longguard 0.2.1.dev0__tar.gz → 0.2.3__tar.gz
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.
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/workflows/ci.yml +55 -24
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/workflows/docs.yml +2 -2
- longguard-0.2.3/.github/workflows/release.yml +190 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/CHANGELOG.md +39 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/PKG-INFO +39 -21
- {longguard-0.2.1.dev0 → longguard-0.2.3}/README.md +34 -18
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/compatibility.md +1 -1
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/index.md +3 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/pyproject.toml +9 -2
- longguard-0.2.3/scripts/check_version.py +203 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/scripts/prepare_merge_release.py +6 -6
- {longguard-0.2.1.dev0 → longguard-0.2.3}/scripts/release_notes.py +5 -4
- {longguard-0.2.1.dev0 → longguard-0.2.3}/scripts/verify_release.py +14 -3
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/__init__.py +1 -1
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/benchmarks/run.py +1 -1
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_benchmark.py +2 -2
- longguard-0.2.3/tests/test_check_version.py +124 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_prepare_merge_release.py +7 -6
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_report_persistence.py +4 -4
- {longguard-0.2.1.dev0 → longguard-0.2.3}/uv.lock +406 -31
- longguard-0.2.1.dev0/.github/workflows/release-on-merge.yml +0 -102
- longguard-0.2.1.dev0/.github/workflows/release.yml +0 -75
- longguard-0.2.1.dev0/tests/test_merge_release_workflow.py +0 -99
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/ISSUE_TEMPLATE/bug_report.yml +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/ISSUE_TEMPLATE/config.yml +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/ISSUE_TEMPLATE/feature_request.yml +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/PR_v0.2.0_DRAFT.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/PULL_REQUEST_TEMPLATE.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/dependabot.yml +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.github/release.yml +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/.gitignore +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/CODE_OF_CONDUCT.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/CONTRIBUTING.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/LICENSE +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/SECURITY.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/SUPPORT.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/LongGuard-Logo.png +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/architecture.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/icon.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/linkedin_card.png +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/linkedin_launch.jpg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/logo-dark.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/logo-light.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/assets/logo.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/assets/LongGuard-Logo.png +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/assets/architecture.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/assets/icon.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/assets/logo-dark.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/assets/logo-light.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/assets/logo.svg +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/changelog.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/concepts/detectors.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/concepts/how-it-works.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/concepts/reflect-and-pivot.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/contributing.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/getting-started.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/benchmarks.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/configuration.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/decorator.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/development-evidence.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/model-pricing-sources.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/guides/reporting.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/integrations/crewai.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/integrations/langchain.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/integrations/langgraph.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/docs/integrations/raw-client.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/examples/crewai_guarded.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/examples/custom_detector.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/examples/langchain_agent.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/examples/langgraph_basic.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/examples/langgraph_resume.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/examples/openai_raw_guard.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/mkdocs.yml +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/config.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/__init__.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/breaker.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/decorator.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/detectors/__init__.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/detectors/base.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/detectors/dead_end.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/detectors/semantic_osc.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/detectors/token_velocity.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/detectors/tool_repeat.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/pivot.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/reporter.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/core/step.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/__init__.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/crewai.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/graph_runner.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/langchain.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/langgraph.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/messages.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/integrations/strands.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/pricing.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/src/longguard/py.typed +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/__init__.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/benchmarks/LABEL_REVIEW.md +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/benchmarks/corpus.json +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/benchmarks/results/baseline-hash.json +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/benchmarks/results/current-hash.json +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/conftest.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/fixtures/report_v014.json +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_breaker.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_config.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_crewai_callback_contracts.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_crewai_execution.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_crewai_integration.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_dead_end.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_decorator.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_decorator_contracts.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_langchain_integration.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_langgraph_integration.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_langgraph_node_contracts.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_langgraph_resume.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_pivot.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_policy_edge_contracts.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_pricing.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_public_exports.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_release_notes.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_reporter.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_semantic_osc.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_step.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_step_adapters.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_strands_integration.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_terminal_state.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_token_velocity.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_tool_cleanup.py +0 -0
- {longguard-0.2.1.dev0 → longguard-0.2.3}/tests/test_tool_repeat.py +0 -0
|
@@ -7,20 +7,59 @@ on:
|
|
|
7
7
|
branches: [ main, master ]
|
|
8
8
|
|
|
9
9
|
jobs:
|
|
10
|
+
checks:
|
|
11
|
+
name: Code Quality, Typing, Docs & Package Verification
|
|
12
|
+
runs-on: ubuntu-latest
|
|
13
|
+
steps:
|
|
14
|
+
- uses: actions/checkout@v7
|
|
15
|
+
|
|
16
|
+
- name: Install uv
|
|
17
|
+
uses: astral-sh/setup-uv@v7
|
|
18
|
+
with:
|
|
19
|
+
version: "latest"
|
|
20
|
+
|
|
21
|
+
- name: Set up Python 3.11
|
|
22
|
+
run: uv python install 3.11
|
|
23
|
+
|
|
24
|
+
- name: Install dependencies
|
|
25
|
+
run: uv sync --python 3.11 --extra dev
|
|
26
|
+
|
|
27
|
+
- name: Run Ruff Linter & Formatter
|
|
28
|
+
run: |
|
|
29
|
+
uv run --python 3.11 ruff check src/ tests/ scripts/
|
|
30
|
+
|
|
31
|
+
- name: Run MyPy Type Checker
|
|
32
|
+
run: uv run --python 3.11 mypy src/
|
|
33
|
+
|
|
34
|
+
- name: Check Version Consistency
|
|
35
|
+
run: uv run --python 3.11 python scripts/check_version.py
|
|
36
|
+
|
|
37
|
+
- name: Strict Documentation Build Check
|
|
38
|
+
run: uv run --python 3.11 mkdocs build --strict
|
|
39
|
+
|
|
40
|
+
- name: Verify Package Build
|
|
41
|
+
run: uv build
|
|
42
|
+
|
|
43
|
+
- name: Validate Built Distribution and Metadata
|
|
44
|
+
run: |
|
|
45
|
+
uv run --python 3.11 twine check --strict dist/*
|
|
46
|
+
uv run --python 3.11 python scripts/check_version.py --check-dist
|
|
47
|
+
|
|
10
48
|
test:
|
|
11
49
|
name: Test on Python ${{ matrix.python-version }} (${{ matrix.os }})
|
|
50
|
+
needs: [checks]
|
|
12
51
|
runs-on: ${{ matrix.os }}
|
|
13
52
|
strategy:
|
|
14
53
|
fail-fast: false
|
|
15
54
|
matrix:
|
|
16
|
-
os: [ubuntu-latest, macos-latest]
|
|
17
|
-
python-version: ["3.11", "3.12"]
|
|
55
|
+
os: [ubuntu-latest, macos-latest, windows-latest]
|
|
56
|
+
python-version: ["3.11", "3.12", "3.13"]
|
|
18
57
|
|
|
19
58
|
steps:
|
|
20
|
-
- uses: actions/checkout@
|
|
59
|
+
- uses: actions/checkout@v7
|
|
21
60
|
|
|
22
61
|
- name: Install uv
|
|
23
|
-
uses: astral-sh/setup-uv@
|
|
62
|
+
uses: astral-sh/setup-uv@v7
|
|
24
63
|
with:
|
|
25
64
|
version: "latest"
|
|
26
65
|
|
|
@@ -28,47 +67,39 @@ jobs:
|
|
|
28
67
|
run: uv python install ${{ matrix.python-version }}
|
|
29
68
|
|
|
30
69
|
- name: Install dependencies
|
|
31
|
-
run:
|
|
32
|
-
uv sync --python ${{ matrix.python-version }} --extra dev
|
|
33
|
-
|
|
34
|
-
- name: Run Ruff Linter & Formatter
|
|
35
|
-
run: |
|
|
36
|
-
uv run --python ${{ matrix.python-version }} ruff check src/ tests/
|
|
37
|
-
|
|
38
|
-
- name: Run MyPy Type Checker
|
|
39
|
-
run: |
|
|
40
|
-
uv run --python ${{ matrix.python-version }} mypy src/
|
|
70
|
+
run: uv sync --python ${{ matrix.python-version }} --extra dev
|
|
41
71
|
|
|
42
72
|
- name: Run Pytest Suite with Coverage
|
|
43
|
-
|
|
44
|
-
|
|
73
|
+
if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.12'
|
|
74
|
+
run: uv run --python ${{ matrix.python-version }} pytest tests/ --cov=longguard --cov-report=xml --cov-report=term-missing
|
|
45
75
|
|
|
46
|
-
- name:
|
|
47
|
-
|
|
48
|
-
|
|
76
|
+
- name: Run Pytest Suite
|
|
77
|
+
if: "!(matrix.os == 'ubuntu-latest' && matrix.python-version == '3.12')"
|
|
78
|
+
run: uv run --python ${{ matrix.python-version }} pytest tests/
|
|
49
79
|
|
|
50
80
|
frameworks:
|
|
51
81
|
name: Real frameworks (${{ matrix.dependencies }})
|
|
82
|
+
needs: [checks]
|
|
52
83
|
runs-on: ubuntu-latest
|
|
53
84
|
strategy:
|
|
54
85
|
fail-fast: false
|
|
55
86
|
matrix:
|
|
56
87
|
dependencies: [locked, newest]
|
|
57
88
|
steps:
|
|
58
|
-
- uses: actions/checkout@
|
|
59
|
-
- uses: astral-sh/setup-uv@
|
|
89
|
+
- uses: actions/checkout@v7
|
|
90
|
+
- uses: astral-sh/setup-uv@v7
|
|
60
91
|
- run: uv python install 3.12
|
|
61
92
|
- name: Install framework contracts
|
|
62
93
|
run: uv sync --locked --python 3.12 --extra dev --extra langgraph --extra langchain --extra crewai
|
|
63
94
|
- name: Exercise newest compatible frameworks without changing the lockfile
|
|
64
95
|
if: matrix.dependencies == 'newest'
|
|
65
|
-
run: uv pip install --upgrade 'langgraph>=1.0.0' 'langchain>=1.0.0' 'crewai>=
|
|
96
|
+
run: uv pip install --upgrade 'langgraph>=1.0.0' 'langchain>=1.0.0' 'crewai>=1.15.0,<1.15.29'
|
|
66
97
|
- name: Require framework presence
|
|
67
98
|
run: >-
|
|
68
99
|
uv run --no-sync python -c "from importlib.metadata import version; print({n: version(n) for n in ('langgraph', 'langchain', 'crewai')})"
|
|
69
100
|
- name: Execute gates and require real framework tests
|
|
70
101
|
run: uv run --no-sync python scripts/verify_release.py
|
|
71
|
-
- uses: actions/upload-artifact@
|
|
102
|
+
- uses: actions/upload-artifact@v7
|
|
72
103
|
if: always()
|
|
73
104
|
with:
|
|
74
105
|
name: checks-${{ matrix.dependencies }}
|
|
@@ -77,7 +108,7 @@ jobs:
|
|
|
77
108
|
release-coverage.json
|
|
78
109
|
- name: Reproduce provisional benchmark
|
|
79
110
|
run: uv run --no-sync python tests/benchmarks/run.py --output /tmp/benchmark.json
|
|
80
|
-
- uses: actions/upload-artifact@
|
|
111
|
+
- uses: actions/upload-artifact@v7
|
|
81
112
|
with:
|
|
82
113
|
name: benchmark-${{ matrix.dependencies }}
|
|
83
114
|
path: /tmp/benchmark.json
|
|
@@ -0,0 +1,190 @@
|
|
|
1
|
+
name: Release
|
|
2
|
+
|
|
3
|
+
on:
|
|
4
|
+
workflow_call:
|
|
5
|
+
inputs:
|
|
6
|
+
tag:
|
|
7
|
+
required: true
|
|
8
|
+
type: string
|
|
9
|
+
publish:
|
|
10
|
+
default: false
|
|
11
|
+
type: boolean
|
|
12
|
+
testpypi:
|
|
13
|
+
default: false
|
|
14
|
+
type: boolean
|
|
15
|
+
workflow_dispatch:
|
|
16
|
+
inputs:
|
|
17
|
+
tag:
|
|
18
|
+
description: "Existing version tag to release (for example v0.2.2 or v0.2.2rc1)"
|
|
19
|
+
required: true
|
|
20
|
+
publish:
|
|
21
|
+
description: "Publish to PyPI"
|
|
22
|
+
default: false
|
|
23
|
+
type: boolean
|
|
24
|
+
testpypi:
|
|
25
|
+
description: "Publish dry run to TestPyPI"
|
|
26
|
+
default: false
|
|
27
|
+
type: boolean
|
|
28
|
+
|
|
29
|
+
jobs:
|
|
30
|
+
build-and-release:
|
|
31
|
+
runs-on: ubuntu-latest
|
|
32
|
+
permissions:
|
|
33
|
+
contents: write
|
|
34
|
+
outputs:
|
|
35
|
+
tag: ${{ steps.notes.outputs.tag }}
|
|
36
|
+
version: ${{ steps.meta.outputs.version }}
|
|
37
|
+
prerelease: ${{ steps.notes.outputs.prerelease }}
|
|
38
|
+
publish: ${{ steps.flags.outputs.publish }}
|
|
39
|
+
testpypi: ${{ steps.flags.outputs.testpypi }}
|
|
40
|
+
steps:
|
|
41
|
+
- uses: actions/checkout@v7
|
|
42
|
+
with:
|
|
43
|
+
ref: ${{ inputs.tag || github.ref }}
|
|
44
|
+
fetch-depth: 0
|
|
45
|
+
- uses: astral-sh/setup-uv@v7
|
|
46
|
+
- run: uv python install 3.12
|
|
47
|
+
- run: uv sync --locked --python 3.12 --extra dev --extra langgraph --extra langchain --extra crewai
|
|
48
|
+
- name: Verify tag and source identity
|
|
49
|
+
env:
|
|
50
|
+
RELEASE_TAG: ${{ inputs.tag || github.ref_name }}
|
|
51
|
+
run: |
|
|
52
|
+
git show-ref --verify "refs/tags/$RELEASE_TAG"
|
|
53
|
+
test "$(git rev-list -n 1 "$RELEASE_TAG")" = "$(git rev-parse HEAD)"
|
|
54
|
+
- name: Verify version consistency against tag
|
|
55
|
+
env:
|
|
56
|
+
RELEASE_TAG: ${{ inputs.tag || github.ref_name }}
|
|
57
|
+
run: |
|
|
58
|
+
uv run --no-sync python scripts/check_version.py --tag "$RELEASE_TAG"
|
|
59
|
+
- name: Execute and record release gates
|
|
60
|
+
env:
|
|
61
|
+
UV_CACHE_DIR: /tmp/longguard-uv-cache
|
|
62
|
+
run: uv run --no-sync python scripts/verify_release.py
|
|
63
|
+
- name: Extract metadata
|
|
64
|
+
id: meta
|
|
65
|
+
run: |
|
|
66
|
+
VERSION=$(uv run --no-sync python -c "import tomllib, pathlib; print(tomllib.loads(pathlib.Path('pyproject.toml').read_text())['project']['version'])")
|
|
67
|
+
echo "version=$VERSION" >> "$GITHUB_OUTPUT"
|
|
68
|
+
- name: Generate literal release notes
|
|
69
|
+
id: notes
|
|
70
|
+
env:
|
|
71
|
+
RELEASE_TAG: ${{ inputs.tag || github.ref_name }}
|
|
72
|
+
run: >-
|
|
73
|
+
uv run --no-sync python scripts/release_notes.py --tag "$RELEASE_TAG"
|
|
74
|
+
--evidence release-evidence.json --output release_body.md
|
|
75
|
+
--github-output "$GITHUB_OUTPUT"
|
|
76
|
+
- name: Determine publish flags
|
|
77
|
+
id: flags
|
|
78
|
+
run: |
|
|
79
|
+
PUBLISH="${{ inputs.publish == true }}"
|
|
80
|
+
TESTPYPI="${{ inputs.testpypi == true }}"
|
|
81
|
+
if [ "$PUBLISH" = "true" ]; then
|
|
82
|
+
echo "publish=true" >> "$GITHUB_OUTPUT"
|
|
83
|
+
else
|
|
84
|
+
echo "publish=false" >> "$GITHUB_OUTPUT"
|
|
85
|
+
fi
|
|
86
|
+
echo "testpypi=$TESTPYPI" >> "$GITHUB_OUTPUT"
|
|
87
|
+
- name: Upload check evidence
|
|
88
|
+
if: always()
|
|
89
|
+
uses: actions/upload-artifact@v7
|
|
90
|
+
with:
|
|
91
|
+
name: release-evidence
|
|
92
|
+
path: |
|
|
93
|
+
release-evidence.json
|
|
94
|
+
release-coverage.json
|
|
95
|
+
- name: Upload distribution packages
|
|
96
|
+
uses: actions/upload-artifact@v7
|
|
97
|
+
with:
|
|
98
|
+
name: dist-packages
|
|
99
|
+
path: dist/*
|
|
100
|
+
- name: Create GitHub release
|
|
101
|
+
uses: softprops/action-gh-release@v3
|
|
102
|
+
with:
|
|
103
|
+
tag_name: ${{ steps.notes.outputs.tag }}
|
|
104
|
+
name: LongGuard ${{ steps.notes.outputs.tag }}
|
|
105
|
+
files: |
|
|
106
|
+
dist/*.whl
|
|
107
|
+
dist/*.tar.gz
|
|
108
|
+
release-evidence.json
|
|
109
|
+
release-coverage.json
|
|
110
|
+
body_path: release_body.md
|
|
111
|
+
draft: ${{ steps.flags.outputs.publish != 'true' }}
|
|
112
|
+
prerelease: ${{ steps.notes.outputs.prerelease == 'true' }}
|
|
113
|
+
make_latest: ${{ steps.notes.outputs.latest }}
|
|
114
|
+
|
|
115
|
+
publish-testpypi:
|
|
116
|
+
needs: build-and-release
|
|
117
|
+
if: needs.build-and-release.outputs.testpypi == 'true'
|
|
118
|
+
runs-on: ubuntu-latest
|
|
119
|
+
environment: testpypi
|
|
120
|
+
permissions:
|
|
121
|
+
id-token: write
|
|
122
|
+
steps:
|
|
123
|
+
- name: Download distribution packages
|
|
124
|
+
uses: actions/download-artifact@v8
|
|
125
|
+
with:
|
|
126
|
+
name: dist-packages
|
|
127
|
+
path: dist
|
|
128
|
+
- name: Publish to TestPyPI via Trusted Publishing
|
|
129
|
+
uses: pypa/gh-action-pypi-publish@release/v1
|
|
130
|
+
with:
|
|
131
|
+
repository-url: https://test.pypi.org/legacy/
|
|
132
|
+
attestations: true
|
|
133
|
+
skip-existing: true
|
|
134
|
+
|
|
135
|
+
publish-pypi:
|
|
136
|
+
needs: build-and-release
|
|
137
|
+
if: needs.build-and-release.outputs.publish == 'true' && needs.build-and-release.outputs.prerelease != 'true'
|
|
138
|
+
runs-on: ubuntu-latest
|
|
139
|
+
environment: pypi
|
|
140
|
+
permissions:
|
|
141
|
+
id-token: write
|
|
142
|
+
steps:
|
|
143
|
+
- name: Download distribution packages
|
|
144
|
+
uses: actions/download-artifact@v8
|
|
145
|
+
with:
|
|
146
|
+
name: dist-packages
|
|
147
|
+
path: dist
|
|
148
|
+
- name: Publish to PyPI via Trusted Publishing
|
|
149
|
+
uses: pypa/gh-action-pypi-publish@release/v1
|
|
150
|
+
with:
|
|
151
|
+
attestations: true
|
|
152
|
+
|
|
153
|
+
smoke-test:
|
|
154
|
+
needs: [build-and-release, publish-pypi]
|
|
155
|
+
if: needs.build-and-release.outputs.publish == 'true' && needs.build-and-release.outputs.prerelease != 'true'
|
|
156
|
+
runs-on: ubuntu-latest
|
|
157
|
+
steps:
|
|
158
|
+
- uses: astral-sh/setup-uv@v7
|
|
159
|
+
- name: Verify package installation and breaker trip from PyPI
|
|
160
|
+
env:
|
|
161
|
+
TARGET_VERSION: ${{ needs.build-and-release.outputs.version }}
|
|
162
|
+
run: |
|
|
163
|
+
mkdir /tmp/smoke-test
|
|
164
|
+
cd /tmp/smoke-test
|
|
165
|
+
uv venv .venv --python 3.12
|
|
166
|
+
echo "Waiting for longguard==$TARGET_VERSION to propagate on PyPI index..."
|
|
167
|
+
INSTALLED=false
|
|
168
|
+
for attempt in $(seq 1 6); do
|
|
169
|
+
echo "Attempt $attempt: pip install longguard==$TARGET_VERSION"
|
|
170
|
+
if uv pip install --python .venv/bin/python --no-cache "longguard==$TARGET_VERSION"; then
|
|
171
|
+
INSTALLED=true
|
|
172
|
+
break
|
|
173
|
+
fi
|
|
174
|
+
sleep 15
|
|
175
|
+
done
|
|
176
|
+
if [ "$INSTALLED" != "true" ]; then
|
|
177
|
+
echo "Failed to install longguard==$TARGET_VERSION from PyPI within timeout"
|
|
178
|
+
exit 1
|
|
179
|
+
fi
|
|
180
|
+
.venv/bin/python -c "
|
|
181
|
+
import longguard
|
|
182
|
+
from longguard import CircuitBreaker, GuardConfig, AgentStep
|
|
183
|
+
assert longguard.__version__ == '$TARGET_VERSION', f'Version mismatch: {longguard.__version__} != $TARGET_VERSION'
|
|
184
|
+
breaker = CircuitBreaker(GuardConfig(tool_repeat_threshold=2, max_reflections=1))
|
|
185
|
+
step = AgentStep(step_number=1, thought='loop', action='search', action_input={'q': 'x'}, observation='none')
|
|
186
|
+
d1 = breaker.check(step)
|
|
187
|
+
d2 = breaker.check(step)
|
|
188
|
+
assert d2.action in ('reflect', 'kill'), f'Expected trip, got {d2.action}'
|
|
189
|
+
print('Post-publish smoke test PASSED for longguard', longguard.__version__)
|
|
190
|
+
"
|
|
@@ -7,6 +7,45 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
7
7
|
|
|
8
8
|
---
|
|
9
9
|
|
|
10
|
+
## [0.2.3] - 2026-11-12
|
|
11
|
+
|
|
12
|
+
### Added
|
|
13
|
+
|
|
14
|
+
- 9-cell continuous integration matrix across Ubuntu, macOS, and Windows on Python 3.11, 3.12, and 3.13 (`.github/workflows/ci.yml`).
|
|
15
|
+
- Added official `Programming Language :: Python :: 3.13` classifier to `pyproject.toml`.
|
|
16
|
+
- Configured explicit `asyncio_default_fixture_loop_scope = "function"` in pytest configuration for modern pytest-asyncio compatibility.
|
|
17
|
+
|
|
18
|
+
### Changed
|
|
19
|
+
|
|
20
|
+
- CI efficiency optimization: separated code quality, strict typing (`mypy`), documentation builds (`mkdocs --strict`), and package validation (`twine check`) into a fast standalone quality gates job, eliminating redundant execution across matrix cells.
|
|
21
|
+
- Updated README badges and documentation with verified cross-platform compatibility across Linux, macOS, and Windows.
|
|
22
|
+
|
|
23
|
+
### Fixed
|
|
24
|
+
|
|
25
|
+
- Enforced explicit `utf-8` encoding across report persistence, version validation, benchmark execution, and test fixtures for Windows compatibility (preventing default ANSI/CP1252 decoding failures).
|
|
26
|
+
|
|
27
|
+
### Compatibility and limits
|
|
28
|
+
|
|
29
|
+
- No runtime change: detectors, circuit breaker state machine, defaults, budget boundaries, and report schemas are identical to 0.2.0.
|
|
30
|
+
|
|
31
|
+
## [0.2.2] - 2026-11-10
|
|
32
|
+
|
|
33
|
+
### Added
|
|
34
|
+
|
|
35
|
+
- Tag-driven GitHub Actions release pipeline with PyPI Trusted Publishing (OIDC), digital build provenance attestations, and post-publish clean virtual environment smoke testing.
|
|
36
|
+
- Distribution and version consistency validation (`scripts/check_version.py` and `tests/test_check_version.py`) verifying parity between `pyproject.toml`, `src/longguard/__init__.py`, `uv.lock`, and release distribution artifacts.
|
|
37
|
+
- Twine metadata verification (`twine check --strict`) and built wheel/sdist contents audit integrated into release verification and CI gates.
|
|
38
|
+
|
|
39
|
+
### Changed
|
|
40
|
+
|
|
41
|
+
- Documentation limits: Added an explicit "Compatibility & Known Limits" section to README and docs documenting runtime boundaries for guarded graph resume, CrewAI sequential ReAct execution, and detector cost estimation.
|
|
42
|
+
- Clarified "LongGuard vs. LangGraph recursion_limit" comparison table to state synchronous same-process support for graph resume.
|
|
43
|
+
- Strict documentation build (`mkdocs build --strict`) enforced in standard pull request CI.
|
|
44
|
+
|
|
45
|
+
### Compatibility and limits
|
|
46
|
+
|
|
47
|
+
- No runtime change: detectors, circuit breaker state machine, defaults, budget boundaries, and report schemas are identical to 0.2.0.
|
|
48
|
+
|
|
10
49
|
## [0.2.1] - 2026-11-09
|
|
11
50
|
|
|
12
51
|
### Added
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.5
|
|
2
2
|
Name: longguard
|
|
3
|
-
Version: 0.2.
|
|
3
|
+
Version: 0.2.3
|
|
4
4
|
Summary: Circuit breaker middleware for LangGraph/LangChain agents — detect loops, inject Reflect & Pivot, prevent runaway costs
|
|
5
5
|
Project-URL: Homepage, https://github.com/ENDEVSOLS/LongGuard
|
|
6
6
|
Project-URL: Repository, https://github.com/ENDEVSOLS/LongGuard
|
|
@@ -19,6 +19,7 @@ Classifier: Operating System :: OS Independent
|
|
|
19
19
|
Classifier: Programming Language :: Python :: 3
|
|
20
20
|
Classifier: Programming Language :: Python :: 3.11
|
|
21
21
|
Classifier: Programming Language :: Python :: 3.12
|
|
22
|
+
Classifier: Programming Language :: Python :: 3.13
|
|
22
23
|
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
23
24
|
Classifier: Topic :: Software Development :: Libraries :: Python Modules
|
|
24
25
|
Classifier: Topic :: System :: Monitoring
|
|
@@ -26,7 +27,7 @@ Classifier: Typing :: Typed
|
|
|
26
27
|
Requires-Python: >=3.11
|
|
27
28
|
Requires-Dist: numpy>=1.24.0
|
|
28
29
|
Provides-Extra: all
|
|
29
|
-
Requires-Dist: crewai<1.15.29,>=
|
|
30
|
+
Requires-Dist: crewai<1.15.29,>=1.15.0; extra == 'all'
|
|
30
31
|
Requires-Dist: langchain-core>=1.0.0; extra == 'all'
|
|
31
32
|
Requires-Dist: langchain>=1.0.0; extra == 'all'
|
|
32
33
|
Requires-Dist: langgraph>=1.0.0; extra == 'all'
|
|
@@ -35,7 +36,7 @@ Requires-Dist: mkdocs>=1.5.0; extra == 'all'
|
|
|
35
36
|
Requires-Dist: pymdown-extensions>=10.7; extra == 'all'
|
|
36
37
|
Requires-Dist: sentence-transformers>=2.7.0; extra == 'all'
|
|
37
38
|
Provides-Extra: crewai
|
|
38
|
-
Requires-Dist: crewai<1.15.29,>=
|
|
39
|
+
Requires-Dist: crewai<1.15.29,>=1.15.0; extra == 'crewai'
|
|
39
40
|
Provides-Extra: dev
|
|
40
41
|
Requires-Dist: mkdocs-material>=9.5.0; extra == 'dev'
|
|
41
42
|
Requires-Dist: mkdocs>=1.5.0; extra == 'dev'
|
|
@@ -44,6 +45,7 @@ Requires-Dist: pytest-asyncio>=1.0.0; extra == 'dev'
|
|
|
44
45
|
Requires-Dist: pytest-cov>=5.0.0; extra == 'dev'
|
|
45
46
|
Requires-Dist: pytest>=8.0.0; extra == 'dev'
|
|
46
47
|
Requires-Dist: ruff>=0.5.0; extra == 'dev'
|
|
48
|
+
Requires-Dist: twine>=5.0.0; extra == 'dev'
|
|
47
49
|
Provides-Extra: docs
|
|
48
50
|
Requires-Dist: mkdocs-material>=9.5.0; extra == 'docs'
|
|
49
51
|
Requires-Dist: mkdocs>=1.5.0; extra == 'docs'
|
|
@@ -69,6 +71,7 @@ Description-Content-Type: text/markdown
|
|
|
69
71
|
[](https://pypi.org/project/longguard/)
|
|
70
72
|
[](https://pypi.org/project/longguard/)
|
|
71
73
|
[](https://pypi.org/project/longguard/)
|
|
74
|
+
[](https://github.com/ENDEVSOLS/LongGuard)
|
|
72
75
|
[](https://github.com/ENDEVSOLS/LongGuard/actions/workflows/ci.yml)
|
|
73
76
|
[](https://endevsols.github.io/LongGuard/)
|
|
74
77
|
[](https://opensource.org/licenses/MIT)
|
|
@@ -103,22 +106,6 @@ LangGraph's built-in `recursion_limit` is a **hard crash** (`GraphRecursionError
|
|
|
103
106
|
|
|
104
107
|
---
|
|
105
108
|
|
|
106
|
-
<a id="part-of-the-long-suite"></a>
|
|
107
|
-
## 🌐 Part of the Long Suite
|
|
108
|
-
|
|
109
|
-
LongGuard is part of the **[EnDevSols Long Suite](https://endevsols.com/open-source)** of open-source production AI tools:
|
|
110
|
-
|
|
111
|
-
- **[LongParser](https://github.com/ENDEVSOLS/LongParser)** — High-speed, privacy-first local document ingestion & chunking (PDF, DOCX, PPTX, XLSX)
|
|
112
|
-
- **[LongTrainer](https://github.com/ENDEVSOLS/Long-Trainer)** — Production multi-tenant RAG chatbot and agent framework
|
|
113
|
-
- **[LongCache](https://github.com/ENDEVSOLS/LongCache)** — Production semantic similarity caching for LLM and RAG pipelines
|
|
114
|
-
- **[LongTracer](https://github.com/ENDEVSOLS/LongTracer)** — Post-generation hallucination detection via hybrid STS + NLI claim verification
|
|
115
|
-
- **[LongProbe](https://github.com/ENDEVSOLS/LongProbe)** — Sub-second RAG retrieval regression testing with pytest
|
|
116
|
-
- **[LongGuard](https://github.com/ENDEVSOLS/LongGuard)** — **In-flight runtime agent circuit breaker & reasoning loop recovery** ← *You are here*
|
|
117
|
-
|
|
118
|
-
Together, the Long Suite covers the full AI lifecycle from data ingestion and retrieval CI regression to runtime agent safety and post-generation verification.
|
|
119
|
-
|
|
120
|
-
---
|
|
121
|
-
|
|
122
109
|
<a id="why-longguard"></a>
|
|
123
110
|
## 💡 Why LongGuard?
|
|
124
111
|
|
|
@@ -170,7 +157,7 @@ pip install longguard[embeddings]
|
|
|
170
157
|
pip install longguard[all]
|
|
171
158
|
```
|
|
172
159
|
|
|
173
|
-
> **Python 3.11
|
|
160
|
+
> **Python 3.11 – 3.13 tested and supported** on Linux, macOS, and Windows.
|
|
174
161
|
|
|
175
162
|
### 1. LangGraph Integration (1 Line)
|
|
176
163
|
|
|
@@ -351,13 +338,27 @@ print(breaker.report.summary())
|
|
|
351
338
|
| **Detects Semantic Reasoning Loops** | ❌ No | ✅ **Yes** |
|
|
352
339
|
| **Detects Sudden Cost / Token Spikes** | ❌ No | ✅ **Yes** |
|
|
353
340
|
| **Auto-Injects Recovery Prompts** | ❌ No | ✅ **Yes** |
|
|
354
|
-
| **Safe Graph Resume (no tool replay)** | ❌ No | ✅ **Yes** |
|
|
341
|
+
| **Safe Graph Resume (no tool replay)** | ❌ No | ✅ **Yes** *(sync, same-process)* |
|
|
355
342
|
| **Exit Behavior** | 💥 Unhandled Exception (`Crash`) | 🛡️ **Graceful State Preservation** |
|
|
356
343
|
| **Run Reporting & Telemetry** | ❌ No | ✅ **JSON & Summary Reports** |
|
|
357
344
|
| **Configurable Thresholds** | ❌ Single integer | ✅ **Granular `GuardConfig`** |
|
|
358
345
|
|
|
359
346
|
---
|
|
360
347
|
|
|
348
|
+
<a id="compatibility-and-limits"></a>
|
|
349
|
+
## ⚠️ Compatibility & Known Limits
|
|
350
|
+
|
|
351
|
+
LongGuard runtime contracts are tested and verified under specific execution boundaries:
|
|
352
|
+
|
|
353
|
+
- **Guarded LangGraph Resume (`compile_guarded_graph` / `GuardedGraph.resume`):** Supports explicit continuation for synchronous two-node graphs (`START -> agent -> tools/END`, `tools -> agent`) within the same controller process using `InMemorySaver`. Async graphs, streaming, subgraphs, parallel tool dispatch, namespaces, and distributed checkpoint continuation are outside current support.
|
|
354
|
+
- **CrewAI Execution (`add_guard_to_crew`):** Explicitly bounded to the tested sequential, synchronous ReAct `CrewAgentExecutor`. Async, hierarchical, batch, planning, memory, or native tool-batch modes fail clearly; callback-only `CrewGuard` remains available.
|
|
355
|
+
- **Detector Evaluation & Cost Tracking:** Held-out detector evaluation metrics are provisional. Dollar cost estimation applies a 50/50 input/output token split when provider-specific token breakdowns are unavailable and is not an invoice billing guarantee.
|
|
356
|
+
- **Non-Invasive Telemetry:** Telemetry reports and checkpoints record execution history; they do not authorize replay or modify external service state.
|
|
357
|
+
|
|
358
|
+
See the full [Tested Compatibility Guide](https://endevsols.github.io/LongGuard/guides/compatibility/) for complete version matrices and verification details.
|
|
359
|
+
|
|
360
|
+
---
|
|
361
|
+
|
|
361
362
|
<a id="configuration"></a>
|
|
362
363
|
## ⚙️ Configuration at a Glance
|
|
363
364
|
|
|
@@ -388,6 +389,23 @@ config = GuardConfig(
|
|
|
388
389
|
|
|
389
390
|
👉 *For detailed documentation on custom detectors, embedding backends, and LangSmith telemetry, see the [Full Documentation](https://endevsols.github.io/LongGuard/).*
|
|
390
391
|
|
|
392
|
+
---
|
|
393
|
+
|
|
394
|
+
## Part of the Long Suite
|
|
395
|
+
|
|
396
|
+
LongGuard is part of the **Long Suite** — a collection of tools for building, testing, and monitoring production RAG systems.
|
|
397
|
+
|
|
398
|
+
| Project | Description |
|
|
399
|
+
|---|---|
|
|
400
|
+
| **[LongParser](https://github.com/ENDEVSOLS/LongParser)** | Document ingestion and chunking |
|
|
401
|
+
| **[LongTrainer](https://github.com/ENDEVSOLS/Long-Trainer)** | RAG chatbot framework |
|
|
402
|
+
| **[LongGuard](https://github.com/ENDEVSOLS/LongGuard)** | In-flight circuit breaker and reasoning loop recovery for agents *(you are here)* |
|
|
403
|
+
| **[LongCache](https://github.com/ENDEVSOLS/LongCache)** | Semantic similarity cache for LLM and RAG pipelines |
|
|
404
|
+
| **[LongTracer](https://github.com/ENDEVSOLS/LongTracer)** | Hallucination detection and tracing |
|
|
405
|
+
| **[LongProbe](https://github.com/ENDEVSOLS/LongProbe)** | Retrieval regression testing |
|
|
406
|
+
|
|
407
|
+
---
|
|
408
|
+
|
|
391
409
|
## 🤝 Contributing
|
|
392
410
|
|
|
393
411
|
Contributions are welcome! Please see [CONTRIBUTING.md](https://github.com/ENDEVSOLS/LongGuard/blob/main/CONTRIBUTING.md) for guidelines on code formatting, running tests, and opening pull requests.
|
|
@@ -9,6 +9,7 @@
|
|
|
9
9
|
[](https://pypi.org/project/longguard/)
|
|
10
10
|
[](https://pypi.org/project/longguard/)
|
|
11
11
|
[](https://pypi.org/project/longguard/)
|
|
12
|
+
[](https://github.com/ENDEVSOLS/LongGuard)
|
|
12
13
|
[](https://github.com/ENDEVSOLS/LongGuard/actions/workflows/ci.yml)
|
|
13
14
|
[](https://endevsols.github.io/LongGuard/)
|
|
14
15
|
[](https://opensource.org/licenses/MIT)
|
|
@@ -43,22 +44,6 @@ LangGraph's built-in `recursion_limit` is a **hard crash** (`GraphRecursionError
|
|
|
43
44
|
|
|
44
45
|
---
|
|
45
46
|
|
|
46
|
-
<a id="part-of-the-long-suite"></a>
|
|
47
|
-
## 🌐 Part of the Long Suite
|
|
48
|
-
|
|
49
|
-
LongGuard is part of the **[EnDevSols Long Suite](https://endevsols.com/open-source)** of open-source production AI tools:
|
|
50
|
-
|
|
51
|
-
- **[LongParser](https://github.com/ENDEVSOLS/LongParser)** — High-speed, privacy-first local document ingestion & chunking (PDF, DOCX, PPTX, XLSX)
|
|
52
|
-
- **[LongTrainer](https://github.com/ENDEVSOLS/Long-Trainer)** — Production multi-tenant RAG chatbot and agent framework
|
|
53
|
-
- **[LongCache](https://github.com/ENDEVSOLS/LongCache)** — Production semantic similarity caching for LLM and RAG pipelines
|
|
54
|
-
- **[LongTracer](https://github.com/ENDEVSOLS/LongTracer)** — Post-generation hallucination detection via hybrid STS + NLI claim verification
|
|
55
|
-
- **[LongProbe](https://github.com/ENDEVSOLS/LongProbe)** — Sub-second RAG retrieval regression testing with pytest
|
|
56
|
-
- **[LongGuard](https://github.com/ENDEVSOLS/LongGuard)** — **In-flight runtime agent circuit breaker & reasoning loop recovery** ← *You are here*
|
|
57
|
-
|
|
58
|
-
Together, the Long Suite covers the full AI lifecycle from data ingestion and retrieval CI regression to runtime agent safety and post-generation verification.
|
|
59
|
-
|
|
60
|
-
---
|
|
61
|
-
|
|
62
47
|
<a id="why-longguard"></a>
|
|
63
48
|
## 💡 Why LongGuard?
|
|
64
49
|
|
|
@@ -110,7 +95,7 @@ pip install longguard[embeddings]
|
|
|
110
95
|
pip install longguard[all]
|
|
111
96
|
```
|
|
112
97
|
|
|
113
|
-
> **Python 3.11
|
|
98
|
+
> **Python 3.11 – 3.13 tested and supported** on Linux, macOS, and Windows.
|
|
114
99
|
|
|
115
100
|
### 1. LangGraph Integration (1 Line)
|
|
116
101
|
|
|
@@ -291,13 +276,27 @@ print(breaker.report.summary())
|
|
|
291
276
|
| **Detects Semantic Reasoning Loops** | ❌ No | ✅ **Yes** |
|
|
292
277
|
| **Detects Sudden Cost / Token Spikes** | ❌ No | ✅ **Yes** |
|
|
293
278
|
| **Auto-Injects Recovery Prompts** | ❌ No | ✅ **Yes** |
|
|
294
|
-
| **Safe Graph Resume (no tool replay)** | ❌ No | ✅ **Yes** |
|
|
279
|
+
| **Safe Graph Resume (no tool replay)** | ❌ No | ✅ **Yes** *(sync, same-process)* |
|
|
295
280
|
| **Exit Behavior** | 💥 Unhandled Exception (`Crash`) | 🛡️ **Graceful State Preservation** |
|
|
296
281
|
| **Run Reporting & Telemetry** | ❌ No | ✅ **JSON & Summary Reports** |
|
|
297
282
|
| **Configurable Thresholds** | ❌ Single integer | ✅ **Granular `GuardConfig`** |
|
|
298
283
|
|
|
299
284
|
---
|
|
300
285
|
|
|
286
|
+
<a id="compatibility-and-limits"></a>
|
|
287
|
+
## ⚠️ Compatibility & Known Limits
|
|
288
|
+
|
|
289
|
+
LongGuard runtime contracts are tested and verified under specific execution boundaries:
|
|
290
|
+
|
|
291
|
+
- **Guarded LangGraph Resume (`compile_guarded_graph` / `GuardedGraph.resume`):** Supports explicit continuation for synchronous two-node graphs (`START -> agent -> tools/END`, `tools -> agent`) within the same controller process using `InMemorySaver`. Async graphs, streaming, subgraphs, parallel tool dispatch, namespaces, and distributed checkpoint continuation are outside current support.
|
|
292
|
+
- **CrewAI Execution (`add_guard_to_crew`):** Explicitly bounded to the tested sequential, synchronous ReAct `CrewAgentExecutor`. Async, hierarchical, batch, planning, memory, or native tool-batch modes fail clearly; callback-only `CrewGuard` remains available.
|
|
293
|
+
- **Detector Evaluation & Cost Tracking:** Held-out detector evaluation metrics are provisional. Dollar cost estimation applies a 50/50 input/output token split when provider-specific token breakdowns are unavailable and is not an invoice billing guarantee.
|
|
294
|
+
- **Non-Invasive Telemetry:** Telemetry reports and checkpoints record execution history; they do not authorize replay or modify external service state.
|
|
295
|
+
|
|
296
|
+
See the full [Tested Compatibility Guide](https://endevsols.github.io/LongGuard/guides/compatibility/) for complete version matrices and verification details.
|
|
297
|
+
|
|
298
|
+
---
|
|
299
|
+
|
|
301
300
|
<a id="configuration"></a>
|
|
302
301
|
## ⚙️ Configuration at a Glance
|
|
303
302
|
|
|
@@ -328,6 +327,23 @@ config = GuardConfig(
|
|
|
328
327
|
|
|
329
328
|
👉 *For detailed documentation on custom detectors, embedding backends, and LangSmith telemetry, see the [Full Documentation](https://endevsols.github.io/LongGuard/).*
|
|
330
329
|
|
|
330
|
+
---
|
|
331
|
+
|
|
332
|
+
## Part of the Long Suite
|
|
333
|
+
|
|
334
|
+
LongGuard is part of the **Long Suite** — a collection of tools for building, testing, and monitoring production RAG systems.
|
|
335
|
+
|
|
336
|
+
| Project | Description |
|
|
337
|
+
|---|---|
|
|
338
|
+
| **[LongParser](https://github.com/ENDEVSOLS/LongParser)** | Document ingestion and chunking |
|
|
339
|
+
| **[LongTrainer](https://github.com/ENDEVSOLS/Long-Trainer)** | RAG chatbot framework |
|
|
340
|
+
| **[LongGuard](https://github.com/ENDEVSOLS/LongGuard)** | In-flight circuit breaker and reasoning loop recovery for agents *(you are here)* |
|
|
341
|
+
| **[LongCache](https://github.com/ENDEVSOLS/LongCache)** | Semantic similarity cache for LLM and RAG pipelines |
|
|
342
|
+
| **[LongTracer](https://github.com/ENDEVSOLS/LongTracer)** | Hallucination detection and tracing |
|
|
343
|
+
| **[LongProbe](https://github.com/ENDEVSOLS/LongProbe)** | Retrieval regression testing |
|
|
344
|
+
|
|
345
|
+
---
|
|
346
|
+
|
|
331
347
|
## 🤝 Contributing
|
|
332
348
|
|
|
333
349
|
Contributions are welcome! Please see [CONTRIBUTING.md](https://github.com/ENDEVSOLS/LongGuard/blob/main/CONTRIBUTING.md) for guidelines on code formatting, running tests, and opening pull requests.
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
# Tested execution and compatibility
|
|
2
2
|
|
|
3
|
-
|
|
3
|
+
LongGuard 0.2.x extends the existing circuit breaker. Package dependency minimums
|
|
4
4
|
remain unchanged; they do not guarantee that every permitted framework release supports
|
|
5
5
|
these execution contracts. Bare `import longguard` keeps framework dependencies optional.
|
|
6
6
|
|
|
@@ -47,6 +47,9 @@ CLOSED ──(loop detected)──▶ REFLECTING ──(persists)──▶ HALF_
|
|
|
47
47
|
|
|
48
48
|
---
|
|
49
49
|
|
|
50
|
+
!!! warning "Runtime Boundaries & Known Limits"
|
|
51
|
+
LongGuard contracts are tested for synchronous execution boundaries (synchronous same-process LangGraph resume and sequential ReAct CrewAI execution). Review the full [Tested Compatibility & Limits Guide](guides/compatibility.md) for detailed runtime boundaries and usage side effects.
|
|
52
|
+
|
|
50
53
|
## 30-Second Example (LangGraph)
|
|
51
54
|
|
|
52
55
|
LongGuard wraps your agent nodes in a single line of code:
|