A tiny friend in your Mac's notch and on your iPhone that keeps an eye on your AI coding agents: Claude Code, Codex, Cursor, Gemini CLI, Antigravity and more. Approve from the notch or your Lock Screen.
Swift ai-agentsanthropicantigravityclaudeclaude-codecodexcursordynamic-islandgemini-cliiosiphonlinuxmacosmacos-appmenubar-appnotchopen-sourceswiftswiftuiwindows
#if !APPSTORE
import Foundation
// MARK: - EntityResolver
/// Fuzzy-matches a voice entity name against a pill list.
/// Resolution order: aliases → exact match → word-match → Levenshtein ≤ 2.
/// No AppKit.
enum EntityResolver {
// MARK: - Aliases
/// Fixed aliases for names that don't appear in pill names verbatim.
/// Key: normalised alias. Value: pill id.
static let aliases: [String: String] = [
// VS Code pill is the Claude Code integration
"claude": "integration_claude",
"claude code": "integration_claude",
// How speech recognition writes "VS Code" (FR and EN)
"vs code": "integration_claude",
"vscode": "integration_claude",
"v s code": "integration_claude",
"visual studio code": "integration_claude",
"visual studio": "integration_claude",
"ves code": "integration_claude",
"vee s code": "integration_claude",
"vi s code": "integration_claude",
"vsco": "integration_claude",
// Agents
"gemini": "agent_gemini",
"copilot": "agent_copilot",
"muse": "agent_muse",
"hermes": "agent_hermes",
"amp": "agent_amp",
"opencode": "agent_opencode",
"antigravity": "agent_antigravity",
"codex": "agent_codex",
"claude desktop": "agent_claude-desktop",
// AI
"google": "ai_google",
"openai": "ai_openai",
"open ai": "ai_openai",
"ollama": "ai_ollama",
"lmstudio": "ai_lmstudio",
"anthropic": "ai_anthropic",
// Services
"github": "integration_github",
// GitHub phonetic variants (speech recognition mishears)
"gt": "integration_github",
"g t": "integration_github",
"git hub": "integration_github",
"jit hub": "integration_github",
"guitare hub": "integration_github",
"guitar hub": "integration_github",
"gitub": "integration_github",
"guit hub": "integration_github",
"vercel": "integration_vercel",
"notion": "integration_notion",
"stripe": "integration_stripe",
"resend": "integration_resend",
"n8n": "integration_n8n",
"calcom": "integration_calcom",
"cal": "integration_calcom",
"spotify": "integration_spotify",
"apple music": "integration_music",
"music": "integration_music",
"musique": "integration_music",
]
// MARK: - Public API
/// Resolve `query` to a pill id from `pills`.
/// Pass `category` to restrict to workspace pills (for main-pill commands).
/// Returns nil when no match within tolerance, or when the match is ambiguous.
static func resolve(
_ query: String,
from pills: [PillDefinition],
category: PillCategory? = nil
) -> String? {
let q = IntentParser.normalise(query)
var pool = pills
if let cat = category { pool = pool.filter { $0.category == cat } }
guard !pool.isEmpty, !q.isEmpty else { return nil }
// 1. Alias table (handles "claude", "gemini", "google", etc.)
if let aliasId = aliases[q] {
if pool.contains(where: { $0.id == aliasId }) { return aliasId }
}
// 2. Exact name match (case-insensitive + diacritic-insensitive)
if let exact = pool.first(where: { IntentParser.normalise($0.name) == q }) {
return exact.id
}
// 3. Word-match: q is a complete word inside the pill's normalised name
// e.g. "gemini" matches "gemini cli", "copilot" matches "copilot cli"
let wordMatches = pool.filter { def in
let nameWords = IntentParser.normalise(def.name).split(separator: " ").map(String.init)
return nameWords.contains(q)
}
if wordMatches.count == 1 { return wordMatches[0].id }
if wordMatches.count > 1 { return nil } // ambiguous
// 4. Levenshtein ≤ 2 (≤ 1 for short queries)
var best: (id: String, dist: Int)?
var secondBest: (id: String, dist: Int)?
for def in pool {
let target = IntentParser.normalise(def.name)
let d = levenshtein(q, target)
if d < (best?.dist ?? Int.max) {
secondBest = best
best = (def.id, d)
} else if d < (secondBest?.dist ?? Int.max) {
secondBest = (def.id, d)
}
}
guard let winner = best else { return nil }
let tolerance = q.count <= 4 ? 1 : 2
guard winner.dist <= tolerance else { return nil }
if let runner = secondBest, runner.dist <= tolerance { return nil }
return winner.id
}
// MARK: - Levenshtein distance
/// Standard Wagner–Fischer DP. Operates on Character arrays.
static func levenshtein(_ a: String, _ b: String) -> Int {
let ac = Array(a), bc = Array(b)
let m = ac.count, n = bc.count
if m == 0 { return n }
if n == 0 { return m }
var row = Array(0...n)
for i in 1...m {
var prev = row[0]
row[0] = i
for j in 1...n {
let temp = row[j]
row[j] = ac[i-1] == bc[j-1]
? prev
: Swift.min(prev, Swift.min(row[j], row[j-1])) + 1
prev = temp
}
}
return row[n]
}
}
#endif