mirror of
https://github.com/PX4/PX4-Autopilot.git
synced 2026-11-22 22:50:49 +08:00
Tools/ci/ai_review reviews a PR with Claude Code on Amazon Bedrock and writes a pr-review artifact for pr-review-poster.py. - Pipeline: gather (metadata, linked issues, commits, CI status, per-file patches, earlier comments), a reviewer pass returning line findings, PR-level findings and a contribution checklist, an independent validator pass per finding, and code-owned routing: inline only at high confidence, otherwise a before-merge list or a collapsed section. No PR is skipped for size; generated files are listed but not diffed. - Trust: the PR is untrusted data. The agent runs --bare --restricted with read-only tools. Code it writes to check its math runs only in a px4-dev container (no network, no credentials, read-only checkout, gVisor in CI), and `harden` blocks the runner's metadata service first. Output containing a credential is never written. - A model refusal produces no review and passes the provider's exact reason, category and request ID through. - Models and per-role defaults live in models.json; models-check verifies access before switching. - Process checks (tests, description, upgrade notes, docs) live in a CI-only prompt so the interactive skill stays lean. Assisted-by: Claude:claude-opus-5-5 Signed-off-by: Ramon Roche <mrpollo@gmail.com>
94 lines
3.6 KiB
Python
94 lines
3.6 KiB
Python
"""The agent reads untrusted PR content while holding Bedrock credentials.
|
|
These tests pin the sandboxing flags, the environment it gets, and that
|
|
any failure is an error rather than an empty (clean-looking) review."""
|
|
|
|
import json
|
|
import subprocess
|
|
import unittest
|
|
from pathlib import Path
|
|
from typing import Any, Dict, List, Mapping
|
|
|
|
from ai_review import agent
|
|
|
|
CFG = agent.AgentConfig(model='global.anthropic.claude-opus-5-5',
|
|
effort='xhigh', max_turns=10, timeout_s=60)
|
|
|
|
|
|
class FakeRunner:
|
|
|
|
def __init__(self, stdout: str, returncode: int = 0) -> None:
|
|
self.stdout = stdout
|
|
self.returncode = returncode
|
|
self.calls: List[Dict[str, Any]] = []
|
|
|
|
def __call__(self, cmd: List[str], stdin: str, cwd: Path,
|
|
env: Mapping[str, str],
|
|
timeout: float) -> 'subprocess.CompletedProcess[str]':
|
|
self.calls.append({'cmd': cmd, 'stdin': stdin, 'env': dict(env)})
|
|
return subprocess.CompletedProcess(cmd, self.returncode, self.stdout,
|
|
'boom')
|
|
|
|
|
|
class TestCommand(unittest.TestCase):
|
|
|
|
def setUp(self) -> None:
|
|
self.cmd = agent.command(CFG, Path('/p.md'), {'type': 'object'},
|
|
'task')
|
|
|
|
def test_sandbox_flags(self) -> None:
|
|
for flag in ('--bare', '--restricted', '--strict-mcp-config'):
|
|
self.assertIn(flag, self.cmd)
|
|
i = self.cmd.index('--permission-mode')
|
|
self.assertEqual(self.cmd[i + 1], 'dontAsk')
|
|
|
|
def test_only_read_only_tools(self) -> None:
|
|
i = self.cmd.index('--tools')
|
|
self.assertEqual(self.cmd[i + 1], 'Read,Bash')
|
|
j = self.cmd.index('--allowedTools')
|
|
allowed = self.cmd[j + 1:self.cmd.index('--effort')]
|
|
self.assertEqual(allowed, list(agent.ALLOWED_TOOLS))
|
|
for entry in allowed:
|
|
# nothing that writes, runs code or reaches the network
|
|
self.assertRegex(entry, r'^(Read|Bash\((grep|ls|git (log|blame|'
|
|
r'show)) \*\))$')
|
|
|
|
|
|
class TestRun(unittest.TestCase):
|
|
|
|
def run_agent(self, runner: FakeRunner) -> agent.AgentResult:
|
|
return agent.run(CFG, Path('/p.md'), {}, 'task', 'big context',
|
|
Path('.'), runner=runner)
|
|
|
|
def test_parses_structured_output_and_cost(self) -> None:
|
|
runner = FakeRunner(json.dumps({
|
|
'structured_output': {'findings': []}, 'total_cost_usd': 1.5,
|
|
'usage': {'input_tokens': 10}}))
|
|
result = self.run_agent(runner)
|
|
self.assertEqual(result.output, {'findings': []})
|
|
self.assertEqual(result.cost_usd, 1.5)
|
|
self.assertEqual(runner.calls[0]['stdin'], 'big context')
|
|
|
|
def test_environment_selects_bedrock_and_drops_github_token(
|
|
self) -> None:
|
|
env = agent.environment(CFG, {'GITHUB_TOKEN': 'x', 'GH_TOKEN': 'y',
|
|
'PATH': '/bin'})
|
|
self.assertEqual(env['CLAUDE_CODE_USE_BEDROCK'], '1')
|
|
self.assertEqual(env['ANTHROPIC_MODEL'], CFG.model)
|
|
self.assertEqual(env['ANTHROPIC_DEFAULT_HAIKU_MODEL'], CFG.model)
|
|
self.assertNotIn('GITHUB_TOKEN', env)
|
|
self.assertNotIn('GH_TOKEN', env)
|
|
|
|
def test_failures_raise(self) -> None:
|
|
cases = [FakeRunner('{}', returncode=1),
|
|
FakeRunner('not json'),
|
|
FakeRunner(json.dumps({'is_error': True})),
|
|
FakeRunner(json.dumps({'result': 'text only'}))]
|
|
for runner in cases:
|
|
with self.subTest(stdout=runner.stdout):
|
|
with self.assertRaises(agent.AgentError):
|
|
self.run_agent(runner)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|