Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(dead-code): align default min_confidence floor to RISK_CAP_CONFIDENCE (#1084) (#1087)

Akshat malik151491dbbc82
diff --git a/docs/agent/MCP_TOOLS.md b/docs/agent/MCP_TOOLS.md
index 093cd4cd4..e26c15dae 100644
--- a/docs/agent/MCP_TOOLS.md
+++ b/docs/agent/MCP_TOOLS.md
@@ -441,7 +441,7 @@ Unreachable code, unused exports, unused internals, and zombie packages, sorted
|-----------|------|----------|-------------|
| `repo` | string | No | *(workspace only)* Target repo alias |
| `kind` | string | No | Restrict to one finding kind: `unreachable_file` \| `unused_export` \| `unused_internal` \| `zombie_package` |
-| `min_confidence` | float | No | Minimum confidence floor (default `0.5`; `0.7`+ is cleanup-ready only) |
+| `min_confidence` | float | No | Minimum confidence floor (default `0.4`; `0.7`+ is cleanup-ready only) |
| `safe_only` | boolean | No | Deletion-ready findings only, excluding anything with runtime-load risk (default `false`) |
| `limit` | int | No | Max findings per tier, clamped to 25 (default 20) |
| `tier` | string | No | Restrict to one tier: `high` (>= 0.8) \| `medium` \| `low` |
diff --git a/docs/architecture/deep-dives.md b/docs/architecture/deep-dives.md
index 20511f7b5..0ac018007 100644
--- a/docs/architecture/deep-dives.md
+++ b/docs/architecture/deep-dives.md
@@ -254,7 +254,7 @@ DeadCodeAnalyzer.analyze()
├── _detect_unused_exports() → findings with confidence + safe_to_delete
├── _detect_zombie_packages() → findings at 0.5 confidence, never safe
│
- ▼ apply min_confidence filter (default 0.5)
+ ▼ apply min_confidence filter (default 0.4)
│
▼ persist to dead_code_findings table
│
diff --git a/docs/layers/DEAD_CODE.md b/docs/layers/DEAD_CODE.md
index a45a1e9c5..3555644ef 100644
--- a/docs/layers/DEAD_CODE.md
+++ b/docs/layers/DEAD_CODE.md
@@ -101,10 +101,10 @@ compare numbers:
| Surface | High | Medium | Low | Default floor |
|---------|------|--------|-----|---------------|
-| CLI (`repowise dead-code`) | `>= 0.7` | `0.5` to `0.7` | `< 0.5` | `--min-confidence 0.5` |
-| MCP (`get_dead_code`) | `>= 0.8` | `0.5` to `0.8` | `< 0.5` | `min_confidence=0.5` |
+| CLI (`repowise dead-code`) | `>= 0.7` | `0.4` to `0.7` | `< 0.4` | `--min-confidence 0.4` |
+| MCP (`get_dead_code`) | `>= 0.8` | `0.4` to `0.8` | `< 0.4` | `min_confidence=0.4` |
-Both surfaces now share the same default floor (`0.5`). The MCP high/medium
+Both surfaces now share the same default floor (`0.4`). The MCP high/medium
cutovers stay stricter: an agent acting on a finding is riskier than a human
reading a table.
@@ -230,7 +230,7 @@ before deleting anything.
| Flag | Description |
|------|-------------|
-| `--min-confidence` | Minimum confidence threshold (default `0.5`) |
+| `--min-confidence` | Minimum confidence threshold (default `0.4`) |
| `--safe-only` | Only findings marked safe to delete |
| `--kind` | `unreachable_file`, `unused_export`, `unused_internal`, `zombie_package` |
| `--format` | `table` (default), `json`, `md` |
@@ -251,7 +251,7 @@ path, kind, confidence, line count, and a cleanup impact estimate.
| Parameter | Default | Notes |
|-----------|---------|-------|
| `kind` | all | One of the four finding kinds |
-| `min_confidence` | `0.5` | `0.7` and above is cleanup-ready only |
+| `min_confidence` | `0.4` | `0.7` and above is cleanup-ready only |
| `tier` | all | `high` (`>= 0.8`), `medium`, `low` |
| `safe_only` | `false` | Deletion-ready only, excluding runtime-load risk |
| `limit` | `20` | Per tier, clamped to 25 |
diff --git a/docs/reference/CLI_REFERENCE.md b/docs/reference/CLI_REFERENCE.md
index 42d81b889..b652edaed 100644
--- a/docs/reference/CLI_REFERENCE.md
+++ b/docs/reference/CLI_REFERENCE.md
@@ -449,7 +449,7 @@ Detect dead and unused code.
| Flag | Description |
|------|-------------|
-| `--min-confidence` | Minimum confidence threshold (default: 0.5) |
+| `--min-confidence` | Minimum confidence threshold (default: 0.4) |
| `--safe-only` | Only show findings marked safe to delete |
| `--kind` | Filter: `unreachable_file`, `unused_export`, `unused_internal`, `zombie_package` |
| `--format` | Output: `table` (default), `json`, `md` |
diff --git a/packages/cli/src/repowise/cli/commands/dead_code_cmd.py b/packages/cli/src/repowise/cli/commands/dead_code_cmd.py
index 175ff7387..a961e990f 100644
--- a/packages/cli/src/repowise/cli/commands/dead_code_cmd.py
+++ b/packages/cli/src/repowise/cli/commands/dead_code_cmd.py
@@ -14,11 +14,17 @@ from repowise.cli.helpers import (
run_async,
silence_logs_for_machine_output,
)
+from repowise.core.analysis.dead_code.risk_factors import RISK_CAP_CONFIDENCE
@click.command("dead-code")
@click.argument("path", required=False, type=click.Path(exists=True))
[email protected]("--min-confidence", default=0.5, type=float, help="Minimum confidence threshold.")
+ "--min-confidence",
+ default=RISK_CAP_CONFIDENCE,
+ type=float,
+ help="Minimum confidence threshold.",
+)
@click.option(
"--safe-only",
is_flag=True,
@@ -117,9 +123,7 @@ def dead_code_command(
if primary is None:
raise click.ClickException("Workspace has no primary repo configured.")
repo_path = primary
- console.print(
- "[dim] (Tip: pass --repo <alias> to analyze a different repo.)[/dim]"
- )
+ console.print("[dim] (Tip: pass --repo <alias> to analyze a different repo.)[/dim]")
else:
assert target.repo_path is not None
repo_path = target.repo_path
diff --git a/packages/core/src/repowise/core/analysis/dead_code/risk_factors.py b/packages/core/src/repowise/core/analysis/dead_code/risk_factors.py
index 898784cb0..df4b2a534 100644
--- a/packages/core/src/repowise/core/analysis/dead_code/risk_factors.py
+++ b/packages/core/src/repowise/core/analysis/dead_code/risk_factors.py
@@ -34,9 +34,10 @@ from pathlib import PurePosixPath
SAFE_CONFIDENCE_THRESHOLD: float = 0.7
# Confidence ceiling applied to a finding that carries a runtime-load risk
-# factor. 0.4 is the default ``min_confidence`` floor, so the finding still
-# surfaces (as a medium / review-required candidate) but never reads as
-# deletion-ready.
+# factor. 0.4 is the default ``min_confidence`` floor across CLI, REST router,
+# and MCP tools (which import this value as their single source of truth), so the
+# finding still surfaces (as a medium / review-required candidate) but never
+# reads as deletion-ready.
RISK_CAP_CONFIDENCE: float = 0.4
# Filename-stem tokens → risk-factor tag. Matched against the basename split on
diff --git a/packages/server/src/repowise/server/mcp_server/tool_dead_code.py b/packages/server/src/repowise/server/mcp_server/tool_dead_code.py
index 653b4a931..da43b4d7c 100644
--- a/packages/server/src/repowise/server/mcp_server/tool_dead_code.py
+++ b/packages/server/src/repowise/server/mcp_server/tool_dead_code.py
@@ -10,6 +10,7 @@ from typing import Any
from sqlalchemy import select
from repowise.core.analysis.dead_code.risk_factors import (
+ RISK_CAP_CONFIDENCE,
effective_safe_to_delete,
path_risk_factors,
)
@@ -215,7 +216,7 @@ _TIER_DESC_LOW = (
async def get_dead_code(
repo: str | None = None,
kind: str | None = None,
- min_confidence: float = 0.5,
+ min_confidence: float = RISK_CAP_CONFIDENCE,
safe_only: bool = False,
limit: int = 20,
tier: str | None = None,
@@ -236,7 +237,7 @@ async def get_dead_code(
Args:
repo: usually omitted.
kind: unreachable_file | unused_export | unused_internal | zombie_package.
- min_confidence: floor, default 0.5 (0.7 = cleanup-ready only).
+ min_confidence: floor, default 0.4 (0.7 = cleanup-ready only).
safe_only: deletion-ready findings only (no runtime-load risk).
limit: max findings per tier (clamped to 25).
tier: "high" (>=0.8) | "medium" | "low".
diff --git a/packages/server/src/repowise/server/routers/dead_code.py b/packages/server/src/repowise/server/routers/dead_code.py
index a2b495f41..9324d74c4 100644
--- a/packages/server/src/repowise/server/routers/dead_code.py
+++ b/packages/server/src/repowise/server/routers/dead_code.py
@@ -5,7 +5,10 @@ from __future__ import annotations
from fastapi import APIRouter, Depends, HTTPException, Query, Request
from sqlalchemy.ext.asyncio import AsyncSession
-from repowise.core.analysis.dead_code.risk_factors import effective_safe_to_delete
+from repowise.core.analysis.dead_code.risk_factors import (
+ RISK_CAP_CONFIDENCE,
+ effective_safe_to_delete,
+)
from repowise.core.persistence import crud
from repowise.server.deps import get_db_session, verify_api_key
from repowise.server.schemas import (
@@ -27,7 +30,7 @@ router = APIRouter(
async def list_dead_code(
repo_id: str,
kind: str | None = Query(None, description="Filter by finding kind"),
- min_confidence: float = Query(0.5, ge=0.0, le=1.0),
+ min_confidence: float = Query(RISK_CAP_CONFIDENCE, ge=0.0, le=1.0),
status: str = Query("open"),
safe_only: bool = Query(False),
limit: int = Query(100, ge=1, le=500),
diff --git a/packages/types/src/dead-code.ts b/packages/types/src/dead-code.ts
index 7552538b5..a49ed8b37 100644
--- a/packages/types/src/dead-code.ts
+++ b/packages/types/src/dead-code.ts
@@ -26,7 +26,11 @@ export type DeadCodeStatus = "open" | "acknowledged" | "resolved" | "false_posit
export const DEAD_CODE_CONFIDENCE = {
/** Deletion-ready floor; matches SAFE_CONFIDENCE_THRESHOLD. */
HIGH: 0.7,
- /** The list endpoint's default floor; below this findings are not fetched. */
+ /**
+ * The list endpoint & CLI's default `min_confidence` floor (0.4); mirrors
+ * `RISK_CAP_CONFIDENCE` in `core/analysis/dead_code/risk_factors.py`. Below this
+ * findings are not fetched.
+ */
MEDIUM: 0.4,
} as const;
diff --git a/packages/ui/src/dead-code/findings-table.tsx b/packages/ui/src/dead-code/findings-table.tsx
index 9ee08e380..9988d2384 100644
--- a/packages/ui/src/dead-code/findings-table.tsx
+++ b/packages/ui/src/dead-code/findings-table.tsx
@@ -117,8 +117,8 @@ export function FindingsTable({
isLoading,
}: FindingsTableProps) {
const [activeTab, setActiveTab] = useState<string | null>(null);
- // The slider floor is the server's own `min_confidence` default: nothing
- // below it is ever fetched, so a lower floor would be a control over nothing.
+ // The slider floor is the server's own `min_confidence` default (0.4, derived from
+ // RISK_CAP_CONFIDENCE in risk_factors.py): nothing below it is ever fetched.
const [minConfidence, setMinConfidence] = useState<number>(DEAD_CODE_CONFIDENCE.MEDIUM);
const [safeOnly, setSafeOnly] = useState(false);
const [query, setQuery] = useState("");
diff --git a/plugins/claude-code/commands/dead-code.md b/plugins/claude-code/commands/dead-code.md
index c2fab1269..1e0fdb0c2 100644
--- a/plugins/claude-code/commands/dead-code.md
+++ b/plugins/claude-code/commands/dead-code.md
@@ -23,7 +23,7 @@ should be proposed for removal.
- "files" → `--kind unreachable_file`
- "exports" → `--kind unused_export`
- "packages" / "zombie" → `--kind zombie_package`
-- "strict" → `--min-confidence 0.7` (default is 0.5)
+- "strict" → `--min-confidence 0.7` (default is 0.4)
Other flags: `--format json`, `--include-internals` (aggressive, private-symbol
scan; higher false-positive rate), `--include-zombie-packages` /
diff --git a/tests/unit/analysis/test_ts_never_flag.py b/tests/unit/analysis/test_ts_never_flag.py
index 010fc910c..c17bb9381 100644
--- a/tests/unit/analysis/test_ts_never_flag.py
+++ b/tests/unit/analysis/test_ts_never_flag.py
@@ -14,6 +14,7 @@ import fnmatch
def _matches(path: str) -> bool:
from repowise.core.analysis.dead_code.constants import _NEVER_FLAG_PATTERNS
+
return any(fnmatch.fnmatch(path, p) for p in _NEVER_FLAG_PATTERNS)
@@ -92,6 +93,7 @@ class TestNeverFlagDoesNotLeak:
class TestEntryPointSymbolNames:
def test_nextjs_route_exports_included(self):
from repowise.core.analysis.dead_code.analyzer import _ENTRY_POINT_SYMBOL_NAMES
+
for name in (
"generateStaticParams",
"generateMetadata",
@@ -103,6 +105,7 @@ class TestEntryPointSymbolNames:
def test_remix_route_exports_included(self):
from repowise.core.analysis.dead_code.analyzer import _ENTRY_POINT_SYMBOL_NAMES
+
for name in ("ErrorBoundary", "HydrateFallback", "clientLoader", "shouldRevalidate"):
assert name in _ENTRY_POINT_SYMBOL_NAMES
@@ -110,6 +113,7 @@ class TestEntryPointSymbolNames:
# These names are too common to blanket-exempt — they'd mask
# genuine dead code in non-route files across every language.
from repowise.core.analysis.dead_code.analyzer import _ENTRY_POINT_SYMBOL_NAMES
+
for name in ("load", "action", "loader", "meta", "metadata", "config", "headers"):
assert name not in _ENTRY_POINT_SYMBOL_NAMES
@@ -117,6 +121,7 @@ class TestEntryPointSymbolNames:
class TestTsUnusedExportConstants:
def test_non_importable_kinds_allows_ts_const_and_var(self):
from repowise.core.analysis.dead_code.analyzer import _non_importable_kinds
+
kinds = _non_importable_kinds("typescript")
assert "constant" not in kinds
assert "variable" not in kinds
@@ -125,13 +130,14 @@ class TestTsUnusedExportConstants:
def test_non_importable_kinds_allows_js_const_and_var(self):
from repowise.core.analysis.dead_code.analyzer import _non_importable_kinds
+
kinds = _non_importable_kinds("javascript")
assert "constant" not in kinds
assert "variable" not in kinds
def test_other_languages_retain_universal_non_importable(self):
from repowise.core.analysis.dead_code.analyzer import _non_importable_kinds
+
python_kinds = _non_importable_kinds("python")
assert "constant" in python_kinds
assert "variable" in python_kinds
-
diff --git a/tests/unit/dead_code/test_confidence_parity.py b/tests/unit/dead_code/test_confidence_parity.py
new file mode 100644
index 000000000..9ce98c766
--- /dev/null
+++ b/tests/unit/dead_code/test_confidence_parity.py
@@ -0,0 +1,55 @@
+"""Cross-language contract test for dead-code confidence threshold alignment.
+
+Verifies that TypeScript's DEAD_CODE_CONFIDENCE.MEDIUM in packages/types/src/dead-code.ts
+matches Python's RISK_CAP_CONFIDENCE in repowise.core.analysis.dead_code.risk_factors.
+"""
+
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+from repowise.core.analysis.dead_code.risk_factors import (
+ RISK_CAP_CONFIDENCE,
+ SAFE_CONFIDENCE_THRESHOLD,
+)
+
+
+def test_ts_confidence_floor_matches_python_risk_cap() -> None:
+ """Ensure DEAD_CODE_CONFIDENCE.MEDIUM in TypeScript matches RISK_CAP_CONFIDENCE (0.4)."""
+ repo_root = Path(__file__).parents[3]
+ ts_file = repo_root / "packages" / "types" / "src" / "dead-code.ts"
+ assert ts_file.exists(), f"TypeScript types file not found at {ts_file}"
+
+ content = ts_file.read_text(encoding="utf-8")
+
+ medium_match = re.search(
+ r"DEAD_CODE_CONFIDENCE\s*=\s*\{[\s\S]*?MEDIUM:\s*([0-9.]+)", content
+ )
+ assert medium_match is not None, "Could not parse DEAD_CODE_CONFIDENCE.MEDIUM from dead-code.ts"
+ ts_medium = float(medium_match.group(1))
+
+ assert ts_medium == RISK_CAP_CONFIDENCE, (
+ f"Cross-language drift detected! TypeScript DEAD_CODE_CONFIDENCE.MEDIUM is {ts_medium}, "
+ f"but Python RISK_CAP_CONFIDENCE is {RISK_CAP_CONFIDENCE}."
+ )
+
+
+def test_ts_confidence_high_matches_python_safe_threshold() -> None:
+ """Ensure DEAD_CODE_CONFIDENCE.HIGH in TypeScript matches SAFE_CONFIDENCE_THRESHOLD (0.7)."""
+ repo_root = Path(__file__).parents[3]
+ ts_file = repo_root / "packages" / "types" / "src" / "dead-code.ts"
+ assert ts_file.exists(), f"TypeScript types file not found at {ts_file}"
+
+ content = ts_file.read_text(encoding="utf-8")
+
+ high_match = re.search(
+ r"DEAD_CODE_CONFIDENCE\s*=\s*\{[\s\S]*?HIGH:\s*([0-9.]+)", content
+ )
+ assert high_match is not None, "Could not parse DEAD_CODE_CONFIDENCE.HIGH from dead-code.ts"
+ ts_high = float(high_match.group(1))
+
+ assert ts_high == SAFE_CONFIDENCE_THRESHOLD, (
+ f"Cross-language drift detected! TypeScript DEAD_CODE_CONFIDENCE.HIGH is {ts_high}, "
+ f"but Python SAFE_CONFIDENCE_THRESHOLD is {SAFE_CONFIDENCE_THRESHOLD}."
+ )