chore: fix 66 ruff violations (unblock SDK CI)

src/gradata/_core.py

@@ -989,7 +989,6 @@ def _cloud_sync_session(
     try:
         import hashlib
         import os
-
         from pathlib import Path
 
         # 1. Resolve cloud credentials: ~/.gradata/config.toml or env var

src/gradata/_events.py

@@ -414,7 +414,7 @@ class RetainOrchestrator:
     position on the next call.
     """
 
-    def __init__(self, brain_dir: "str | Path") -> None:
+    def __init__(self, brain_dir: str | Path) -> None:
         from pathlib import Path as _Path
         self.brain_dir = _Path(brain_dir)
         self.events_path = self.brain_dir / "events.jsonl"

src/gradata/cli.py

@@ -20,6 +20,7 @@
 from __future__ import annotations
 
 import argparse
+import contextlib
 import json
 import logging
 import sys
@@ -516,7 +517,7 @@ def _sanitize_toml_value(val: str) -> str:
     return val.replace("\n", "").replace("\r", "").replace("[", "").replace("]", "").replace('"', "").replace("\\", "").strip()
 
 
-def _check_config_permissions(config_path: "Path") -> None:
+def _check_config_permissions(config_path: Path) -> None:
     """Finding 4: warn if config file is world-readable (Unix only)."""
     import os
     import stat
@@ -538,17 +539,18 @@ def cmd_login(args):
     import os
     import time
     import webbrowser
+    from urllib.error import HTTPError, URLError
     from urllib.request import Request, urlopen
-    from urllib.error import URLError, HTTPError
 
     api_url = os.environ.get("GRADATA_API_URL", "https://api.gradata.ai/api/v1")
 
     # Finding 5: reject non-HTTPS API URLs (allow localhost for development)
-    if not api_url.startswith("https://"):
-        if not (api_url.startswith("http://localhost") or api_url.startswith("http://127.0.0.1")):
-            print(f"Error: GRADATA_API_URL must use HTTPS (got: {api_url})")
-            print("  HTTP is only allowed for localhost/127.0.0.1 during development.")
-            sys.exit(1)
+    if not api_url.startswith("https://") and not (
+        api_url.startswith("http://localhost") or api_url.startswith("http://127.0.0.1")
+    ):
+        print(f"Error: GRADATA_API_URL must use HTTPS (got: {api_url})")
+        print("  HTTP is only allowed for localhost/127.0.0.1 during development.")
+        sys.exit(1)
 
     # Step 1: Request a device code
     try:
@@ -614,10 +616,9 @@ def cmd_login(args):
             config_path.parent.mkdir(parents=True, exist_ok=True)
 
             # Finding 4: restrict directory permissions (Unix)
-            try:
+            # Windows — os.chmod has limited effect, suppress the error
+            with contextlib.suppress(OSError, AttributeError):
                 os.chmod(config_path.parent, 0o700)
-            except (OSError, AttributeError):
-                pass  # Windows — os.chmod has limited effect
 
             # Finding 12: sanitize values before writing TOML
             safe_key = _sanitize_toml_value(api_key)
@@ -634,16 +635,15 @@ def cmd_login(args):
             )
 
             # Finding 4: restrict file permissions (Unix)
-            try:
+            # Windows — os.chmod has limited effect, suppress the error
+            with contextlib.suppress(OSError, AttributeError):
                 os.chmod(config_path, 0o600)
-            except (OSError, AttributeError):
-                pass  # Windows — os.chmod has limited effect
 
             # Also set in the environment for the current process (so
             # subsequent brain operations in the same shell can auto-sync)
             os.environ["GRADATA_API_KEY"] = api_key
 
-            print(f"  Connected! Your brain will sync to app.gradata.ai")
+            print("  Connected! Your brain will sync to app.gradata.ai")
             print(f"  Config saved to {config_path}")
 
             # Finding 4: check permissions on startup

src/gradata/correction_detector.py

@@ -24,7 +24,7 @@
 from __future__ import annotations
 
 import re
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 from enum import StrEnum
 from typing import Any
 
@@ -149,7 +149,7 @@ def to_dict(self) -> dict[str, Any]:
         }
 
     @classmethod
-    def from_dict(cls, data: dict[str, Any]) -> "StructuredCorrection":
+    def from_dict(cls, data: dict[str, Any]) -> StructuredCorrection:
         """Deserialize from a plain dict."""
         return cls(
             what_wrong=data.get("what_wrong", ""),
@@ -416,7 +416,6 @@ def extract_structured_correction(
             # one/both of draft/final are empty. Nothing to extract.
             return None
 
-    correction_text = final or context
     full_text = " ".join(filter(None, [context, final]))
 
     what_wrong = _extract_what_wrong(draft, final)

src/gradata/enhancements/bandits/collaborative_filter.py

@@ -34,7 +34,7 @@
 
 import hashlib
 import math
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 
 
 @dataclass
@@ -61,7 +61,7 @@ class BrainFingerprint:
     category_distribution: dict[str, int]   # category -> rule count
 
     @classmethod
-    def from_lessons(cls, lessons: list, domain: str = "", total_sessions: int = 0) -> "BrainFingerprint":
+    def from_lessons(cls, lessons: list, domain: str = "", total_sessions: int = 0) -> BrainFingerprint:
         """Build a fingerprint from a list of Lesson objects."""
 
         rules = []
@@ -122,7 +122,7 @@ def compute_brain_similarity(a: BrainFingerprint, b: BrainFingerprint) -> float:
     vec_b = [b.category_distribution.get(c, 0) for c in sorted(all_cats)]
 
     # Cosine similarity
-    dot = sum(x * y for x, y in zip(vec_a, vec_b))
+    dot = sum(x * y for x, y in zip(vec_a, vec_b, strict=True))
     mag_a = math.sqrt(sum(x * x for x in vec_a))
     mag_b = math.sqrt(sum(x * x for x in vec_b))
 

src/gradata/enhancements/carl.py

@@ -1,6 +1,6 @@
 """Backward-compatibility shim — all code moved to behavioral_engine.py."""
 
-from gradata.enhancements.behavioral_engine import (  # noqa: F401
+from gradata.enhancements.behavioral_engine import (
     BehavioralContract,
     ConstraintViolation,
     ContractRegistry,

src/gradata/enhancements/clustering.py

@@ -156,7 +156,7 @@ def promote_instinct_clusters(
         by_category[lesson.category].append(lesson)
 
     promoted: list[str] = []
-    for category, members in by_category.items():
+    for _category, members in by_category.items():
         if len(members) < min_cluster_size:
             continue
 

src/gradata/enhancements/graduation/agent_graduation.py

@@ -48,26 +48,25 @@
 import json
 import re
 from dataclasses import dataclass, field
-from datetime import datetime, timezone
+from datetime import UTC, datetime
 from pathlib import Path
 
 # Import graduation constants from self_improvement (same research-backed values)
 from gradata.enhancements.self_improvement import (
-    SURVIVAL_BONUS,
     ACCEPTANCE_BONUS,
+    MIN_APPLICATIONS_FOR_PATTERN,
+    MIN_APPLICATIONS_FOR_RULE,
     MISFIRE_PENALTY,
     PATTERN_THRESHOLD,
     RULE_THRESHOLD,
-    MIN_APPLICATIONS_FOR_PATTERN,
-    MIN_APPLICATIONS_FOR_RULE,
-    LessonState,
+    SURVIVAL_BONUS,
     Lesson,
-    parse_lessons,
+    LessonState,
     format_lessons,
     get_maturity_phase,
+    parse_lessons,
 )
 
-
 # ---------------------------------------------------------------------------
 # Approval Gate Thresholds
 # ---------------------------------------------------------------------------
@@ -250,7 +249,7 @@ def compile_deterministic_rule(lesson: Lesson) -> DeterministicRule | None:
 
 
 def _now() -> str:
-    return datetime.now(timezone.utc).isoformat()
+    return datetime.now(UTC).isoformat()
 
 
 # ---------------------------------------------------------------------------
@@ -450,7 +449,7 @@ def _extract_agent_lesson(
         category = "AGENT_" + profile.agent_type.upper()
         if edit_category:
             category = edit_category.upper()
-        today = datetime.now(timezone.utc).strftime("%Y-%m-%d")
+        today = datetime.now(UTC).strftime("%Y-%m-%d")
 
         # Check for duplicate (same edit pattern already captured)
         edit_lower = edits.lower()[:100]
@@ -773,7 +772,7 @@ def compute_quality_scores(self) -> dict:
             "best_agent": best,
         }
 
-    def get_deterministic_rules(self, agent_type: str, task_type: str = "") -> list["DeterministicRule"]:
+    def get_deterministic_rules(self, agent_type: str, task_type: str = "") -> list[DeterministicRule]:
         """Get RULE-tier lessons compiled into enforceable guard logic.
 
         Only RULE-tier lessons with an enforceable pattern are returned.
@@ -820,7 +819,7 @@ def get_deterministic_rules(self, agent_type: str, task_type: str = "") -> list[
 
         return rules
 
-    def enforce_rules(self, agent_type: str, output: str, task_type: str = "") -> "EnforcementResult":
+    def enforce_rules(self, agent_type: str, output: str, task_type: str = "") -> EnforcementResult:
         """Apply all deterministic RULE-tier guards to agent output.
 
         Returns an EnforcementResult with pass/fail and details of any violations.

src/gradata/enhancements/graduation/judgment_decay.py

@@ -35,12 +35,11 @@
 from typing import Any
 
 from gradata.enhancements.self_improvement import (
-    LessonState,
-    Lesson,
     PATTERN_THRESHOLD,
+    Lesson,
+    LessonState,
 )
 
-
 # Decay constants
 DECAY_PER_IDLE_SESSION = 0.02
 CONFIDENCE_FLOOR = 0.10

src/gradata/enhancements/meta_rules.py

@@ -681,7 +681,7 @@ def _try_llm_principle(rules: list[Lesson], category: str) -> str | None:
                 api_base=b,
                 model=os.environ.get("GRADATA_LLM_MODEL", "gpt-4o-mini"),
             )
-        except Exception as exc:  # noqa: BLE001 -- degrade to deterministic
+        except Exception as exc:
             _log.debug("OpenAI-compat synthesis failed for %s: %s", category, exc)
             return None
 

src/gradata/enhancements/profiling/tone_profile.py

@@ -33,7 +33,6 @@
 from dataclasses import dataclass, field
 from typing import Any
 
-
 # ---------------------------------------------------------------------------
 # Tone Feature Extraction
 # ---------------------------------------------------------------------------
@@ -118,7 +117,7 @@ def to_dict(self) -> dict[str, Any]:
         }
 
     @classmethod
-    def from_dict(cls, d: dict[str, Any]) -> "ToneFeatures":
+    def from_dict(cls, d: dict[str, Any]) -> ToneFeatures:
         """Deserialize from dict."""
         return cls(**{k: v for k, v in d.items() if k in cls.__dataclass_fields__})
 
@@ -250,7 +249,7 @@ def to_dict(self) -> dict[str, Any]:
         }
 
     @classmethod
-    def from_dict(cls, d: dict[str, Any]) -> "ToneProfile":
+    def from_dict(cls, d: dict[str, Any]) -> ToneProfile:
         features = ToneFeatures.from_dict(d.get("features", {}))
         return cls(
             task_type=d["task_type"],

src/gradata/enhancements/rule_pipeline.py

@@ -187,7 +187,7 @@ def _generate_skill_file(
     return skill_path
 
 
-def review_generated_skill(skill_path: "Path") -> dict:
+def review_generated_skill(skill_path: Path) -> dict:
     """Review a generated skill file for quality issues.
 
     Returns dict with:
@@ -370,8 +370,9 @@ def run_rule_pipeline(
                 )
                 if already_exists:
                     continue
-                from gradata._types import Lesson as _Lesson
                 from datetime import date as _date
+
+                from gradata._types import Lesson as _Lesson
                 candidate = _Lesson(
                     date=_date.today().isoformat(),
                     state=LessonState.INSTINCT,
@@ -404,18 +405,26 @@ def run_rule_pipeline(
     # ── Phase 2: Atomic writes ────────────────────────────────────────────────
     # Graduate rules, update confidence, create meta-rules.
     for lesson in all_lessons:
-        if lesson.state.name == "INSTINCT" and lesson.confidence >= PATTERN_THRESHOLD:
-            if lesson.fire_count >= MIN_APPLICATIONS_FOR_PATTERN:
-                lesson.state = LessonState.PATTERN
-                result.graduated.append(f"{lesson.category}:{lesson.description[:30]}")
-        elif lesson.state.name == "PATTERN" and lesson.confidence >= RULE_THRESHOLD:
-            if lesson.fire_count >= MIN_APPLICATIONS_FOR_RULE:
-                lesson.state = LessonState.RULE
-                result.graduated.append(f"{lesson.category}:{lesson.description[:30]}")
+        if (
+            lesson.state.name == "INSTINCT"
+            and lesson.confidence >= PATTERN_THRESHOLD
+            and lesson.fire_count >= MIN_APPLICATIONS_FOR_PATTERN
+        ):
+            lesson.state = LessonState.PATTERN
+            result.graduated.append(f"{lesson.category}:{lesson.description[:30]}")
+        elif (
+            lesson.state.name == "PATTERN"
+            and lesson.confidence >= RULE_THRESHOLD
+            and lesson.fire_count >= MIN_APPLICATIONS_FOR_RULE
+        ):
+            lesson.state = LessonState.RULE
+            result.graduated.append(f"{lesson.category}:{lesson.description[:30]}")
 
     # Synthesize meta-rules from graduated rules
     try:
-        from gradata.enhancements.meta_rules import synthesize_meta_rules_agentic  # type: ignore[import]
+        from gradata.enhancements.meta_rules import (
+            synthesize_meta_rules_agentic,  # type: ignore[import]
+        )
         from gradata.enhancements.meta_rules_storage import (  # type: ignore[import]
             load_meta_rules,
             save_meta_rules,
@@ -470,7 +479,9 @@ def run_rule_pipeline(
 
     # Disposition updates from this session's corrections
     try:
-        from gradata.enhancements.behavioral_engine import DispositionTracker  # type: ignore[import]
+        from gradata.enhancements.behavioral_engine import (
+            DispositionTracker,  # type: ignore[import]
+        )
 
         tracker = DispositionTracker()
         disp_path = lessons_path.parent / "disposition.json"
@@ -613,7 +624,9 @@ def build_knowledge_graph(lessons_path: Path, db_path: Path) -> dict:
 
     # Cross-domain candidates
     try:
-        from gradata.enhancements.meta_rules import detect_cross_domain_candidates  # type: ignore[import]
+        from gradata.enhancements.meta_rules import (
+            detect_cross_domain_candidates,  # type: ignore[import]
+        )
         graph["cross_domain"] = detect_cross_domain_candidates(lessons)
     except (ImportError, Exception):
         pass

src/gradata/enhancements/scoring/brain_scores.py

@@ -21,11 +21,6 @@
 import sqlite3
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    pass
-
 
 # ---------------------------------------------------------------------------
 # Data model
@@ -233,7 +228,7 @@ def compute_brain_scores(
     """
     # Primary path: delegate to the authoritative implementation
     try:
-        import gradata._events as _events  # noqa: PLC0415
+        import gradata._events as _events
 
         _fn = getattr(_events, "compute_brain_scores", None)
         if _fn is None:

src/gradata/enhancements/scoring/calibration.py

@@ -28,8 +28,7 @@
 
 from __future__ import annotations
 
-import math
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 
 
 @dataclass