rosett-ai 1.8.1 → 1.11.9
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/CHANGELOG.md +253 -0
- data/README.md +2 -2
- data/conf/schemas/rai_config_schema.json +10 -0
- data/lib/rosett_ai/compiler/backends/agents_md_backend.rb +10 -3
- data/lib/rosett_ai/compiler/backends/claude_backend.rb +11 -1
- data/lib/rosett_ai/compiler/compilation_pipeline.rb +25 -0
- data/lib/rosett_ai/compiler/lean_bundle.rb +150 -0
- data/lib/rosett_ai/compiler/lean_emitter.rb +76 -0
- data/lib/rosett_ai/desktop/gui_logger.rb +1 -1
- data/lib/rosett_ai/exit_codes.rb +2 -1
- data/lib/rosett_ai/init/mcp_registrar.rb +56 -19
- data/lib/rosett_ai/mcp/governance.rb +1 -0
- data/lib/rosett_ai/mcp/tools/compile_status_tool.rb +58 -13
- data/lib/rosett_ai/mcp/tools/config_compile_tool.rb +40 -7
- data/lib/rosett_ai/mcp/tools/optimize_tool.rb +73 -0
- data/lib/rosett_ai/optimize/corpus.rb +79 -0
- data/lib/rosett_ai/optimize/detector.rb +27 -0
- data/lib/rosett_ai/optimize/detectors/conflict.rb +50 -0
- data/lib/rosett_ai/optimize/detectors/overlap.rb +74 -0
- data/lib/rosett_ai/optimize/detectors/staleness.rb +61 -0
- data/lib/rosett_ai/optimize/detectors/token_cost.rb +44 -0
- data/lib/rosett_ai/optimize/finding.rb +36 -0
- data/lib/rosett_ai/optimize/proposal_writer.rb +54 -0
- data/lib/rosett_ai/optimize/report_writer.rb +72 -0
- data/lib/rosett_ai/optimize/runner.rb +38 -0
- data/lib/rosett_ai/optimize/thresholds.rb +52 -0
- data/lib/rosett_ai/optimize.rb +33 -0
- data/lib/rosett_ai/path_resolver.rb +25 -0
- data/lib/rosett_ai/plugins/registry.rb +45 -15
- data/lib/rosett_ai/provenance/entry.rb +1 -2
- data/lib/rosett_ai/rai_config.rb +3 -0
- data/lib/rosett_ai/rules/status_checker.rb +144 -0
- data/lib/rosett_ai/telemetry/json_lines_writer.rb +1 -1
- data/lib/rosett_ai/thor/cli.rb +7 -61
- data/lib/rosett_ai/thor/command_tree.rb +79 -0
- data/lib/rosett_ai/thor/tasks/compile.rb +137 -2
- data/lib/rosett_ai/thor/tasks/optimize.rb +103 -0
- data/lib/rosett_ai/thor/tasks/rules.rb +125 -0
- data/lib/rosett_ai/version.rb +1 -1
- data/lib/rosett_ai.rb +2 -0
- data/locales/en.yml +15 -0
- metadata +22 -3
|
@@ -0,0 +1,44 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
module RosettAi
|
|
7
|
+
module Optimize
|
|
8
|
+
module Detectors
|
|
9
|
+
# Flags rules whose source description exceeds
|
|
10
|
+
# {Thresholds#verbose_bytes_min} bytes. Compile-time payload
|
|
11
|
+
# leanness is a separate C2/C3 concern — this detector targets
|
|
12
|
+
# authoring-time verbosity so authors can trim before it inflates
|
|
13
|
+
# every compiled artifact.
|
|
14
|
+
class TokenCost < Detector
|
|
15
|
+
# Approximation: ~4 chars/token for typical English narrative.
|
|
16
|
+
# Not exact — a hint for the reader, never a hard budget.
|
|
17
|
+
CHARS_PER_TOKEN = 4
|
|
18
|
+
|
|
19
|
+
# @return [Array<Finding>]
|
|
20
|
+
def call
|
|
21
|
+
corpus.rules.filter_map do |rule|
|
|
22
|
+
next if rule.source_bytes <= thresholds.verbose_bytes_min
|
|
23
|
+
|
|
24
|
+
build_finding(rule)
|
|
25
|
+
end
|
|
26
|
+
end
|
|
27
|
+
|
|
28
|
+
private
|
|
29
|
+
|
|
30
|
+
def build_finding(rule)
|
|
31
|
+
approx_tokens = rule.source_bytes / CHARS_PER_TOKEN
|
|
32
|
+
Finding.new(
|
|
33
|
+
detector: :token_cost,
|
|
34
|
+
severity: rule.source_bytes >= (thresholds.verbose_bytes_min * 2) ? :high : :medium,
|
|
35
|
+
behaviour: rule.behaviour,
|
|
36
|
+
rule_id: rule.id,
|
|
37
|
+
summary: "Verbose rule body: #{rule.source_bytes} bytes (~#{approx_tokens} tokens)",
|
|
38
|
+
details: { source_bytes: rule.source_bytes, approx_tokens: approx_tokens }
|
|
39
|
+
)
|
|
40
|
+
end
|
|
41
|
+
end
|
|
42
|
+
end
|
|
43
|
+
end
|
|
44
|
+
end
|
|
@@ -0,0 +1,36 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
module RosettAi
|
|
7
|
+
module Optimize
|
|
8
|
+
# Immutable value object describing one optimization finding.
|
|
9
|
+
#
|
|
10
|
+
# @!attribute [r] detector
|
|
11
|
+
# @return [Symbol] :overlap | :staleness | :conflict | :token_cost
|
|
12
|
+
# @!attribute [r] severity
|
|
13
|
+
# @return [Symbol] :low | :medium | :high
|
|
14
|
+
# @!attribute [r] behaviour
|
|
15
|
+
# @return [String] behaviour file basename (without .yml)
|
|
16
|
+
# @!attribute [r] rule_id
|
|
17
|
+
# @return [String, nil] rule id, or nil for behaviour-level findings
|
|
18
|
+
# @!attribute [r] summary
|
|
19
|
+
# @return [String] one-line finding summary
|
|
20
|
+
# @!attribute [r] details
|
|
21
|
+
# @return [Hash] detector-specific context (peer rule, score, terms, etc.)
|
|
22
|
+
Finding = Data.define(:detector, :severity, :behaviour, :rule_id, :summary, :details) do
|
|
23
|
+
# @return [Hash] JSON-safe representation for report / MCP output
|
|
24
|
+
def to_h
|
|
25
|
+
{
|
|
26
|
+
detector: detector.to_s,
|
|
27
|
+
severity: severity.to_s,
|
|
28
|
+
behaviour: behaviour,
|
|
29
|
+
rule_id: rule_id,
|
|
30
|
+
summary: summary,
|
|
31
|
+
details: details
|
|
32
|
+
}
|
|
33
|
+
end
|
|
34
|
+
end
|
|
35
|
+
end
|
|
36
|
+
end
|
|
@@ -0,0 +1,54 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
require 'fileutils'
|
|
7
|
+
require 'yaml'
|
|
8
|
+
|
|
9
|
+
module RosettAi
|
|
10
|
+
module Optimize
|
|
11
|
+
# Writes proposal YAML patch files (one per behaviour) to a directory.
|
|
12
|
+
#
|
|
13
|
+
# Consumed manually by the operator (or by a future review UI). This
|
|
14
|
+
# writer NEVER modifies the source corpus itself — the human gate is
|
|
15
|
+
# mandatory (invisible-user-rules ship-breaker lesson).
|
|
16
|
+
class ProposalWriter
|
|
17
|
+
# @param findings [Array<Finding>]
|
|
18
|
+
def initialize(findings:)
|
|
19
|
+
@findings = findings
|
|
20
|
+
end
|
|
21
|
+
|
|
22
|
+
# Write one YAML patch file per behaviour into `dir`.
|
|
23
|
+
#
|
|
24
|
+
# @param dir [String, Pathname]
|
|
25
|
+
# @return [Array<String>] paths written
|
|
26
|
+
def write(dir)
|
|
27
|
+
FileUtils.mkdir_p(dir)
|
|
28
|
+
grouped.map do |behaviour, findings|
|
|
29
|
+
path = File.join(dir.to_s, "#{behaviour}.proposal.yml")
|
|
30
|
+
File.write(path, YAML.dump('behaviour' => behaviour, 'proposals' => proposals_for(findings)))
|
|
31
|
+
path
|
|
32
|
+
end
|
|
33
|
+
end
|
|
34
|
+
|
|
35
|
+
private
|
|
36
|
+
|
|
37
|
+
def grouped
|
|
38
|
+
@findings.group_by(&:behaviour).sort_by { |k, _| k.to_s }
|
|
39
|
+
end
|
|
40
|
+
|
|
41
|
+
def proposals_for(findings)
|
|
42
|
+
findings.map do |f|
|
|
43
|
+
{
|
|
44
|
+
'detector' => f.detector.to_s,
|
|
45
|
+
'rule_id' => f.rule_id,
|
|
46
|
+
'severity' => f.severity.to_s,
|
|
47
|
+
'summary' => f.summary,
|
|
48
|
+
'details' => f.details
|
|
49
|
+
}
|
|
50
|
+
end
|
|
51
|
+
end
|
|
52
|
+
end
|
|
53
|
+
end
|
|
54
|
+
end
|
|
@@ -0,0 +1,72 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
module RosettAi
|
|
7
|
+
module Optimize
|
|
8
|
+
# Renders {Finding} lists as Markdown reports.
|
|
9
|
+
#
|
|
10
|
+
# Report-only mode is the default output: a single Markdown file
|
|
11
|
+
# grouped by detector, with a summary count table at the top.
|
|
12
|
+
class ReportWriter
|
|
13
|
+
# @param findings [Array<Finding>]
|
|
14
|
+
# @param generated_at [String] ISO8601 timestamp for the header
|
|
15
|
+
def initialize(findings:, generated_at:)
|
|
16
|
+
@findings = findings
|
|
17
|
+
@generated_at = generated_at
|
|
18
|
+
end
|
|
19
|
+
|
|
20
|
+
# @return [String] full Markdown report
|
|
21
|
+
def render
|
|
22
|
+
sections = [header, summary_table]
|
|
23
|
+
grouped.each { |detector, items| sections << detector_section(detector, items) }
|
|
24
|
+
"#{sections.join("\n\n").rstrip}\n"
|
|
25
|
+
end
|
|
26
|
+
|
|
27
|
+
# Write the rendered report to `path` (creates parent dirs as needed).
|
|
28
|
+
#
|
|
29
|
+
# @param path [String, Pathname]
|
|
30
|
+
# @return [String] the path written
|
|
31
|
+
def write(path)
|
|
32
|
+
require 'fileutils'
|
|
33
|
+
FileUtils.mkdir_p(File.dirname(path))
|
|
34
|
+
File.write(path, render)
|
|
35
|
+
path.to_s
|
|
36
|
+
end
|
|
37
|
+
|
|
38
|
+
private
|
|
39
|
+
|
|
40
|
+
def grouped
|
|
41
|
+
@findings.group_by(&:detector).sort_by { |k, _| k.to_s }
|
|
42
|
+
end
|
|
43
|
+
|
|
44
|
+
def header
|
|
45
|
+
"# rai_optimize report\n\nGenerated: #{@generated_at}\n\nTotal findings: #{@findings.size}"
|
|
46
|
+
end
|
|
47
|
+
|
|
48
|
+
def summary_table
|
|
49
|
+
counts = @findings.group_by(&:detector).transform_values(&:size)
|
|
50
|
+
rows = counts.sort_by { |k, _| k.to_s }.map { |k, v| "| #{k} | #{v} |" }.join("\n")
|
|
51
|
+
<<~TABLE.strip
|
|
52
|
+
## Summary
|
|
53
|
+
|
|
54
|
+
| Detector | Count |
|
|
55
|
+
|---|---|
|
|
56
|
+
#{rows.empty? ? '| _(none)_ | 0 |' : rows}
|
|
57
|
+
TABLE
|
|
58
|
+
end
|
|
59
|
+
|
|
60
|
+
def detector_section(detector, items)
|
|
61
|
+
rows = items.map { |f| "| #{f.severity} | #{f.behaviour} | #{f.rule_id} | #{f.summary} |" }
|
|
62
|
+
<<~SECTION.strip
|
|
63
|
+
## #{detector}
|
|
64
|
+
|
|
65
|
+
| Severity | Behaviour | Rule | Summary |
|
|
66
|
+
|---|---|---|---|
|
|
67
|
+
#{rows.join("\n")}
|
|
68
|
+
SECTION
|
|
69
|
+
end
|
|
70
|
+
end
|
|
71
|
+
end
|
|
72
|
+
end
|
|
@@ -0,0 +1,38 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
module RosettAi
|
|
7
|
+
module Optimize
|
|
8
|
+
# Orchestrates the four mechanical detectors and returns a flat
|
|
9
|
+
# Array<{Finding}> for the caller (report writer, proposal writer,
|
|
10
|
+
# CLI, MCP tool).
|
|
11
|
+
class Runner
|
|
12
|
+
DETECTORS = [
|
|
13
|
+
Detectors::Overlap,
|
|
14
|
+
Detectors::Staleness,
|
|
15
|
+
Detectors::Conflict,
|
|
16
|
+
Detectors::TokenCost
|
|
17
|
+
].freeze
|
|
18
|
+
|
|
19
|
+
attr_reader :corpus, :thresholds
|
|
20
|
+
|
|
21
|
+
# @param corpus [Corpus]
|
|
22
|
+
# @param thresholds [Thresholds]
|
|
23
|
+
# @param detectors [Array<Class>] override for tests
|
|
24
|
+
def initialize(corpus:, thresholds:, detectors: DETECTORS)
|
|
25
|
+
@corpus = corpus
|
|
26
|
+
@thresholds = thresholds
|
|
27
|
+
@detectors = detectors
|
|
28
|
+
end
|
|
29
|
+
|
|
30
|
+
# @return [Array<Finding>]
|
|
31
|
+
def run
|
|
32
|
+
@detectors.flat_map do |klass|
|
|
33
|
+
klass.new(corpus: corpus, thresholds: thresholds).call
|
|
34
|
+
end
|
|
35
|
+
end
|
|
36
|
+
end
|
|
37
|
+
end
|
|
38
|
+
end
|
|
@@ -0,0 +1,52 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
require 'yaml'
|
|
7
|
+
|
|
8
|
+
module RosettAi
|
|
9
|
+
module Optimize
|
|
10
|
+
# Detector thresholds — loaded from conf/rai-optimize/thresholds.yml or
|
|
11
|
+
# constructed with defaults matching the ratified design (Jaccard 0.4,
|
|
12
|
+
# inflated priority ≥ 80, verbose byte length > 3000).
|
|
13
|
+
class Thresholds
|
|
14
|
+
DEFAULT_JACCARD_MIN = 0.4
|
|
15
|
+
DEFAULT_INFLATED_PRIORITY_MIN = 80
|
|
16
|
+
DEFAULT_VERBOSE_BYTES_MIN = 3000
|
|
17
|
+
|
|
18
|
+
attr_reader :jaccard_min, :inflated_priority_min, :verbose_bytes_min
|
|
19
|
+
|
|
20
|
+
# @param jaccard_min [Float]
|
|
21
|
+
# @param inflated_priority_min [Integer]
|
|
22
|
+
# @param verbose_bytes_min [Integer]
|
|
23
|
+
def initialize(jaccard_min:, inflated_priority_min:, verbose_bytes_min:)
|
|
24
|
+
@jaccard_min = jaccard_min
|
|
25
|
+
@inflated_priority_min = inflated_priority_min
|
|
26
|
+
@verbose_bytes_min = verbose_bytes_min
|
|
27
|
+
end
|
|
28
|
+
|
|
29
|
+
# @return [Thresholds] instance with ratified design defaults
|
|
30
|
+
def self.defaults
|
|
31
|
+
new(
|
|
32
|
+
jaccard_min: DEFAULT_JACCARD_MIN,
|
|
33
|
+
inflated_priority_min: DEFAULT_INFLATED_PRIORITY_MIN,
|
|
34
|
+
verbose_bytes_min: DEFAULT_VERBOSE_BYTES_MIN
|
|
35
|
+
)
|
|
36
|
+
end
|
|
37
|
+
|
|
38
|
+
# Load thresholds from a YAML file; missing keys fall back to defaults.
|
|
39
|
+
#
|
|
40
|
+
# @param path [String, Pathname]
|
|
41
|
+
# @return [Thresholds]
|
|
42
|
+
def self.load(path)
|
|
43
|
+
raw = YAML.safe_load_file(path.to_s, permitted_classes: [Symbol]) || {}
|
|
44
|
+
new(
|
|
45
|
+
jaccard_min: raw.dig('overlap', 'jaccard_min') || DEFAULT_JACCARD_MIN,
|
|
46
|
+
inflated_priority_min: raw.dig('conflict', 'inflated_priority_min') || DEFAULT_INFLATED_PRIORITY_MIN,
|
|
47
|
+
verbose_bytes_min: raw.dig('token_cost', 'verbose_bytes_min') || DEFAULT_VERBOSE_BYTES_MIN
|
|
48
|
+
)
|
|
49
|
+
end
|
|
50
|
+
end
|
|
51
|
+
end
|
|
52
|
+
end
|
|
@@ -0,0 +1,33 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
module RosettAi
|
|
7
|
+
# Mechanical self-optimizer for the rai behaviour corpus.
|
|
8
|
+
#
|
|
9
|
+
# Sibling to {RosettAi::Adopter} (LLM-based analysis) — this module is
|
|
10
|
+
# mechanical only: token overlap, priority-inflation heuristics, staleness
|
|
11
|
+
# vocabulary matches, and byte-cost. No network calls, no LLM.
|
|
12
|
+
#
|
|
13
|
+
# Report-only by default; `--propose` writes YAML patch files to $TMPDIR.
|
|
14
|
+
# This module NEVER auto-applies changes (invisible-user-rules ship-breaker
|
|
15
|
+
# lesson — human gate is mandatory).
|
|
16
|
+
#
|
|
17
|
+
# @see RosettAi::Optimize::Runner
|
|
18
|
+
# @author hugo
|
|
19
|
+
# @author claude
|
|
20
|
+
module Optimize
|
|
21
|
+
# Convenience entry-point: run all detectors against a corpus dir.
|
|
22
|
+
#
|
|
23
|
+
# @param behaviour_dir [String, Pathname] behaviour YAML directory
|
|
24
|
+
# @param thresholds [Thresholds, nil] thresholds (defaults if nil)
|
|
25
|
+
# @return [Array<Finding>] flat list of findings from all detectors
|
|
26
|
+
def self.analyze(behaviour_dir:, thresholds: nil)
|
|
27
|
+
Runner.new(
|
|
28
|
+
corpus: Corpus.load(behaviour_dir),
|
|
29
|
+
thresholds: thresholds || Thresholds.defaults
|
|
30
|
+
).run
|
|
31
|
+
end
|
|
32
|
+
end
|
|
33
|
+
end
|
|
@@ -28,6 +28,12 @@ module RosettAi
|
|
|
28
28
|
# @return [Pathname] ~/.claude/rules/
|
|
29
29
|
def rules_dir = global_dir.join('rules')
|
|
30
30
|
|
|
31
|
+
# Lean and role-scoped context bundle output directory. Kept separate from
|
|
32
|
+
# {#rules_dir} so a session loads either the full reference rules or one
|
|
33
|
+
# lean bundle, never both.
|
|
34
|
+
# @return [Pathname] ~/.claude/rules-lean/
|
|
35
|
+
def lean_rules_dir = global_dir.join('rules-lean')
|
|
36
|
+
|
|
31
37
|
# Project-local configuration directory.
|
|
32
38
|
# @return [Pathname] .claude/ in the user's working directory
|
|
33
39
|
def local_dir = Pathname.new(user_pwd).join('.claude')
|
|
@@ -54,6 +60,17 @@ module RosettAi
|
|
|
54
60
|
# @return [Pathname] /opt/rosett-ai/app/conf
|
|
55
61
|
def packaged_conf_dir = Pathname.new('/opt/rosett-ai/app/conf')
|
|
56
62
|
|
|
63
|
+
# Compile lockfile path (`--vendor` mode). ENV-overridable so vendor
|
|
64
|
+
# writes always target the user config tree, never a packaged
|
|
65
|
+
# install's (possibly root-owned, read-only) conf/ directory.
|
|
66
|
+
# @return [Pathname] $ROSETT_AI_COMPILE_LOCK or ~/.config/rosett-ai/conf/compile.lock.yml
|
|
67
|
+
def compile_lockfile_path
|
|
68
|
+
override = ENV.fetch('ROSETT_AI_COMPILE_LOCK', nil)
|
|
69
|
+
return Pathname.new(override) if override
|
|
70
|
+
|
|
71
|
+
rai_conf_dir.join('compile.lock.yml')
|
|
72
|
+
end
|
|
73
|
+
|
|
57
74
|
# Encrypted secrets store.
|
|
58
75
|
# @return [Pathname] secrets.yml inside the rosett-ai config directory
|
|
59
76
|
def secrets_file = rai_config_dir.join('secrets.yml')
|
|
@@ -70,6 +87,10 @@ module RosettAi
|
|
|
70
87
|
# @return [Pathname] cache/content/ inside the rosett-ai config directory
|
|
71
88
|
def content_cache_dir = rai_config_dir.join('cache', 'content')
|
|
72
89
|
|
|
90
|
+
# Generated report output directory (e.g. rai_optimize reports).
|
|
91
|
+
# @return [Pathname] reports/ inside the rosett-ai config directory
|
|
92
|
+
def reports_dir = rai_config_dir.join('reports')
|
|
93
|
+
|
|
73
94
|
# Engine settings target paths
|
|
74
95
|
|
|
75
96
|
# System-wide managed settings (enterprise).
|
|
@@ -98,6 +119,10 @@ module RosettAi
|
|
|
98
119
|
# @return [String] ~/.claude/rules
|
|
99
120
|
def rules_display_path = '~/.claude/rules'
|
|
100
121
|
|
|
122
|
+
# Tilde-form display path for lean bundle output.
|
|
123
|
+
# @return [String] ~/.claude/rules-lean
|
|
124
|
+
def lean_rules_display_path = '~/.claude/rules-lean'
|
|
125
|
+
|
|
101
126
|
# The user's original working directory.
|
|
102
127
|
# When running from the .deb wrapper, Dir.pwd is /opt/rosett-ai/app, so
|
|
103
128
|
# RAI_ORIGINAL_PWD (set by the wrapper) preserves the real directory.
|
|
@@ -120,20 +120,40 @@ module RosettAi
|
|
|
120
120
|
|
|
121
121
|
# Scan `/opt/rosett-ai/embedded/lib/ruby/gems/*/gems/*/lib/` for plugin
|
|
122
122
|
# register.rb files. Handles Ruby ABI version mismatches between
|
|
123
|
-
# the embedded Ruby and engine gems built with a different minor version
|
|
123
|
+
# the embedded Ruby and engine gems built with a different minor version
|
|
124
|
+
# — both halves of the SAME `/opt/rosett-ai` installation.
|
|
125
|
+
#
|
|
126
|
+
# Gated by {embedded_install_active?} so that this never reaches into an
|
|
127
|
+
# unrelated `/opt/rosett-ai` installation found on the same host: a
|
|
128
|
+
# source checkout or CI run uses its own Ruby and its own gem home,
|
|
129
|
+
# never the embedded one, so it must never self-discover (and
|
|
130
|
+
# double-load) that installation's plugin gems.
|
|
124
131
|
#
|
|
125
132
|
# @param prefix [String] plugin namespace prefix (e.g. 'rosett_ai_engine')
|
|
126
133
|
def discover_embedded_gems(prefix)
|
|
134
|
+
return unless embedded_install_active?
|
|
135
|
+
|
|
127
136
|
embedded_gems = Pathname.new('/opt/rosett-ai/embedded/lib/ruby/gems')
|
|
128
137
|
return unless embedded_gems.directory?
|
|
129
138
|
|
|
130
|
-
|
|
131
|
-
|
|
132
|
-
|
|
133
|
-
|
|
134
|
-
|
|
135
|
-
|
|
136
|
-
|
|
139
|
+
pattern = embedded_gems.join('*', 'gems', '*', 'lib', prefix, '*', 'register.rb').to_s
|
|
140
|
+
Dir.glob(pattern).each { |path| require_plugin_file(path, source: 'embedded') }
|
|
141
|
+
end
|
|
142
|
+
|
|
143
|
+
# Is the Ruby process running *this code* itself the hermetic Ruby
|
|
144
|
+
# shipped inside the `/opt/rosett-ai` `.deb` install?
|
|
145
|
+
#
|
|
146
|
+
# The packaged wrapper (`packaging/wrapper.sh.template`) exports
|
|
147
|
+
# `GEM_HOME` to the embedded gem tree before exec-ing the embedded
|
|
148
|
+
# Ruby, so `Gem.dir` is the one reliable runtime signal: it reflects
|
|
149
|
+
# the active bundle actually in effect, not a hardcoded guess. (The
|
|
150
|
+
# embedded Ruby's own compiled-in `RbConfig::CONFIG['prefix']` is
|
|
151
|
+
# NOT usable here — ruby-build bakes in the packaging *staging* path,
|
|
152
|
+
# which does not exist at runtime; see `packaging/scripts/postinst`.)
|
|
153
|
+
#
|
|
154
|
+
# @return [Boolean]
|
|
155
|
+
def embedded_install_active?
|
|
156
|
+
Gem.dir.start_with?('/opt/rosett-ai/embedded')
|
|
137
157
|
end
|
|
138
158
|
|
|
139
159
|
# Scan directories listed in `RAI_ENGINE_PATH` (colon-separated)
|
|
@@ -145,15 +165,25 @@ module RosettAi
|
|
|
145
165
|
return unless custom_path
|
|
146
166
|
|
|
147
167
|
custom_path.split(':').each do |dir|
|
|
148
|
-
|
|
149
|
-
|
|
150
|
-
$LOAD_PATH.unshift(lib_dir) unless $LOAD_PATH.include?(lib_dir)
|
|
151
|
-
require path
|
|
152
|
-
rescue LoadError => e
|
|
153
|
-
RosettAi.logger.warn("Failed to load custom plugin from #{path}: #{e.message}")
|
|
154
|
-
end
|
|
168
|
+
pattern = File.join(dir, 'lib', prefix, '*', 'register.rb')
|
|
169
|
+
Dir.glob(pattern).each { |path| require_plugin_file(path, source: 'custom') }
|
|
155
170
|
end
|
|
156
171
|
end
|
|
172
|
+
|
|
173
|
+
# Require a discovered plugin `register.rb`, prepending its lib dir to
|
|
174
|
+
# `$LOAD_PATH` first. Shared by the embedded-gem-tree and custom-path
|
|
175
|
+
# discovery strategies; logs (rather than raises) on LoadError so one
|
|
176
|
+
# broken plugin never blocks discovery of the others.
|
|
177
|
+
#
|
|
178
|
+
# @param path [String] absolute path to a discovered register.rb
|
|
179
|
+
# @param source [String] discovery strategy name, for the warning message
|
|
180
|
+
def require_plugin_file(path, source:)
|
|
181
|
+
lib_dir = File.expand_path('../../..', path)
|
|
182
|
+
$LOAD_PATH.unshift(lib_dir) unless $LOAD_PATH.include?(lib_dir)
|
|
183
|
+
require path
|
|
184
|
+
rescue LoadError => e
|
|
185
|
+
RosettAi.logger.warn("Failed to load #{source} plugin from #{path}: #{e.message}")
|
|
186
|
+
end
|
|
157
187
|
end
|
|
158
188
|
end
|
|
159
189
|
end
|
|
@@ -27,7 +27,7 @@ module RosettAi
|
|
|
27
27
|
# @param ai_role [String] one of {ALLOWED_ROLES}
|
|
28
28
|
# @param files [Array<Hash>] file-level source references
|
|
29
29
|
# @param timestamp [String, nil] ISO 8601 timestamp (defaults to now)
|
|
30
|
-
# rubocop:disable Metrics/ParameterLists -- provenance entry requires all fields
|
|
30
|
+
# rubocop:disable-next Metrics/ParameterLists -- provenance entry requires all fields
|
|
31
31
|
def initialize(commit:, contributor:, ai_tool:, ai_role:, files: [], timestamp: nil)
|
|
32
32
|
validate_role!(ai_role)
|
|
33
33
|
|
|
@@ -38,7 +38,6 @@ module RosettAi
|
|
|
38
38
|
@files = files.freeze
|
|
39
39
|
@timestamp = (timestamp || Time.now.utc.strftime('%Y-%m-%dT%H:%M:%SZ')).freeze
|
|
40
40
|
end
|
|
41
|
-
# rubocop:enable Metrics/ParameterLists
|
|
42
41
|
|
|
43
42
|
# @return [Hash] serializable representation for YAML output
|
|
44
43
|
def to_h
|
data/lib/rosett_ai/rai_config.rb
CHANGED
|
@@ -35,6 +35,9 @@ module RosettAi
|
|
|
35
35
|
def cache_ttl_hours = @data.dig('cache', 'ttl_hours')
|
|
36
36
|
def detect_on_init? = @data.dig('detect', 'on_init')
|
|
37
37
|
def dbus_allow_plugin_management? = @data.dig('dbus', 'allow_plugin_management') == true
|
|
38
|
+
# Path to the git-tracked mirror of the LIVE behaviour directory.
|
|
39
|
+
# @return [String, nil]
|
|
40
|
+
def rules_mirror_dir = @data.dig('rules', 'mirror_dir')
|
|
38
41
|
|
|
39
42
|
# Load the license key from persistent storage.
|
|
40
43
|
#
|
|
@@ -0,0 +1,144 @@
|
|
|
1
|
+
# frozen_string_literal: true
|
|
2
|
+
|
|
3
|
+
# SPDX-License-Identifier: GPL-3.0-only
|
|
4
|
+
# Copyright (C) 2026 Hugo Antonio Sepulveda Manriquez / NeatNerds
|
|
5
|
+
|
|
6
|
+
require 'digest'
|
|
7
|
+
require 'open3'
|
|
8
|
+
|
|
9
|
+
module RosettAi
|
|
10
|
+
# Drift detection between the LIVE behaviour-rule source tree and its
|
|
11
|
+
# git-tracked MIRROR copy.
|
|
12
|
+
module Rules
|
|
13
|
+
# Compares every behaviour-rule YAML file in a LIVE directory (not
|
|
14
|
+
# git-tracked; the compile source) against its counterpart in a git-
|
|
15
|
+
# tracked MIRROR directory, by content hash.
|
|
16
|
+
#
|
|
17
|
+
# For files present on both sides with differing content, direction is
|
|
18
|
+
# derived from the MIRROR's own git history: if LIVE's current content
|
|
19
|
+
# matches a past commit of the MIRROR file, the MIRROR has moved on
|
|
20
|
+
# since LIVE was last in sync (:mirror_newer); otherwise LIVE holds
|
|
21
|
+
# content the MIRROR's history never captured (:live_newer). This
|
|
22
|
+
# reflects which side has moved, not which side is correct — a human
|
|
23
|
+
# must still inspect the diff before acting on it.
|
|
24
|
+
#
|
|
25
|
+
# Read-only: never writes to either tree, never touches the network.
|
|
26
|
+
#
|
|
27
|
+
# @author claude
|
|
28
|
+
class StatusChecker
|
|
29
|
+
# @return [Array<Symbol>] all possible per-file drift states.
|
|
30
|
+
STATES = [:identical, :live_only, :mirror_only, :live_newer, :mirror_newer].freeze
|
|
31
|
+
|
|
32
|
+
# Git repo-location env vars that git exports into hook subprocesses
|
|
33
|
+
# (pre-commit, pre-push, etc.) so the hook script can find the repo
|
|
34
|
+
# that invoked it. If inherited here, GIT_DIR/GIT_WORK_TREE override
|
|
35
|
+
# the `-C` flag below for some git subcommands (notably `log`'s
|
|
36
|
+
# pathspec resolution) and redirect this class's queries to whichever
|
|
37
|
+
# repo is running the hook, instead of the MIRROR repo it is asked to
|
|
38
|
+
# inspect. Clearing them keeps every subprocess call hermetic to
|
|
39
|
+
# @mirror_dir, regardless of what invoked this process.
|
|
40
|
+
GIT_ENV_OVERRIDE = {
|
|
41
|
+
'GIT_DIR' => nil,
|
|
42
|
+
'GIT_WORK_TREE' => nil,
|
|
43
|
+
'GIT_INDEX_FILE' => nil,
|
|
44
|
+
'GIT_COMMON_DIR' => nil,
|
|
45
|
+
'GIT_OBJECT_DIRECTORY' => nil,
|
|
46
|
+
'GIT_ALTERNATE_OBJECT_DIRECTORIES' => nil
|
|
47
|
+
}.freeze
|
|
48
|
+
|
|
49
|
+
# @param live_dir [String, Pathname] the LIVE behaviour-rule directory
|
|
50
|
+
# @param mirror_dir [String, Pathname] the git-tracked MIRROR directory
|
|
51
|
+
def initialize(live_dir:, mirror_dir:)
|
|
52
|
+
@live_dir = Pathname.new(live_dir)
|
|
53
|
+
@mirror_dir = Pathname.new(mirror_dir)
|
|
54
|
+
end
|
|
55
|
+
|
|
56
|
+
# Classify every file present in either tree.
|
|
57
|
+
#
|
|
58
|
+
# @return [Array<Hash>] one entry per file: { name:, state: }, sorted by name
|
|
59
|
+
def check
|
|
60
|
+
all_names.map { |name| { name: name, state: classify(name) } }
|
|
61
|
+
end
|
|
62
|
+
|
|
63
|
+
# @return [Boolean] true if any file is in a non-identical state
|
|
64
|
+
def drift?
|
|
65
|
+
check.any? { |entry| entry[:state] != :identical }
|
|
66
|
+
end
|
|
67
|
+
|
|
68
|
+
private
|
|
69
|
+
|
|
70
|
+
def all_names
|
|
71
|
+
(yml_names_in(@live_dir) + yml_names_in(@mirror_dir)).uniq.sort
|
|
72
|
+
end
|
|
73
|
+
|
|
74
|
+
def yml_names_in(dir)
|
|
75
|
+
return [] unless dir.directory?
|
|
76
|
+
|
|
77
|
+
Dir.children(dir).select { |f| f.end_with?('.yml') }
|
|
78
|
+
end
|
|
79
|
+
|
|
80
|
+
def classify(name)
|
|
81
|
+
live_path = @live_dir.join(name)
|
|
82
|
+
mirror_path = @mirror_dir.join(name)
|
|
83
|
+
live_exists = live_path.exist?
|
|
84
|
+
mirror_exists = mirror_path.exist?
|
|
85
|
+
|
|
86
|
+
return :live_only if live_exists && !mirror_exists
|
|
87
|
+
return :mirror_only if mirror_exists && !live_exists
|
|
88
|
+
|
|
89
|
+
classify_both_exist(live_path, mirror_path, name)
|
|
90
|
+
end
|
|
91
|
+
|
|
92
|
+
def classify_both_exist(live_path, mirror_path, name)
|
|
93
|
+
live_hash = content_hash(live_path)
|
|
94
|
+
return :identical if live_hash == content_hash(mirror_path)
|
|
95
|
+
|
|
96
|
+
mirror_history_hashes(name).include?(live_hash) ? :mirror_newer : :live_newer
|
|
97
|
+
end
|
|
98
|
+
|
|
99
|
+
# SHA256 blob hashes of every historical commit of the MIRROR file,
|
|
100
|
+
# discovered via the MIRROR directory's own git repository.
|
|
101
|
+
def mirror_history_hashes(name)
|
|
102
|
+
repo_root = mirror_repo_root
|
|
103
|
+
return [] unless repo_root
|
|
104
|
+
|
|
105
|
+
relative_path = @mirror_dir.relative_path_from(repo_root).join(name).to_s
|
|
106
|
+
mirror_log_commits(repo_root, relative_path).filter_map do |sha|
|
|
107
|
+
mirror_show_hash(repo_root, sha, relative_path)
|
|
108
|
+
end
|
|
109
|
+
end
|
|
110
|
+
|
|
111
|
+
def mirror_repo_root
|
|
112
|
+
out, status = Open3.capture2(GIT_ENV_OVERRIDE, 'git', '-C', @mirror_dir.to_s, 'rev-parse', '--show-toplevel')
|
|
113
|
+
return nil unless status.success?
|
|
114
|
+
|
|
115
|
+
Pathname.new(out.strip)
|
|
116
|
+
rescue Errno::ENOENT
|
|
117
|
+
# git binary not available — treat as no history rather than fail
|
|
118
|
+
nil
|
|
119
|
+
end
|
|
120
|
+
|
|
121
|
+
def mirror_log_commits(repo_root, relative_path)
|
|
122
|
+
out, status = Open3.capture2(
|
|
123
|
+
GIT_ENV_OVERRIDE, 'git', '-C', repo_root.to_s, 'log', '--format=%H', '--', relative_path
|
|
124
|
+
)
|
|
125
|
+
return [] unless status.success?
|
|
126
|
+
|
|
127
|
+
out.lines(chomp: true)
|
|
128
|
+
end
|
|
129
|
+
|
|
130
|
+
def mirror_show_hash(repo_root, sha, relative_path)
|
|
131
|
+
content, status = Open3.capture2(
|
|
132
|
+
GIT_ENV_OVERRIDE, 'git', '-C', repo_root.to_s, 'show', "#{sha}:#{relative_path}"
|
|
133
|
+
)
|
|
134
|
+
return nil unless status.success?
|
|
135
|
+
|
|
136
|
+
Digest::SHA256.hexdigest(content)
|
|
137
|
+
end
|
|
138
|
+
|
|
139
|
+
def content_hash(path)
|
|
140
|
+
Digest::SHA256.file(path).hexdigest
|
|
141
|
+
end
|
|
142
|
+
end
|
|
143
|
+
end
|
|
144
|
+
end
|