completion-kit 0.27.3 → 0.27.5
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/app/jobs/completion_kit/generate_row_job.rb +23 -8
- data/app/jobs/completion_kit/judge_review_job.rb +32 -16
- data/app/models/completion_kit/metric_version.rb +2 -0
- data/app/services/completion_kit/judge_service.rb +1 -1
- data/app/views/completion_kit/responses/show.html.erb +2 -0
- data/db/migrate/20260715000001_dedupe_and_unique_index_metric_versions.rb +56 -0
- data/lib/completion_kit/errors.rb +18 -0
- data/lib/completion_kit/version.rb +1 -1
- metadata +2 -1
checksums.yaml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
---
|
|
2
2
|
SHA256:
|
|
3
|
-
metadata.gz:
|
|
4
|
-
data.tar.gz:
|
|
3
|
+
metadata.gz: f36bc9d9d0e9aa1e0e6ed29dec74bef0021e112c1c2198f737313ccabd02c827
|
|
4
|
+
data.tar.gz: a3dd4f7b50e7bf96389b390fa985954fed6644b0cd33d1eda1edfa92613b6a04
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: 74ca0f3de22ea9af05a5afc86771bad4e12e018d2342d76110fc699ad6630345cd7664c2e2169b98bf2c4120cfbd96369b98be42c425ea6206344e2348052819
|
|
7
|
+
data.tar.gz: 42ce6279ed741e065dbeb4e51dba9342f08e7059690852787fbd83a1e0b279435b867d92825f4714e9594c39e5fe117d92d1445a4d651db7f72896d86965501c
|
|
@@ -12,20 +12,30 @@ module CompletionKit
|
|
|
12
12
|
30 * executions
|
|
13
13
|
end
|
|
14
14
|
|
|
15
|
+
rescue_from(StandardError) do |error|
|
|
16
|
+
Rails.error.report(error, handled: true, context: { job: self.class.name, run_id: @run_id, response_id: @response_id })
|
|
17
|
+
record_terminal_failure!(error)
|
|
18
|
+
enqueue_completion_check
|
|
19
|
+
end
|
|
20
|
+
|
|
15
21
|
retry_on Faraday::TimeoutError,
|
|
16
22
|
Faraday::ConnectionFailed,
|
|
17
|
-
wait: :polynomially_longer, attempts: 5
|
|
23
|
+
wait: :polynomially_longer, attempts: 5 do |job, error|
|
|
24
|
+
job.send(:record_terminal_failure!, error)
|
|
25
|
+
job.send(:enqueue_completion_check)
|
|
26
|
+
end
|
|
18
27
|
|
|
19
28
|
retry_on CompletionKit::RateLimitError,
|
|
20
|
-
wait:
|
|
29
|
+
wait: ->(executions) { rate_limit_wait(executions) }, attempts: 5 do |job, error|
|
|
30
|
+
job.send(:record_terminal_failure!, error)
|
|
31
|
+
job.send(:enqueue_completion_check)
|
|
32
|
+
end
|
|
21
33
|
|
|
22
34
|
discard_on ActiveJob::DeserializationError
|
|
23
|
-
discard_on CompletionKit::ConfigurationError
|
|
24
35
|
|
|
25
|
-
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
enqueue_completion_check
|
|
36
|
+
discard_on CompletionKit::ConfigurationError do |job, error|
|
|
37
|
+
job.send(:record_terminal_failure!, error)
|
|
38
|
+
job.send(:enqueue_completion_check)
|
|
29
39
|
end
|
|
30
40
|
|
|
31
41
|
before_perform do |job|
|
|
@@ -49,7 +59,12 @@ module CompletionKit
|
|
|
49
59
|
raise ConfigurationError, client.configuration_errors.join(", ") unless client.configured?
|
|
50
60
|
|
|
51
61
|
text = client.generate_completion(rendered, model: prompt.llm_model, temperature: run.temperature)
|
|
52
|
-
|
|
62
|
+
|
|
63
|
+
if text.to_s.start_with?("Error:")
|
|
64
|
+
record_terminal_failure!(ProviderError.from_client_error(text))
|
|
65
|
+
enqueue_completion_check
|
|
66
|
+
return
|
|
67
|
+
end
|
|
53
68
|
|
|
54
69
|
if client.respond_to?(:temperature_dropped?) && client.temperature_dropped? && !run.temperature_ignored?
|
|
55
70
|
run.update_columns(temperature_ignored: true)
|
|
@@ -14,20 +14,30 @@ module CompletionKit
|
|
|
14
14
|
30 * executions
|
|
15
15
|
end
|
|
16
16
|
|
|
17
|
+
rescue_from(StandardError) do |error|
|
|
18
|
+
Rails.error.report(error, handled: true, context: { job: self.class.name, run_id: @run_id, review_id: @review_id })
|
|
19
|
+
record_terminal_failure!(error)
|
|
20
|
+
enqueue_completion_check
|
|
21
|
+
end
|
|
22
|
+
|
|
17
23
|
retry_on Faraday::TimeoutError,
|
|
18
24
|
Faraday::ConnectionFailed,
|
|
19
|
-
wait: :polynomially_longer, attempts: 5
|
|
25
|
+
wait: :polynomially_longer, attempts: 5 do |job, error|
|
|
26
|
+
job.send(:record_terminal_failure!, error)
|
|
27
|
+
job.send(:enqueue_completion_check)
|
|
28
|
+
end
|
|
20
29
|
|
|
21
30
|
retry_on CompletionKit::RateLimitError,
|
|
22
|
-
wait:
|
|
31
|
+
wait: ->(executions) { rate_limit_wait(executions) }, attempts: 5 do |job, error|
|
|
32
|
+
job.send(:record_terminal_failure!, error)
|
|
33
|
+
job.send(:enqueue_completion_check)
|
|
34
|
+
end
|
|
23
35
|
|
|
24
36
|
discard_on ActiveJob::DeserializationError
|
|
25
|
-
discard_on CompletionKit::ConfigurationError
|
|
26
37
|
|
|
27
|
-
|
|
28
|
-
|
|
29
|
-
|
|
30
|
-
enqueue_completion_check
|
|
38
|
+
discard_on CompletionKit::ConfigurationError do |job, error|
|
|
39
|
+
job.send(:record_terminal_failure!, error)
|
|
40
|
+
job.send(:enqueue_completion_check)
|
|
31
41
|
end
|
|
32
42
|
|
|
33
43
|
before_perform do |job|
|
|
@@ -52,15 +62,21 @@ module CompletionKit
|
|
|
52
62
|
config = ApiConfig.for_model(run.judge_model).merge(judge_model: run.judge_model)
|
|
53
63
|
judge = JudgeService.new(config)
|
|
54
64
|
|
|
55
|
-
|
|
56
|
-
|
|
57
|
-
|
|
58
|
-
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
|
|
62
|
-
|
|
63
|
-
|
|
65
|
+
begin
|
|
66
|
+
evaluation = judge.evaluate(
|
|
67
|
+
response.response_text,
|
|
68
|
+
response.expected_output,
|
|
69
|
+
run.prompt&.template,
|
|
70
|
+
criteria: metric.instruction.to_s,
|
|
71
|
+
rubric_text: metric.display_rubric_text,
|
|
72
|
+
input_data: response.input_data,
|
|
73
|
+
human_examples: review_examples_for(metric, response)
|
|
74
|
+
)
|
|
75
|
+
rescue CompletionKit::ProviderError => e
|
|
76
|
+
record_terminal_failure!(e)
|
|
77
|
+
enqueue_completion_check
|
|
78
|
+
return
|
|
79
|
+
end
|
|
64
80
|
|
|
65
81
|
review = response.reviews.find_or_initialize_by(metric_id: metric.id)
|
|
66
82
|
current_metric_version = MetricVersion.ensure_current_for(metric)
|
|
@@ -20,7 +20,7 @@ module CompletionKit
|
|
|
20
20
|
human_examples: human_examples)
|
|
21
21
|
|
|
22
22
|
response = @judge_client.generate_completion(judge_prompt, model: @judge_model)
|
|
23
|
-
raise
|
|
23
|
+
raise CompletionKit::ProviderError.from_client_error(response) if response.start_with?("Error:")
|
|
24
24
|
parse_judge_response(response)
|
|
25
25
|
end
|
|
26
26
|
|
|
@@ -125,6 +125,8 @@
|
|
|
125
125
|
<% end %>
|
|
126
126
|
<% elsif review.check? && !review.passed.nil? %>
|
|
127
127
|
<%= ck_check_badge(review.passed) %>
|
|
128
|
+
<% elsif review.status == "failed" %>
|
|
129
|
+
<span class="<%= ck_badge_classes(:failed) %>" title="<%= review.error_message.presence || "Judge failed" %>">Failed</span>
|
|
128
130
|
<% else %>
|
|
129
131
|
<span class="<%= ck_badge_classes(:pending) %>">Pending</span>
|
|
130
132
|
<% end %>
|
|
@@ -0,0 +1,56 @@
|
|
|
1
|
+
class DedupeAndUniqueIndexMetricVersions < ActiveRecord::Migration[8.1]
|
|
2
|
+
class MetricVersion < ActiveRecord::Base
|
|
3
|
+
self.table_name = "completion_kit_metric_versions"
|
|
4
|
+
end
|
|
5
|
+
|
|
6
|
+
class Review < ActiveRecord::Base
|
|
7
|
+
self.table_name = "completion_kit_reviews"
|
|
8
|
+
end
|
|
9
|
+
|
|
10
|
+
class Agreement < ActiveRecord::Base
|
|
11
|
+
self.table_name = "completion_kit_agreements"
|
|
12
|
+
end
|
|
13
|
+
|
|
14
|
+
def up
|
|
15
|
+
collapse_duplicate_versions
|
|
16
|
+
collapse_extra_current
|
|
17
|
+
|
|
18
|
+
remove_index :completion_kit_metric_versions, name: "index_ck_metric_versions_on_metric_vnum", if_exists: true
|
|
19
|
+
add_index :completion_kit_metric_versions, [:metric_id, :version_number],
|
|
20
|
+
unique: true, name: "index_ck_metric_versions_on_metric_vnum"
|
|
21
|
+
end
|
|
22
|
+
|
|
23
|
+
def down
|
|
24
|
+
remove_index :completion_kit_metric_versions, name: "index_ck_metric_versions_on_metric_vnum", if_exists: true
|
|
25
|
+
add_index :completion_kit_metric_versions, [:metric_id, :version_number],
|
|
26
|
+
name: "index_ck_metric_versions_on_metric_vnum"
|
|
27
|
+
end
|
|
28
|
+
|
|
29
|
+
private
|
|
30
|
+
|
|
31
|
+
def collapse_duplicate_versions
|
|
32
|
+
MetricVersion.order(:id).pluck(:id, :metric_id, :version_number)
|
|
33
|
+
.group_by { |(_, metric_id, version_number)| [metric_id, version_number] }
|
|
34
|
+
.each_value do |group|
|
|
35
|
+
next if group.size < 2
|
|
36
|
+
|
|
37
|
+
canonical = group.first.first
|
|
38
|
+
losers = group.drop(1).map(&:first)
|
|
39
|
+
Review.where(metric_version_id: losers).update_all(metric_version_id: canonical)
|
|
40
|
+
Agreement.where(metric_version_id: losers).update_all(metric_version_id: canonical)
|
|
41
|
+
MetricVersion.where(id: losers).delete_all
|
|
42
|
+
end
|
|
43
|
+
end
|
|
44
|
+
|
|
45
|
+
def collapse_extra_current
|
|
46
|
+
MetricVersion.where(current: true)
|
|
47
|
+
.order(metric_id: :asc, version_number: :desc, id: :desc)
|
|
48
|
+
.pluck(:id, :metric_id)
|
|
49
|
+
.group_by { |(_, metric_id)| metric_id }
|
|
50
|
+
.each_value do |rows|
|
|
51
|
+
next if rows.size < 2
|
|
52
|
+
|
|
53
|
+
MetricVersion.where(id: rows.drop(1).map(&:first)).update_all(current: false)
|
|
54
|
+
end
|
|
55
|
+
end
|
|
56
|
+
end
|
|
@@ -13,4 +13,22 @@ module CompletionKit
|
|
|
13
13
|
@retry_after = retry_after
|
|
14
14
|
end
|
|
15
15
|
end
|
|
16
|
+
|
|
17
|
+
class ProviderError < Error
|
|
18
|
+
attr_reader :status
|
|
19
|
+
|
|
20
|
+
def initialize(message = nil, status: nil)
|
|
21
|
+
super(message)
|
|
22
|
+
@status = status
|
|
23
|
+
end
|
|
24
|
+
|
|
25
|
+
def self.from_client_error(text)
|
|
26
|
+
detail = text.to_s.sub(/\AError:\s*/, "")
|
|
27
|
+
if (match = detail.match(/\A(\d{3})\s*-\s*(.*)/m))
|
|
28
|
+
new(match[2], status: match[1].to_i)
|
|
29
|
+
else
|
|
30
|
+
new(detail)
|
|
31
|
+
end
|
|
32
|
+
end
|
|
33
|
+
end
|
|
16
34
|
end
|
metadata
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
|
2
2
|
name: completion-kit
|
|
3
3
|
version: !ruby/object:Gem::Version
|
|
4
|
-
version: 0.27.
|
|
4
|
+
version: 0.27.5
|
|
5
5
|
platform: ruby
|
|
6
6
|
authors:
|
|
7
7
|
- Damien Bastin
|
|
@@ -480,6 +480,7 @@ files:
|
|
|
480
480
|
- db/migrate/20260706000001_add_expected_column_to_completion_kit_runs.rb
|
|
481
481
|
- db/migrate/20260708000001_add_api_version_to_completion_kit_provider_credentials.rb
|
|
482
482
|
- db/migrate/20260713000001_add_catalog_model_count_to_completion_kit_provider_credentials.rb
|
|
483
|
+
- db/migrate/20260715000001_dedupe_and_unique_index_metric_versions.rb
|
|
483
484
|
- lib/completion-kit.rb
|
|
484
485
|
- lib/completion_kit.rb
|
|
485
486
|
- lib/completion_kit/concurrency_check.rb
|