chunk: chunk.join("\n").to_json })
101
+ messages = JSON.parse(prompt)
102
+
103
+ response = SummarizeMeeting::Ai.client.chat(
104
+ parameters: {
105
+ model: "gpt-3.5-turbo",
106
+ messages: messages,
107
+ }
108
+ )
109
+ response.dig("choices", 0, "message", "content")
110
+ end
111
+
112
+ def split_lines_into_equal_size_chunks(lines, max_chunk_word_count)
113
+ chunks = []
114
+ chunk = []
115
+ chunk_word_count = 0
116
+ lines.each do |line|
117
+ line_word_count = line.split.size
118
+ if chunk_word_count + line_word_count > max_chunk_word_count
119
+ chunks << chunk
120
+ chunk = []
121
+ chunk_word_count = 0
122
+ end
123
+ chunk << line
124
+ chunk_word_count += line_word_count
125
+ end
126
+ chunks << chunk
127
+ chunks
128
+ end
129
+ end
130
+ end
@@ -0,0 +1,3 @@
1
+ module SummarizeMeeting
2
+ VERSION = "0.2.0"
3
+ end
@@ -0,0 +1,6 @@
1
+ require_relative "./summarize-meeting/version"
2
+ require_relative "./summarize-meeting/ai"
3
+ require_relative "./summarize-meeting/meeting"
4
+
5
+ module SummarizeMeeting
6
+ end
metadata CHANGED
@@ -1,7 +1,7 @@
1
1
  --- !ruby/object:Gem::Specification
2
2
  name: summarize-meeting
3
3
  version: !ruby/object:Gem::Version
4
- version: 0.1.2
4
+ version: 0.2.0
5
5
  platform: ruby
6
6
  authors:
7
7
  - Sean Devine
@@ -133,8 +133,10 @@ files:
133
133
  - LICENSE.txt
134
134
  - README.md
135
135
  - bin/summarize-meeting
136
- - lib/ai.rb
137
- - lib/meeting.rb
136
+ - lib/summarize-meeting.rb
137
+ - lib/summarize-meeting/ai.rb
138
+ - lib/summarize-meeting/meeting.rb
139
+ - lib/summarize-meeting/version.rb
138
140
  homepage:
139
141
  licenses:
140
142
  - MIT
@@ -142,7 +144,7 @@ metadata: {}
142
144
  post_install_message:
143
145
  rdoc_options: []
144
146
  require_paths:
145
- - "."
147
+ - lib
146
148
  required_ruby_version: !ruby/object:Gem::Requirement
147
149
  requirements:
148
150
  - - ">="
data/lib/ai.rb DELETED
@@ -1,26 +0,0 @@
1
- require "openai"
2
-
3
- module Ai
4
- @@access_token = ENV["OPENAI_KEY"]
5
- @@organization_id = ENV["OPENAI_ORG"]
6
-
7
- def self.client
8
- OpenAI::Client.new(access_token: access_token, organization_id: organization_id)
9
- end
10
-
11
- def self.access_token
12
- @@access_token
13
- end
14
-
15
- def self.organization_id
16
- @@organization_id
17
- end
18
-
19
- def self.access_token=(token)
20
- @@access_token = token
21
- end
22
-
23
- def self.organization_id=(id)
24
- @@organization_id = id
25
- end
26
- end
data/lib/meeting.rb DELETED
@@ -1,130 +0,0 @@
1
- require "json"
2
- require "mustache"
3
- require "openai"
4
-
5
- require_relative "./ai"
6
-
7
- class Meeting
8
- LINE_SUMMARY_PROMPT_TEMPLATE = [
9
- {
10
- role: "system",
11
- content: "You are an assistant summarizing a meeting.",
12
- },
13
- {
14
- role: "system",
15
- content: "The transcript of the meeting is split into {{chunkCount}} chunks. This is the {{chunkIndex}} chunk.",
16
- },
17
- {
18
- role: "assistant",
19
- content: "Please provide me with the next chunk of the transcript.",
20
- },
21
- {
22
- role: "user",
23
- content: "{{chunk}}",
24
- }
25
- ]
26
-
27
- CONSOLIDATED_SUMMARY_PROMPT_TEMPLATE = [
28
- {
29
- role: "system",
30
- content: "You are an assistant summarizing a meeting.",
31
- },
32
- {
33
- role: "system",
34
- content: "Notes about the meeting have been compiled.",
35
- },
36
- {
37
- role: "system",
38
- content: <<~CONTENT
39
- Your job is to write a thorough summary of the meeting.
40
- The summary should start with a brief overview of the meeting.
41
- The summary should be detailed and should extract any action items that were discussed.
42
- The summary should be organized into sections with headings and bullet points.
43
- The summary should include a list of attendees.
44
- The order of the sections should be overview, attendees, action items, and detailed notes by topic.
45
- CONTENT
46
- },
47
- {
48
- role: "assistant",
49
- content: "Please provide me with notes from the meeting.",
50
- },
51
- {
52
- role: "user",
53
- content: "{{notes}}",
54
- }
55
- ]
56
-
57
- def initialize(transcript)
58
- @transcript = transcript
59
- end
60
-
61
- attr_reader :transcript
62
-
63
- def summarize
64
-
65
- # Step 1. Split the transcript into lines.
66
- lines = transcript.split("\n")
67
-
68
- # Step 2. Calculate the maximum chunk size in words.
69
- max_total_tokens = 4000
70
- response_token_reserve = 500
71
- template_tokens = LINE_SUMMARY_PROMPT_TEMPLATE.map { |line| line[:content].split.size }.sum
72
- max_chunk_tokens = max_total_tokens - response_token_reserve - template_tokens
73
- words_per_token = 0.7
74
- max_chunk_word_count = max_chunk_tokens * words_per_token
75
-
76
- # Step 3. Split the transcript into equally sized chunks.
77
- chunks = split_lines_into_equal_size_chunks(lines, max_chunk_word_count)
78
-
79
- # Step 4. Summarize each chunk.
80
- previous_chunks_summary = ""
81
- chunks.each_with_index do |chunk, chunk_index|
82
- chunk_summary = summarize_chunk(chunk, chunk_index, chunks.size, previous_chunks_summary)
83
- previous_chunks_summary += chunk_summary
84
- end
85
-
86
- # Step 5. Write a consolidated summary.
87
- consolidated_template = CONSOLIDATED_SUMMARY_PROMPT_TEMPLATE
88
- prompt = Mustache.render(consolidated_template.to_json, { notes: previous_chunks_summary.to_json })
89
- messages = JSON.parse(prompt)
90
- response = Ai.client.chat(
91
- parameters: {
92
- model: "gpt-3.5-turbo",
93
- messages: messages,
94
- }
95
- )
96
- response.dig("choices", 0, "message", "content")
97
- end
98
-
99
- def summarize_chunk(chunk, chunk_index, chunk_count, previous_chunks_summary)
100
- template = LINE_SUMMARY_PROMPT_TEMPLATE
101
- prompt = Mustache.render(template.to_json, { chunkCount: chunk_count, chunkIndex: chunk_index + 1, chunk: chunk.join("\n").to_json })
102
- messages = JSON.parse(prompt)
103
-
104
- response = Ai.client.chat(
105
- parameters: {
106
- model: "gpt-3.5-turbo",
107
- messages: messages,
108
- }
109
- )
110
- response.dig("choices", 0, "message", "content")
111
- end
112
-
113
- def split_lines_into_equal_size_chunks(lines, max_chunk_word_count)
114
- chunks = []
115
- chunk = []
116
- chunk_word_count = 0
117
- lines.each do |line|
118
- line_word_count = line.split.size
119
- if chunk_word_count + line_word_count > max_chunk_word_count
120
- chunks << chunk
121
- chunk = []
122
- chunk_word_count = 0
123
- end
124
- chunk << line
125
- chunk_word_count += line_word_count
126
- end
127
- chunks << chunk
128
- chunks
129
- end
130
- end