auditgate 0.2.1__tar.gz → 0.3.0__tar.gz

auditgate-0.3.0/CLAUDE.md

@@ -0,0 +1,81 @@
+# CLAUDE.md
+
+## Identity
+AuditGate — tamper-evident audit trail primitive for AI agents.
+Hash-chained logging with severity filtering and retention pruning. Part of the ActionGate governance stack. Apache-2.0. Zero dependencies.
+
+## Stack
+Python 3.12+ · Hatchling · mypy strict · ruff · PyPI: `auditgate` @ `actiongate-oss`
+
+## Layout
+```
+auditgate/           # core.py, engine.py, emitter.py, store.py, cli.py, __main__.py
+tests/               # pytest suite, test_golden_vectors.py, test_cli.py
+SEMANTICS.md         # normative spec (governs over implementation)
+pyproject.toml       # package config
+benchmarks/          # reproducible perf tests
+```
+
+## Commands
+```bash
+pytest tests/ -v                    # test
+mypy auditgate/ --strict            # typecheck
+ruff check auditgate/               # lint
+python -m build && twine upload dist/*  # publish
+git log --oneline -5 && git status  # session start check
+```
+
+## Inviolable Rules
+- NEVER modify SEMANTICS.md without explicit approval
+- NEVER force-push to main
+- Tests pass before every push
+
+## Code Standards
+**Architecture:**
+- Frozen dataclasses with slots: `@dataclass(frozen=True, slots=True)`
+- Enums use `auto()`
+- `_Missing` sentinel pattern with `MISSING` constant and `Result[T]`
+- Engine uses `__slots__`, accepts `store: Store | None`, `clock: Callable`
+- Store is a `Protocol`, MemoryStore uses threading locks
+- Emitter mixin: `on_decision()`, `_emit()`, `listener_errors`
+- Integrity modes: NONE, HASH, CHAIN (tamper-evident)
+- Severity filtering via AuditPolicy
+- Verdict enum: ALLOW, BLOCK, ERROR, OVERRIDE
+
+**Style:**
+- Every public symbol has type annotations. No exceptions.
+- No `Any` unless wrapping genuinely dynamic external input
+- Prefer `collections.abc` over `typing` for abstract types
+- Single responsibility: one class, one job. If a docstring needs "and", split it.
+- Functions ≤25 lines. Classes ≤200 lines. Files ≤400 lines. Exceed = refactor.
+- No magic numbers. Named constants or enum members only.
+- Error paths return typed errors, never raise except for programmer bugs
+- Guard clauses at top, happy path unindented. No deep nesting.
+
+**Naming:**
+- `snake_case` for functions/variables, `PascalCase` for classes, `UPPER_SNAKE` for constants
+- Boolean vars/params: `is_`, `has_`, `should_`, `can_` prefixes
+- Private: single underscore. No dunder methods unless implementing a protocol.
+
+**Testing:**
+- Every public method has at least one test
+- Test names: `test_<method>_<scenario>_<expected>`
+- No mocks unless testing I/O boundaries. Prefer real objects.
+- Edge cases: empty input, boundary values, hash chain verification, retention pruning, CLI output
+
+**Performance:**
+- Hot path (record) must be allocation-minimal
+- SHA-256 hashing on hot path is acceptable (required for integrity)
+- Prefer `__slots__` on all frequently instantiated classes
+- Store operations: O(1) lookup, O(n) scan only for admin ops (clear_all)
+
+## Cross-Gate Invariants
+All four gates (actiongate, budgetgate, rulegate, auditgate) share identical patterns:
+- Same Engine API: register, check, enforce, guard, guard_result, clear, clear_all
+- Same Store protocol: get, set, delete, keys
+- Same Emitter: on_decision, _emit, listener_errors
+- Same Decision/Result types with __bool__, to_dict
+- Same pyproject.toml: hatchling, >=3.12, Apache-2.0, same extras pattern
+
+## Commit Convention
+Imperative mood, scoped: `feat:`, `fix:`, `refactor:`, `docs:`, `chore:`

{auditgate-0.2.1 → auditgate-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: auditgate
-Version: 0.2.1
+Version: 0.3.0
 Summary: Compliance-grade audit logging for agent systems.
 Project-URL: Homepage, https://github.com/actiongate-oss/auditgate
 Project-URL: Documentation, https://github.com/actiongate-oss/auditgate#readme

{auditgate-0.2.1 → auditgate-0.3.0}/auditgate/__init__.py

@@ -25,7 +25,7 @@ from .core import (
 )
 from .emitter import Emitter
 from .engine import AuditError, Engine
-from .store import MemoryStore, QueryFilter, Store
+from .store import AsyncMemoryStore, AsyncStore, MemoryStore, QueryFilter, Store
 
 __all__ = [
     "Trail",
@@ -47,8 +47,10 @@ __all__ = [
     "AuditError",
     "Emitter",
     "Store",
+    "AsyncStore",
     "MemoryStore",
+    "AsyncMemoryStore",
     "QueryFilter",
 ]
 
-__version__ = "0.1.1"
+__version__ = "0.3.0"

{auditgate-0.2.1 → auditgate-0.3.0}/auditgate/cli.py

@@ -19,11 +19,12 @@ import argparse
 import json
 import sys
 from pathlib import Path
+from typing import Any
 
 from .core import compute_hash
 
 
-def _verify_log(entries: list[dict], verbose: bool = False) -> tuple[bool, list[str]]:
+def _verify_log(entries: list[dict[str, Any]], verbose: bool = False) -> tuple[bool, list[str]]:
     """Verify a list of serialized audit entries.
 
     Works with raw dicts from JSON (no AuditEntry objects needed).
@@ -35,7 +36,7 @@ def _verify_log(entries: list[dict], verbose: bool = False) -> tuple[bool, list[
         return True, ["Empty log — nothing to verify."]
 
     # Group by trail
-    trails: dict[str, list[dict]] = {}
+    trails: dict[str, list[dict[str, Any]]] = {}
     for entry in entries:
         trail = entry.get("trail", "<unknown>")
         trails.setdefault(trail, []).append(entry)
@@ -78,8 +79,10 @@ def _verify_log(entries: list[dict], verbose: bool = False) -> tuple[bool, list[
                 prev_entry_hash = prev_entry.get("entry_hash")
                 if prev_entry_hash != prev_hash:
                     all_valid = False
+                    prev_seq = prev_entry.get('sequence', '?')
                     messages.append(
-                        f"  [{seq}] FAIL — chain break (prev_hash does not match entry [{prev_entry.get('sequence', '?')}])"
+                        f"  [{seq}] FAIL — chain break"
+                        f" (prev_hash does not match entry [{prev_seq}])"
                     )
                     if verbose:
                         messages.append(f"         expected prev: {prev_entry_hash}")
@@ -99,7 +102,7 @@ def _verify_log(entries: list[dict], verbose: bool = False) -> tuple[bool, list[
                 messages.append(f"  [{seq}] OK — {verdict} hash:{entry_hash[:12]}...")
 
         if all_valid:
-            messages.append(f"  VALID — all hashes verified")
+            messages.append("  VALID — all hashes verified")
 
     return all_valid, messages
 
@@ -145,7 +148,7 @@ def _cmd_verify(args: argparse.Namespace) -> int:
         if valid:
             print(f"RESULT: VALID ({len(entries)} entries verified)")
         else:
-            print(f"RESULT: INVALID (chain integrity broken)")
+            print("RESULT: INVALID (chain integrity broken)")
 
     return 0 if valid else 1
 

{auditgate-0.2.1 → auditgate-0.3.0}/auditgate/core.py

@@ -11,7 +11,7 @@ from __future__ import annotations
 import hashlib
 import json
 from dataclasses import dataclass, field
-from datetime import datetime, timezone
+from datetime import UTC, datetime
 from enum import Enum, auto
 from typing import Any
 
@@ -170,12 +170,12 @@ class Result[T]:
             raise RuntimeError(
                 f"unwrap() called on dropped audit: {self.decision.reason}"
             )
-        return self._value  # type: ignore[return-value]
+        return self._value
 
     def unwrap_or(self, default: T) -> T:
         if isinstance(self._value, _Missing):
             return default
-        return self._value  # type: ignore[return-value]
+        return self._value
 
 
 # ── Integrity ──
@@ -224,8 +224,7 @@ def verify_chain(entries: list[AuditEntry]) -> tuple[bool, int | None]:
         if entry.entry_hash != expected:
             return False, entry.sequence
 
-        if entry.prev_hash is not None and i > 0:
-            if entries[i - 1].entry_hash != entry.prev_hash:
+        if entry.prev_hash is not None and i > 0 and entries[i - 1].entry_hash != entry.prev_hash:
                 return False, entry.sequence
 
     return True, None
@@ -233,4 +232,4 @@ def verify_chain(entries: list[AuditEntry]) -> tuple[bool, int | None]:
 
 def wall_clock() -> str:
     """Current wall-clock time as ISO 8601 UTC."""
-    return datetime.now(timezone.utc).isoformat()
+    return datetime.now(UTC).isoformat()

{auditgate-0.2.1 → auditgate-0.3.0}/auditgate/emitter.py

@@ -8,7 +8,8 @@
 
 from __future__ import annotations
 
-from typing import Any, Callable
+from collections.abc import Callable
+from typing import Any
 
 
 class Emitter:

auditgate-0.3.0/auditgate/engine.py

@@ -0,0 +1,385 @@
+# Copyright 2026 actiongate-oss
+# Licensed under the Apache License, Version 2.0;
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License in the LICENSE file at the
+# root of this repository.
+
+"""AuditGate engine for compliance-grade audit logging."""
+
+from __future__ import annotations
+
+import contextlib
+import time
+from collections.abc import Callable, Coroutine
+from functools import wraps
+from typing import Any, ParamSpec, TypeVar
+
+from .core import (
+    AuditEntry,
+    AuditPolicy,
+    Decision,
+    IntegrityMode,
+    Mode,
+    Result,
+    Severity,
+    Status,
+    StoreErrorMode,
+    Trail,
+    Verdict,
+    compute_hash,
+    wall_clock,
+)
+from .emitter import Emitter
+from .store import AsyncMemoryStore, AsyncStore, MemoryStore, Store
+
+P = ParamSpec("P")
+T = TypeVar("T")
+
+
+class AuditError(RuntimeError):
+    """Raised when audit logging fails in HARD mode."""
+    def __init__(self, decision: Decision) -> None:
+        super().__init__(decision.reason or f"Audit failure: {decision.status}")
+        self.decision = decision
+
+
+class Engine:
+    """AuditGate engine for compliance-grade audit logging."""
+
+    __slots__ = (
+        "_store", "_async_store", "_clock", "_wall_clock", "_recorded_by",
+        "_policies", "_sequences", "_emitter",
+    )
+
+    def __init__(
+        self,
+        store: Store | None = None,
+        clock: Callable[[], float] | None = None,
+        wall_clock_fn: Callable[[], str] | None = None,
+        recorded_by: str = "",
+        emitter: Emitter | None = None,
+        async_store: AsyncStore | None = None,
+    ) -> None:
+        self._store: Store = store or MemoryStore()
+        self._async_store: AsyncStore = async_store or AsyncMemoryStore()
+        self._clock = clock or time.monotonic
+        self._wall_clock = wall_clock_fn or wall_clock
+        self._recorded_by = recorded_by
+        self._policies: dict[Trail, AuditPolicy] = {}
+        self._sequences: dict[Trail, int] = {}
+        self._emitter = emitter or Emitter()
+
+    def register(self, trail: Trail, policy: AuditPolicy) -> None:
+        self._policies[trail] = policy
+
+    def policy_for(self, trail: Trail) -> AuditPolicy | None:
+        return self._policies.get(trail)
+
+    def on_decision(self, listener: Callable[[Decision], None]) -> None:
+        self._emitter.add(listener)
+
+    @property
+    def listener_errors(self) -> int:
+        return self._emitter.error_count
+
+    def record(
+        self,
+        trail: Trail,
+        *,
+        verdict: Verdict,
+        severity: Severity,
+        gate_type: str,
+        gate_identity: str,
+        reason: str | None = None,
+        detail: dict[str, Any] | None = None,
+        policy: AuditPolicy | None = None,
+    ) -> Decision:
+        """Record an audit entry for a gate decision."""
+        if policy is not None:
+            self._policies[trail] = policy
+        p = self._policies.get(trail)
+        if p is None:
+            p = AuditPolicy()
+            self._policies[trail] = p
+        if _sev_rank(severity) < _sev_rank(p.min_severity):
+            return self._decide(trail, p, Status.DROPPED,
+                                reason=f"Below min_severity ({p.min_severity.name})")
+        now = self._clock()
+        try:
+            entry = self._build_entry(trail, p, now, verdict, severity,
+                                      gate_type, gate_identity, reason, detail or {})
+        except Exception as exc:
+            return self._on_store_error(trail, p, f"Entry build failed: {exc}")
+        try:
+            self._store.append(entry)
+        except Exception as exc:
+            return self._on_store_error(trail, p, f"Store append failed: {exc}")
+        if p.retention_seconds is not None:
+            with contextlib.suppress(Exception):
+                self._store.prune(trail, now - p.retention_seconds)
+        return self._decide(trail, p, Status.RECORDED, entry=entry)
+
+    def enforce(self, decision: Decision) -> None:
+        if decision.dropped and decision.policy.mode == Mode.HARD:
+            raise AuditError(decision)
+
+    def clear(self, trail: Trail) -> None:
+        self._store.clear(trail)
+        self._sequences.pop(trail, None)
+
+    def clear_all(self) -> None:
+        self._store.clear_all()
+        self._sequences.clear()
+
+    def guard(
+        self, trail: Trail, *, policy: AuditPolicy | None = None,
+        verdict: Verdict = Verdict.ALLOW, severity: Severity = Severity.INFO,
+        gate_type: str = "auditgate", meta: dict[str, Any] | None = None,
+    ) -> Callable[[Callable[P, T]], Callable[P, T]]:
+        if policy is not None:
+            self.register(trail, policy)
+        def decorator(fn: Callable[P, T]) -> Callable[P, T]:
+            @wraps(fn)
+            def wrapper(*args: P.args, **kwargs: P.kwargs) -> T:
+                d = self._guard_record(trail, verdict, severity, gate_type, fn, meta)
+                if d.dropped:
+                    self.enforce(d)
+                return fn(*args, **kwargs)
+            return wrapper
+        return decorator
+
+    def guard_result(
+        self, trail: Trail, *, policy: AuditPolicy | None = None,
+        verdict: Verdict = Verdict.ALLOW, severity: Severity = Severity.INFO,
+        gate_type: str = "auditgate", meta: dict[str, Any] | None = None,
+    ) -> Callable[[Callable[P, T]], Callable[P, Result[T]]]:
+        if policy is not None:
+            self.register(trail, policy)
+        def decorator(fn: Callable[P, T]) -> Callable[P, Result[T]]:
+            @wraps(fn)
+            def wrapper(*args: P.args, **kwargs: P.kwargs) -> Result[T]:
+                d = self._guard_record(trail, verdict, severity, gate_type, fn, meta,
+                                       prefix="guard_result")
+                if d.dropped:
+                    return Result(decision=d)
+                return Result(decision=d, _value=fn(*args, **kwargs))
+            return wrapper
+        return decorator
+
+    # ─────────────────────────────────────────────────────────────
+    # Async API
+    # ─────────────────────────────────────────────────────────────
+
+    async def async_record(
+        self,
+        trail: Trail,
+        *,
+        verdict: Verdict,
+        severity: Severity,
+        gate_type: str,
+        gate_identity: str,
+        reason: str | None = None,
+        detail: dict[str, Any] | None = None,
+        policy: AuditPolicy | None = None,
+    ) -> Decision:
+        """Async version of record(). Uses async_store."""
+        if policy is not None:
+            self._policies[trail] = policy
+        p = self._policies.get(trail)
+        if p is None:
+            p = AuditPolicy()
+            self._policies[trail] = p
+        if _sev_rank(severity) < _sev_rank(p.min_severity):
+            return self._decide(trail, p, Status.DROPPED,
+                                reason=f"Below min_severity ({p.min_severity.name})")
+        now = self._clock()
+        try:
+            entry = await self._async_build_entry(
+                trail, p, now, verdict, severity,
+                gate_type, gate_identity, reason, detail or {},
+            )
+        except Exception as exc:
+            return self._on_store_error(trail, p, f"Entry build failed: {exc}")
+        try:
+            await self._async_store.append(entry)
+        except Exception as exc:
+            return self._on_store_error(trail, p, f"Store append failed: {exc}")
+        if p.retention_seconds is not None:
+            with contextlib.suppress(Exception):
+                await self._async_store.prune(trail, now - p.retention_seconds)
+        return self._decide(trail, p, Status.RECORDED, entry=entry)
+
+    async def async_enforce(self, decision: Decision) -> None:
+        """Async version of enforce(). Raises AuditError in HARD mode."""
+        if decision.dropped and decision.policy.mode == Mode.HARD:
+            raise AuditError(decision)
+
+    def async_guard(
+        self, trail: Trail, *, policy: AuditPolicy | None = None,
+        verdict: Verdict = Verdict.ALLOW, severity: Severity = Severity.INFO,
+        gate_type: str = "auditgate", meta: dict[str, Any] | None = None,
+    ) -> Callable[
+        [Callable[P, Coroutine[object, object, T]]],
+        Callable[P, Coroutine[object, object, T]],
+    ]:
+        """Async decorator that raises AuditError on dropped + HARD."""
+        if policy is not None:
+            self.register(trail, policy)
+
+        def decorator(
+            fn: Callable[P, Coroutine[object, object, T]],
+        ) -> Callable[P, Coroutine[object, object, T]]:
+            @wraps(fn)
+            async def wrapper(*args: P.args, **kwargs: P.kwargs) -> T:
+                d = await self._async_guard_record(
+                    trail, verdict, severity, gate_type, fn, meta,
+                )
+                if d.dropped:
+                    await self.async_enforce(d)
+                return await fn(*args, **kwargs)
+            return wrapper
+        return decorator
+
+    def async_guard_result(
+        self, trail: Trail, *, policy: AuditPolicy | None = None,
+        verdict: Verdict = Verdict.ALLOW, severity: Severity = Severity.INFO,
+        gate_type: str = "auditgate", meta: dict[str, Any] | None = None,
+    ) -> Callable[
+        [Callable[P, Coroutine[object, object, T]]],
+        Callable[P, Coroutine[object, object, Result[T]]],
+    ]:
+        """Async decorator that returns Result[T] (never raises)."""
+        if policy is not None:
+            self.register(trail, policy)
+
+        def decorator(
+            fn: Callable[P, Coroutine[object, object, T]],
+        ) -> Callable[P, Coroutine[object, object, Result[T]]]:
+            @wraps(fn)
+            async def wrapper(*args: P.args, **kwargs: P.kwargs) -> Result[T]:
+                d = await self._async_guard_record(
+                    trail, verdict, severity, gate_type, fn, meta,
+                    prefix="async_guard_result",
+                )
+                if d.dropped:
+                    return Result(decision=d)
+                return Result(decision=d, _value=await fn(*args, **kwargs))
+            return wrapper
+        return decorator
+
+    async def _async_guard_record(
+        self, trail: Trail, verdict: Verdict, severity: Severity,
+        gate_type: str, fn: Any, meta: dict[str, Any] | None,
+        prefix: str = "async_guard",
+    ) -> Decision:
+        return await self.async_record(
+            trail=trail, verdict=verdict, severity=severity,
+            gate_type=gate_type, gate_identity=str(trail),
+            reason=f"{prefix}:{fn.__qualname__}", detail=meta or {},
+        )
+
+    async def _async_build_entry(
+        self, trail: Trail, policy: AuditPolicy, now: float,
+        verdict: Verdict, severity: Severity, gate_type: str,
+        gate_identity: str, reason: str | None, detail: dict[str, Any],
+    ) -> AuditEntry:
+        seq = await self._async_next_seq(trail)
+        base = {
+            "trail": str(trail), "ts": now, "verdict": verdict.name,
+            "severity": severity.name, "gate_type": gate_type,
+            "gate_identity": gate_identity, "reason": reason,
+            "detail": detail, "sequence": seq,
+        }
+        entry_hash = None
+        prev_hash = None
+        if policy.integrity == IntegrityMode.HASH:
+            entry_hash = compute_hash(base)
+        elif policy.integrity == IntegrityMode.CHAIN:
+            prev = await self._async_store.last_entry(trail)
+            prev_hash = prev.entry_hash if prev is not None else None
+            entry_hash = compute_hash(base, prev_hash=prev_hash)
+        return AuditEntry(
+            trail=trail, ts=now, wall_ts=self._wall_clock(),
+            verdict=verdict, severity=severity, gate_type=gate_type,
+            gate_identity=gate_identity, recorded_by=self._recorded_by,
+            reason=reason, detail=detail, entry_hash=entry_hash,
+            prev_hash=prev_hash, sequence=seq,
+        )
+
+    async def _async_next_seq(self, trail: Trail) -> int:
+        if trail not in self._sequences:
+            last = await self._async_store.last_entry(trail)
+            self._sequences[trail] = (last.sequence + 1) if last is not None else 0
+        seq = self._sequences[trail]
+        self._sequences[trail] = seq + 1
+        return seq
+
+    # ─────────────────────────────────────────────────────────────
+    # Internal (sync)
+    # ─────────────────────────────────────────────────────────────
+
+    def _guard_record(
+        self, trail: Trail, verdict: Verdict, severity: Severity,
+        gate_type: str, fn: Any, meta: dict[str, Any] | None,
+        prefix: str = "guard",
+    ) -> Decision:
+        return self.record(trail=trail, verdict=verdict, severity=severity,
+                           gate_type=gate_type, gate_identity=str(trail),
+                           reason=f"{prefix}:{fn.__qualname__}", detail=meta or {})
+
+    def _build_entry(
+        self, trail: Trail, policy: AuditPolicy, now: float,
+        verdict: Verdict, severity: Severity, gate_type: str,
+        gate_identity: str, reason: str | None, detail: dict[str, Any],
+    ) -> AuditEntry:
+        seq = self._next_seq(trail)
+        base = {
+            "trail": str(trail), "ts": now, "verdict": verdict.name,
+            "severity": severity.name, "gate_type": gate_type,
+            "gate_identity": gate_identity, "reason": reason,
+            "detail": detail, "sequence": seq,
+        }
+        entry_hash = None
+        prev_hash = None
+        if policy.integrity == IntegrityMode.HASH:
+            entry_hash = compute_hash(base)
+        elif policy.integrity == IntegrityMode.CHAIN:
+            prev = self._store.last_entry(trail)
+            prev_hash = prev.entry_hash if prev is not None else None
+            entry_hash = compute_hash(base, prev_hash=prev_hash)
+        return AuditEntry(
+            trail=trail, ts=now, wall_ts=self._wall_clock(),
+            verdict=verdict, severity=severity, gate_type=gate_type,
+            gate_identity=gate_identity, recorded_by=self._recorded_by,
+            reason=reason, detail=detail, entry_hash=entry_hash,
+            prev_hash=prev_hash, sequence=seq,
+        )
+
+    def _next_seq(self, trail: Trail) -> int:
+        if trail not in self._sequences:
+            last = self._store.last_entry(trail)
+            self._sequences[trail] = (last.sequence + 1) if last is not None else 0
+        seq = self._sequences[trail]
+        self._sequences[trail] = seq + 1
+        return seq
+
+    def _on_store_error(
+        self, trail: Trail, policy: AuditPolicy, msg: str,
+    ) -> Decision:
+        decision = self._decide(trail, policy, Status.DROPPED, reason=msg)
+        if policy.on_store_error == StoreErrorMode.FAIL_CLOSED and policy.mode == Mode.HARD:
+            raise AuditError(decision)
+        return decision
+
+    def _decide(
+        self, trail: Trail, policy: AuditPolicy, status: Status,
+        entry: AuditEntry | None = None, reason: str | None = None,
+    ) -> Decision:
+        d = Decision(status=status, trail=trail, policy=policy, entry=entry, reason=reason)
+        self._emitter.emit(d)
+        return d
+
+
+def _sev_rank(severity: Severity) -> int:
+    return {Severity.DEBUG: 0, Severity.INFO: 1, Severity.WARN: 2,
+            Severity.ERROR: 3, Severity.CRITICAL: 4}[severity]