mostlygeek/llama-swap

2.4

Adjusted Score

2.4

Raw Score

100%

Time Factor

2026-07-14

Last Push

5.0K

Stars

Language

56.3K

Lines of Code

400

Files

Pattern Hits

2026-07-14

Scan Date

0.01

HC Hit Rate

What These Metrics Mean

Adjusted Score: Primary synthetic code indicator. Raw score normalised per 1,000 lines of code and multiplied by the temporal discount factor. This is the definitive comparative metric — use it to rank repositories by AI authorship density.
Raw Score: The unmodified sum of all severity-weighted, context-multiplied pattern match scores before temporal discounting. Reflects the absolute signal strength independent of when the repository was last active.
Time Factor: The temporal discount multiplier (0–100%) applied to the raw score. Repositories last updated before ChatGPT's launch (Nov 2022) receive a 5% factor. Full signal is only assigned to repositories active in the post-adoption era (Jan 2024+).
Pattern Hits: Total count of individual pattern matches across all files and categories. A high hit count with a low score may indicate a very large codebase with isolated AI snippets; a low count with a high score indicates dense, concentrated AI signatures.
HC Hit Rate: High+Critical pattern hits per file, averaged across the repository. This orthogonal signal catches repositories where a few files are densely packed with high-severity AI tells — a strong indicator even when the normalised score appears moderate due to codebase size.
Lines of Code / Files: Total lines and files analysed. The scanner examines 94 file extensions. These denominators are used to normalise the score, enabling fair comparison between repositories of vastly different sizes.

Score History

This chart maps the temporal evolution of the adjusted synthetic code score across successive scan runs. An upward trajectory indicates ongoing incorporation of AI-generated code or expanding LLM-assisted scaffolding; a stable or declining trajectory may reflect active human refactoring, code removal, or the adoption of stricter authorship policies. The dashed secondary line (right axis) independently tracks total raw pattern hit count, which can diverge from the normalised score when codebase size changes significantly between scans.

Severity Breakdown

Classifies detected patterns by their diagnostic confidence and structural impact. CRITICAL patterns (coefficient 10) represent definitive synthetic signatures — hallucinated imports, explicit LLM attribution metadata — virtually never produced by human authors. HIGH (5) indicates strong structural tells such as cross-file repetition or cross-linguistic idioms. MEDIUM (2) covers recognisable conversational padding and AI-specific vocabulary. LOW (1) captures subtle indicators like tautological comments and generic boilerplate that require density to carry independent signal.

CRITICAL 0HIGH 3MEDIUM 14LOW 69

Directory Score Breakdown

This horizontal bar chart decomposes the repository's raw synthetic code score by top-level directory, allowing you to pinpoint precisely which modules or components carry the highest AI authorship density. Directories with disproportionately high scores relative to their size warrant targeted manual review: concentrated AI signatures often trace back to mass-generated configuration layers, auto-ported test suites, LLM-scaffolded boilerplate classes, or entire subsystems authored under heavy copilot assistance. Use this view to prioritise your human code-review effort.

Pattern Findings

The scanner identified 86 distinct pattern matches across 10 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

Reading the findings table: The Severity column indicates the diagnostic confidence level (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — comment-context matches receive a ×1.5 weight because LLMs systematically over-annotate. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier as dense clusters constitute far stronger evidence of synthetic authorship than isolated hits.

Over-Commented Block59 hits · 52 pts

Severity	File	Line	Snippet	Context
LOW	config.example.yaml	1	# add this modeline for validation in vscode	COMMENT
LOW	config.example.yaml	21	# - minimum value is 15 seconds, anything less will be set to this value	COMMENT
LOW	config.example.yaml	41	# handling requests, etc.	COMMENT
LOW	config.example.yaml	61	# captureBuffer: how many MBs to allocate for storing request/response captures	COMMENT
LOW	config.example.yaml	81	# - it is automatically incremented for every model that uses it	COMMENT
LOW	config.example.yaml	101	globalTTL: 0	COMMENT
LOW	config.example.yaml	141	# use environment variable macros to keep secrets out of the config	COMMENT
LOW	config.example.yaml	161	# - the model's ID is available in the ${MODEL_ID} macro, also available in macros defined above	COMMENT
LOW	config.example.yaml	181	${latest-llama}	COMMENT
LOW	config.example.yaml	201	# - in the format: ENV_NAME=value	COMMENT
LOW	config.example.yaml	221	# - a ttl of -1 will use the global TTL value as the default	COMMENT
LOW	config.example.yaml	241	# - recommended to stick to sampling parameters	COMMENT
LOW	config.example.yaml	301	b: 2	COMMENT
LOW	config.example.yaml	321	# - useful for models running on slower hardware that need longer timeouts	COMMENT
LOW	config.example.yaml	361	# context: the maximum token context length supported	COMMENT
LOW	config.example.yaml	381	docker run --name ${MODEL_ID}	COMMENT
LOW	config.example.yaml	401	# on_startup: a dictionary of actions to perform on startup	COMMENT
LOW	config.example.yaml	421	# - matrix: the newer engine. More involved to configure, but far more	COMMENT
LOW	config.example.yaml	441	# - behaviour is set per group with the `swap`, `exclusive` and	COMMENT
LOW	config.example.yaml	501	# costliest ones loaded.	COMMENT
LOW	config.example.yaml	541	# +ref inline the expression of another set	COMMENT
LOW	config.example.yaml	621	keepalive: 30	COMMENT
LOW	docker/build-image.sh	1	#!/bin/bash	COMMENT
LOW	docker/unified/build-image.sh	1	#!/bin/bash	COMMENT
LOW	internal/server/log.go	21	// wiring each one's output per the logToStdout config value. The proxy and	COMMENT
LOW	internal/chain/chain.go	1	// Package chain composes http.Handler middleware into a single handler.	COMMENT
LOW	internal/perf/d3dkmt_windows.go	141	// sizeof(D3DKMT_QUERYSTATISTICS) == 0x328 (808 bytes) on x64.	COMMENT
LOW	internal/process/runtime_windows.go	21	// rooted at cmd.Process. Windows has no SIGTERM or process-group signalling, so	COMMENT
LOW	internal/process/process_command.go	21		COMMENT
LOW	internal/process/process_command.go	561	// immediately, which force-kills the process WaitDelay after the signal	COMMENT
LOW	internal/process/process_command_test.go	321	}	COMMENT
LOW	internal/process/process_command_forking_test.go	21	// `"$@"`). After SIGTERM the wrapper dies but the grandchild inherits the	COMMENT
LOW	internal/router/base.go	141	}	COMMENT
LOW	internal/router/base.go	361	}	COMMENT
LOW	internal/router/scheduler/fifo.go	81	//	COMMENT
LOW	internal/router/scheduler/scheduler.go	21	"github.com/mostlygeek/llama-swap/internal/shared"	COMMENT
LOW	internal/router/scheduler/scheduler.go	41	// set EvictionFor was given for this decision. Planners may log their	COMMENT
LOW	internal/router/scheduler/scheduler.go	61	// OnUnload reconciles scheduler state for an unload, stops the targeted	COMMENT
LOW	docs/configuration.md	101	# copying this file into an LLM and asking it to explain or generate	COMMENT
LOW	docs/configuration.md	121	# - optional, default (disabled): ""	COMMENT
LOW	docs/configuration.md	141	# - useful for limiting memory usage when processing large volumes of metrics	COMMENT
LOW	docs/configuration.md	161	enable: true	COMMENT
LOW	docs/configuration.md	181		COMMENT
LOW	docs/configuration.md	201	# - macro values can be numbers, bools, or strings	COMMENT
LOW	docs/configuration.md	221		COMMENT
LOW	docs/configuration.md	241	models:	COMMENT
LOW	docs/configuration.md	261	--ctx-size ${default_ctx}	COMMENT
LOW	docs/configuration.md	281	- "CUDA_VISIBLE_DEVICES=0,1,2"	COMMENT
LOW	docs/configuration.md	301	# - a value of 0 disables automatic unloading of the model	COMMENT
LOW	docs/configuration.md	321		COMMENT
LOW	docs/configuration.md	381	# becomes: c: [0.7, false, "model: llama"]	COMMENT
LOW	docs/configuration.md	421		COMMENT
LOW	docs/configuration.md	441	# - processes have 5 seconds to shutdown until forceful termination is attempted	COMMENT
LOW	docs/configuration.md	461	# 4. For each candidate set, compute cost: sum of evict_costs for	COMMENT
LOW	docs/configuration.md	481	v: voxtral-model	COMMENT
LOW	docs/configuration.md	501	# "a \| b" → [a], [b]	COMMENT
LOW	docs/configuration.md	521		COMMENT
LOW	docs/configuration.md	541	# - peers can be any server that provides the /v1/ generative api endpoints supported by llama-swap	COMMENT
LOW	docs/configuration.md	581	# filters: a dictionary of filter settings for peer requests	COMMENT

Decorative Section Separators9 hits · 36 pts

Severity	File	Line	Snippet	Context
MEDIUM⚡	config.example.yaml	5	# -------------------------------------	COMMENT
MEDIUM⚡	config.example.yaml	8	# ====================================	COMMENT
MEDIUM⚡	config.example.yaml	12	# ====================================	COMMENT
MEDIUM	…/src/components/playground/ConcurrencyInterface.svelte	130	// Detect and split the llama-swap loading block (wrapped in ━━━━━ markers,	COMMENT
MEDIUM⚡	docs/configuration.md	96	# -------------------------------------	COMMENT
MEDIUM⚡	docs/configuration.md	99	# ====================================	COMMENT
MEDIUM⚡	docs/configuration.md	103	# ====================================	COMMENT
MEDIUM	docs/configuration.md	444	# =============================================================================	COMMENT
MEDIUM	docs/configuration.md	446	# =============================================================================	COMMENT

Synthetic Comment Markers3 hits · 20 pts

Severity	File	Line	Snippet	Context
HIGH	config-schema.json	490	"description": "Controls what is logged to stdout. 'proxy': logs generated by llama-swap, 'upstream': copy o	CODE
HIGH	config.example.yaml	40	# - "proxy": logs generated by llama-swap when swapping models,	COMMENT
HIGH	docs/configuration.md	131	# - "proxy": logs generated by llama-swap when swapping models,	COMMENT

Modern AI Meta-Vocabulary3 hits · 8 pts

Severity	File	Line	Snippet	Context
MEDIUM	internal/event/README.md	3	The original code uses a `time.Ticker` to process the event queue which caused a large increase in CPU usage ([#189](htt	CODE
MEDIUM	docs/newrouter-todo.md	41	## Phase 4 — `internal/server` package scaffolding (ProxyManager replacement) -- Completed.	COMMENT
MEDIUM	docs/newrouter-todo.md	68	### Phase 4a — package scaffolding -- Completed.	COMMENT

Redundant / Tautological Comments5 hits · 8 pts

Severity	File	Line	Snippet	Context
LOW	docker/build-container.sh	32	# Check if ARCH is in the allowed list	COMMENT
LOW	docker/build-container.sh	38	# Check if GITHUB_TOKEN is set and not empty	COMMENT
LOW	docker/build-container.sh	97	# Check if response is empty array (no more pages)	COMMENT
LOW	docker/build-image.sh	227	# Check if our custom builder exists with the right config, create/update if needed	COMMENT
LOW	.github/workflows/config-schema.yml	28	# Check if the file is valid JSON	COMMENT

Example Usage Blocks2 hits · 3 pts

Severity	File	Line	Snippet	Context
LOW	docker/build-image.sh	5	# Usage:	COMMENT
LOW	docker/unified/build-image.sh	5	# Usage:	COMMENT

Self-Referential Comments1 hit · 3 pts

Severity	File	Line	Snippet	Context
MEDIUM	docker/build-image.sh	237	# Create the builder with the config	COMMENT

Fake / Example Data2 hits · 2 pts

Severity	File	Line	Snippet	Context
LOW	cmd/fake-model/main.go	21	"Lorem ipsum dolor sit amet consectetur adipiscing elit sed do eiusmod tempor " +	CODE
LOW	cmd/fake-model/main.go	21	"Lorem ipsum dolor sit amet consectetur adipiscing elit sed do eiusmod tempor " +	CODE

Slop Phrases1 hit · 2 pts

Severity	File	Line	Snippet	Context
MEDIUM	internal/router/loading_remarks.go	85	"The model saw \"As an AI language model\" and won't stop saying it now",	CODE

Hyper-Verbose Identifiers1 hit · 1 pts

Severity	File	Line	Snippet	Context
LOW	ui-svelte/src/lib/playgroundInterface.ts	22	export function createPlaygroundInterface(	CODE

Analysis Overview

What These Metrics Mean

Score History

Severity Breakdown

Directory Score Breakdown

Pattern Findings