agentbridge-core 0.1.0__tar.gz

agentbridge_core-0.1.0/.env.example

@@ -0,0 +1,9 @@
+# AgentBridge does not read API keys itself. Provider CLIs use their own
+# authentication. Copy to .env only if your provider CLI expects these variables,
+# and never commit the .env file.
+
+# Claude Code can use an API key instead of an interactive login:
+# ANTHROPIC_API_KEY=
+
+# Optional: point AgentBench at a different benchmarks directory.
+# AGENTBENCH_DIR=./benchmarks

agentbridge_core-0.1.0/.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,23 @@
+---
+name: Bug report
+about: Something does not work as documented
+labels: bug
+---
+
+**What happened**
+
+**What you expected**
+
+**Steps to reproduce**
+
+1.
+2.
+
+**Environment**
+
+- AgentBridge version (`agentbridge --version`):
+- OS:
+- Python / Git versions:
+- Provider and version (e.g. `claude --version`):
+
+**Logs** (`agentbridge logs <task_id>`; please check that no secrets are included)

agentbridge_core-0.1.0/.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,5 @@
+blank_issues_enabled: false
+contact_links:
+  - name: Security vulnerability
+    url: https://docs.github.com/en/code-security/security-advisories/guidance-on-reporting-and-writing-information-about-vulnerabilities/privately-reporting-a-security-vulnerability
+    about: Please report security issues privately (see SECURITY.md).

agentbridge_core-0.1.0/.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,14 @@
+---
+name: Feature request
+about: Propose an improvement
+labels: enhancement
+---
+
+**Problem to solve**
+
+**Proposed solution**
+
+**Alternatives considered**
+
+**Provider capability involved?** If this depends on a provider feature, link the
+official documentation that describes it.

agentbridge_core-0.1.0/.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,13 @@
+## Summary
+
+<!-- What does this change and why? Link the issue. -->
+
+## Checklist
+
+- [ ] `make check` passes (ruff, format, mypy, pytest)
+- [ ] Tests added or updated
+- [ ] Documentation matches the new behaviour
+- [ ] CHANGELOG.md updated under "Unreleased"
+- [ ] No credentials, tokens or personal data added
+- [ ] No security check weakened; no agent/repository text is executed
+- [ ] Benchmark data (if touched) is real and unknown values remain `null`

agentbridge_core-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,33 @@
+name: ci
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+permissions:
+  contents: read
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+      - run: python -m pip install -e ".[dev]" -e packages/agentbridge-bench
+      - run: ruff check src packages tests
+      - run: ruff format --check src packages tests
+      - run: mypy
+
+  bench-controls:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+      - run: python -m pip install -e ".[dev]" -e packages/agentbridge-bench
+      - name: AgentBench harness controls (no paid APIs)
+        run: agentbridge-bench run smoke -c mock-oracle -c mock-noop --output /tmp/agentbench

agentbridge_core-0.1.0/.github/workflows/release.yml

@@ -0,0 +1,95 @@
+name: release
+
+# Push a tag like v0.1.0 to build, smoke-test, publish to PyPI and create a GitHub release.
+# PyPI publishing uses trusted publishing (OIDC): no PyPI password or token is stored
+# anywhere. See docs/releasing.md for the one-time PyPI setup.
+
+on:
+  push:
+    tags: ["v*"]
+
+permissions:
+  contents: read
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+      - name: Check that the tag matches both package versions
+        run: |
+          tag="${GITHUB_REF_NAME#v}"
+          for f in pyproject.toml packages/agentbridge-bench/pyproject.toml; do
+            v=$(python -c "import tomllib,sys;print(tomllib.load(open(sys.argv[1],'rb'))['project']['version'])" "$f")
+            [ "$v" = "$tag" ] || { echo "::error::$f has version $v but the tag is v$tag"; exit 1; }
+          done
+      - run: python -m pip install build
+      - name: Build agentbridge-core and agentbridge-bench
+        run: |
+          python -m build --outdir dist/core .
+          python -m build --wheel --outdir dist/bench packages/agentbridge-bench  # wheel only: it bundles ../../benchmarks
+      - name: Smoke-test both wheels in a clean environment
+        run: |
+          python -m venv /tmp/clean
+          /tmp/clean/bin/pip install dist/core/*.whl dist/bench/*.whl
+          /tmp/clean/bin/agentbridge --version
+          cd /tmp && /tmp/clean/bin/agentbridge-bench list
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  # PyPI allows one trusted-publisher configuration per project, so each package
+  # publishes from its own environment.
+  publish-core:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi
+      url: https://pypi.org/project/agentbridge-core/
+    permissions:
+      id-token: write # trusted publishing; no secrets involved
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: dist/core/
+
+  publish-bench:
+    needs: publish-core # agentbridge-bench depends on agentbridge-core
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi-bench
+      url: https://pypi.org/project/agentbridge-bench/
+    permissions:
+      id-token: write
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: dist/bench/
+
+  github-release:
+    needs: publish-bench
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - name: Create GitHub release
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: gh release create "$GITHUB_REF_NAME" dist/core/* dist/bench/* --notes-file CHANGELOG.md --verify-tag

agentbridge_core-0.1.0/.github/workflows/tests.yml

@@ -0,0 +1,30 @@
+name: tests
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+permissions:
+  contents: read
+
+jobs:
+  pytest:
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, windows-latest, macos-latest]
+        python-version: ["3.12", "3.13"]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: ${{ matrix.python-version }}
+      - run: python -m pip install -e ".[dev]" -e packages/agentbridge-bench
+      - name: Configure git identity for test repositories
+        run: |
+          git config --global user.name "CI"
+          git config --global user.email "ci@example.com"
+          git config --global init.defaultBranch main
+      - run: python -m pytest -p no:cacheprovider --cov=agentbridge --cov=agentbridge_bench --cov-report=term

agentbridge_core-0.1.0/.gitignore

@@ -0,0 +1,32 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+build/
+dist/
+.venv/
+venv/
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+
+# AgentBridge state (when AgentBridge is used on this repository)
+.agentbridge/
+agentbridge.yaml
+
+# AgentBench raw experiment runs (large; publish curated reports under benchmarks/reports/)
+benchmarks/experiments/
+
+# Secrets
+.env
+
+# OS / editor
+.DS_Store
+Thumbs.db
+.idea/
+.vscode/
+
+# Dashboard frontend dependencies (the built bundle in src/agentbridge/dashboard/static is committed)
+dashboard/node_modules/

agentbridge_core-0.1.0/CHANGELOG.md

@@ -0,0 +1,72 @@
+# Changelog
+
+All notable changes are documented here. The format follows
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and the project uses
+[Semantic Versioning](https://semver.org/).
+
+## Unreleased
+
+_Nothing yet._
+
+## 0.1.0 - 2026-11-08
+
+First release on PyPI.
+
+### Changed
+
+- Split into two distributions: **`agentbridge-core`** (the `agentbridge` command;
+  import name unchanged) and **`agentbridge-bench`** (AgentBench, import
+  `agentbridge_bench`, command `agentbridge-bench`). The PyPI name `agentbridge` belongs
+  to an unrelated project. `agentbridge benchmark …` still works when
+  `agentbridge-bench` is installed and prints an install hint otherwise.
+- Release workflow publishes both packages to PyPI with trusted publishing (no stored
+  tokens) after a tag/version check; see docs/releasing.md.
+- The `agentbridge-bench` wheel bundles the benchmark suite and works outside a source
+  checkout.
+
+### Fixed
+
+- AgentBench refuses to run without pytest, and a test command that fails without a
+  pytest summary now counts as a failure instead of silently reporting 0/0.
+
+### Added
+
+- Local web dashboard: `agentbridge dashboard` (FastAPI backend, React + TypeScript
+  frontend shipped prebuilt). Overview, runs, task graph, task detail with diff,
+  validation steps, agent output and event timeline, live updates over server-sent
+  events, token-protected approve/reject/cancel, AgentBench results. Loopback only.
+- Alembic schema migrations; pre-Alembic databases are upgraded in place.
+- `validation.sandbox: docker`: validation commands run in a network-less,
+  resource-limited container and fail closed if Docker is unavailable.
+  `validation.docker.command: [wsl, docker]` supports Docker inside WSL on Windows.
+
+- Claude Code container isolation (`providers.claude.container`): only the worktree
+  is mounted, a single token is passed by variable name, read-only root filesystem,
+  dropped capabilities, shell/web tools disallowed, fail-closed.
+
+### Fixed
+
+- Execution records no longer stay `running` when an adapter raises.
+- Third-party test-client deprecation warning silenced.
+
+### Added (foundations)
+
+- `agentbridge` CLI: `init`, `providers list|check`, `run`, `tasks
+  list|status|inspect|resume|cancel|approve|reject|cleanup`, `logs`, `report`,
+  `remote packet|import`, `benchmark list|run|report`, with stable exit codes.
+- Validated `agentbridge.yaml` configuration with safe defaults.
+- SQLite persistence (WAL), explicit task state machine with atomic transitions and
+  event history, restart recovery and safe resume.
+- Adapter contract with declared capabilities; mock, Claude Code CLI, Antigravity
+  remote handoff and generic CLI adapters.
+- Planner (single, standard template, provider JSON plan), explainable router,
+  dependency-aware scheduler, retries with exponential backoff, cancellation.
+- Git worktree isolation, diff capture, overlap and merge-conflict detection,
+  approval-gated merge.
+- Validation pipeline: diff, scope, isolation, Ruff, Mypy, Pytest, heuristic review,
+  optional agent review.
+- Security: policy profiles, command allowlists, protected git operations, secret
+  redaction, path traversal protection.
+- Task and run reports (JSON + Markdown).
+- AgentBench: 12 deterministic tasks across 8 categories, controls, runner, metrics,
+  JSON/CSV/Markdown/SVG reports.

agentbridge_core-0.1.0/CODE_OF_CONDUCT.md

@@ -0,0 +1,16 @@
+# Code of Conduct
+
+This project adopts the [Contributor Covenant, version 2.1](https://www.contributor-covenant.org/version/2/1/code_of_conduct/)
+as its code of conduct.
+
+In short: be respectful and constructive, assume good intent, welcome newcomers, and
+focus criticism on ideas and code rather than people. Harassment, discrimination and
+personal attacks are not tolerated in issues, pull requests, discussions or any other
+project space.
+
+## Reporting
+
+Report unacceptable behaviour privately to the project maintainers through the
+contact listed in the repository's security policy or profile. Reports are handled
+confidentially. Maintainers may remove content and restrict participation of anyone
+who violates this code.

agentbridge_core-0.1.0/CONTRIBUTING.md

@@ -0,0 +1,39 @@
+# Contributing to AgentBridge
+
+Thank you for helping. AgentBridge values correctness, reliability and security over
+feature count: a small, well-tested change is better than a large incomplete one.
+
+## Setup
+
+```bash
+python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
+pip install -e ".[dev]" -e packages/agentbridge-bench
+make check
+```
+
+## Workflow
+
+1. Open an issue for anything non-trivial so the design can be discussed first.
+2. Create a branch, keep the change focused, and add tests.
+3. Run `make check` (Ruff, format check, Mypy strict, Pytest). All must pass.
+4. Update documentation when behaviour changes. Docs must describe what the code does
+   today, not what it might do.
+5. Add an entry under "Unreleased" in `CHANGELOG.md`.
+6. Open a pull request using the template.
+
+## Ground rules
+
+- Never add code that executes text produced by agents, planners or repositories.
+- Never weaken a security check to make a test pass.
+- Provider adapters must declare only capabilities verified against official
+  documentation or the installed tool, and tests must use fake executables (no paid
+  API calls in CI).
+- Benchmark changes must not fabricate or impute results; unknown values stay `null`.
+- New benchmark tasks need a fixture, hidden tests, a reference solution, and must
+  pass under `mock-oracle` and fail under `mock-noop`.
+
+## Code style
+
+Ruff (lint + format, line length 100) and Mypy in strict mode for `src/`. Prefer small
+modules, typed Pydantic models at boundaries and explicit errors from
+`agentbridge.errors`.

agentbridge_core-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 AgentBridge contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

agentbridge_core-0.1.0/Makefile

@@ -0,0 +1,35 @@
+PYTHON ?= python
+PYTEST_ARGS ?=
+
+.PHONY: install lint format typecheck test check bench-smoke bench-controls clean
+
+install:
+	$(PYTHON) -m pip install -e ".[dev]" -e packages/agentbridge-bench
+
+lint:
+	$(PYTHON) -m ruff check src packages tests
+	$(PYTHON) -m ruff format --check src packages tests
+
+format:
+	$(PYTHON) -m ruff format src packages tests
+	$(PYTHON) -m ruff check --fix src packages tests
+
+typecheck:
+	$(PYTHON) -m mypy
+
+test:
+	$(PYTHON) -m pytest -p no:cacheprovider $(PYTEST_ARGS)
+
+coverage:
+	$(PYTHON) -m pytest -p no:cacheprovider --cov=agentbridge --cov=agentbridge_bench --cov-report=term-missing $(PYTEST_ARGS)
+
+check: lint typecheck test
+
+bench-smoke:
+	$(PYTHON) -m agentbridge_bench.cli run smoke -c mock-oracle -c mock-noop
+
+bench-controls:
+	$(PYTHON) -m agentbridge_bench.cli run core -c mock-oracle -c mock-noop -c antigravity-remote
+
+clean:
+	$(PYTHON) -c "import shutil; [shutil.rmtree(p, ignore_errors=True) for p in ('build', 'dist', 'htmlcov')]"