ashlrai/lexicon

You said:        "tell Ashlr.AI to deploy the Kubernetes auth service"
STT heard:       "tell Ashler to deploy the Cooper Nettie's off service"
Agent received:  "tell Ashlr.AI to deploy the Kubernetes auth service"

That is the live demo running this repo’s real matcher on your text, in your browser, with nothing installed. (Its Dictate button uses your browser’s own speech recognizer, which in Chrome sends audio to Google.)

Measured

Method, full tables and every failing case are in docs/BENCHMARK.md. Reproduce with npm run bench (no setup beyond a clone), or npm run bench:audio && npm run bench:compare (needs macOS and whisper.cpp).

Against the alternatives. 330 clips of synthesized (text-to-speech) audio through whisper.cpp small.en, three voices. Same audio, same recognizer, same 70-term lexicon in every row; the only thing that changes is how the proper nouns get fixed.

Every row uses the same curated seventy-term lexicon (bench/lexicon.yaml). The prose column is a property of the terms in the file as much as of the matcher, so a lexicon assembled some other way is a different measurement.

Exact substitution recovers the spellings someone already wrote down, and nothing else. It cannot reach Versal, Superbase, CloudFloor or pedantic, because no table written by hand contains the mistake you have not heard yet. The phonetic and fuzzy tiers exist for that gap and recover 31 of the 279 term slots on their own, which is 11 points. The remaining 9 points over the casing-aware row come from the alias tier’s tolerance for how the recognizer breaks a name into words, since that row already matches case. --prompt is a complement rather than a rival: stacked with the lexicon it reaches 95.7%.

Two honest notes about that table. Raw whisper.cpp cannot wrongly change prose because nothing runs, which is the absence of the feature rather than an advantage. --prompt is not the same case: it biases the recognizer itself, so whatever it changes is already in the transcript before scoring begins, while the metric counts sentences a post-pass altered. That cell is unmeasured rather than zero. Telling which way it goes would mean diffing prompted transcripts against unprompted ones, which this harness does not do.

Before and after.

Latency is about 0.3 ms per sentence. The TTS-audio rows use macOS text-to-speech read into whisper.cpp, so they are cleaner than a phone microphone.

The synthetic 5.1% is not a claim that speech-to-text gets 5% of proper nouns right in general. Every sentence in that corpus was written to contain a mis-hearing, so 5.1% is only the handful that came out right anyway. The honest “before” number is the TTS-audio one, 41.9%.

The last column counts ordinary prose only. Each corpus also contains sentences deliberately built to trip the matcher (a bare “llama” next to an Ollama term, sound-alikes, code spans), marked expected-hard; with those included the false-positive rate is 15.3% (19 of 124) synthetic and 16.7% (15 of 90) on audio. Both numbers, and every failing case, are in docs/BENCHMARK.md.

Held-out check (separate internal run, unseen voice). The tables above use a lexicon written for the test terms. To test the harder case, aliases were learned only from other sentences and voices, then scored on a voice and sentences never used for learning (63 clips). Synthetic speech from three open text-to-speech voices read into whisper.cpp, clips counted correct when every target term was recovered.

Clean prose changed by Lexicon: 0 of 60 in that run.

Whisper-prompt caveat. On the same 378 clips, whisper’s --prompt hint list alone lifted base.en from 97 to 228 and small.en from 141 to 320 clips correct, so a prompt is a strong baseline and Lexicon adds less on top of it (base.en 271, small.en 339 with Lexicon). The prompt also made base.en produce 55 degenerate transcripts (empty or looping text) against 5 without it, so a prompt is not free on small models. All audio here is synthetic, not microphone speech, and the harness is not yet published, so treat these as internal numbers until it is.

Install

The package is @ashlr/lexicon; the command is lexicon.

curl -fsSL https://ashlrai.github.io/lexicon/install.sh | sh   # CLI + the setup wizard
brew install ashlrai/tap/lexicon                               # or Homebrew (macOS, Linux)
npm i -g @ashlr/lexicon                                        # or npm (Node 20+)

Then open Claude Code and say a sentence with your company name in it. Done.

The install script runs lexicon setup for you (LEXICON_NO_SETUP=1 skips it); after a Homebrew or npm install, run it yourself. Every step is optional and safe to rerun, and lexicon setup --dry-run writes nothing while describing the run you would get from the same command without it: the steps it would perform, and the ones it would stop and ask about, with the answer pressing Enter gives each.

With Homebrew, always use the full tap name ashlrai/tap/lexicon. Plain brew install lexicon installs dns-lexicon, an unrelated DNS tool in homebrew-core.

Or skip the wizard and add one term by hand. The first argument is the canonical spelling, the rest are what STT actually produces:

lexicon add Ashlr.AI Ashler Ashlar "Ashler AI" --phonetic ASH-ler
lexicon normalize "tell Ashler to ship it"
# tell Ashlr.AI to ship it

Ten minutes in, teach it the names you actually say. lexicon import --guided asks where your proper nouns live (your contacts, your calendar, your GitHub), shows each candidate with where it was seen, and writes only the ones you approve. Nothing is read until you check its box, and email and chat stay out entirely.

Claude Code plugin, if you would rather not install a CLI at all. No Node install step, no build:

claude plugin marketplace add ashlrai/lexicon
claude plugin install lexicon@ashlrai

lexicon doctor checks the install. There is no telemetry. Vocabulary and application state are local files; ordinary hooks, correction, MCP, local API and extension paths do not send transcripts to a remote service. Explicit network operations include lexicon voice fetching a whisper model on first use, lexicon pack fetching community packs from a --registry URL with checksum verification, and opting into the GitHub source in lexicon import --guided. That source calls the GitHub API through your existing gh configuration and login (normally api.github.com) for profile, organization membership, member-login and repository-name metadata; no new OAuth scopes are requested. The source is selected before inspection, and candidates are reviewed before writes. The install script, npm and Homebrew fetch the package itself. See SECURITY.md.

The guided GitHub importer ships in 0.5.5 (PR #25); npm latest is currently 0.5.4.

MCP setup

lexicon install --apply does this for you. To wire it by hand:

# Claude Code
claude mcp add lexicon -- npx -y -p @ashlr/lexicon lexicon-mcp
# Codex
codex mcp add lexicon -- npx -y -p @ashlr/lexicon lexicon-mcp

Cursor (~/.cursor/mcp.json):

{ "mcpServers": { "lexicon": { "command": "npx", "args": ["-y", "-p", "@ashlr/lexicon", "lexicon-mcp"] } } }

Why

Speech-to-text is about 95% accurate on ordinary English and much worse on invented names. In the benchmark above, raw whisper.cpp base.en transcribed 117 of 279 dictated proper nouns correctly. “Ashlr.AI” becomes “Ashler”, “Kubernetes” becomes “Cooper Nettie’s”, “SaaS” becomes “sauce”, “auth” becomes “off”. Those are exactly the words an agent needs to get right.

Dictation apps (Wispr Flow, Superwhisper, Aqua) each keep their own dictionary and none of them share it. Agents (Claude Code /voice, ChatGPT voice, Codex, local Whisper) run their own recognizer with no user vocabulary at all. This is the portable layer in between: corrections happen after STT and before the model, wherever the text passes through.

This is not a dictation app. It sits between whatever dictation you already use and whatever agent you talk to. The research behind that call, including the kill criteria, is in docs/RESEARCH.md.

What you get

• Twenty MCP tools, two resources and two prompts, for Claude Code, Codex, Cursor, Windsurf, Gemini CLI, VS Code and Claude Desktop. Your agent can run its own setup: setup_lexicon, lexicon_doctor, install_client, trust_project, import_dictionary, import_vocabulary and suggest_terms mean “set up my lexicon” works without a terminal. The tools that change your machine preview first: setup_lexicon and install_client return a plan and write nothing until the agent passes apply: true, trust_project asks you directly (an MCP elicitation prompt the agent cannot answer) before pinning a file, import_dictionary takes dryRun, and import_vocabulary reads only explicitly selected sources, then requires a saved preview token and approved candidate ids with apply: true.
• A Claude Code plugin: MCP server, SessionStart and UserPromptSubmit hooks, a lexicon skill and a /lexicon command. Installs from this repo’s marketplace with no build step.
• A CLI with 25 commands, from lexicon add to lexicon voice.
• 176 starter terms in four packs (developer, AI, business, voice tools), one command each, plus community packs from registry indexes (lexicon pack search cardiology --registry ): checksums pinned at install, every install and update previewed before it writes.
• Fifteen export formats (Wispr Flow, Superwhisper, macOS Text Replacement, espanso, Whisper and OpenAI prompts, Deepgram, AssemblyAI, Azure, Google, CLAUDE.md, markdown, text, CSV, JSON) and eight importers (including the CLAUDE.md table it writes) for the dictionary you already trained.
• Repo harvesting, a guided vocabulary import (lexicon import --guided: your contacts, calendar and GitHub, each source opt-in, every candidate approved before it is written), correction learning (“it’s Ashlr.AI not Ashler”), usage stats, suggestions mined from your voice history, and a trust gate for project lexicons.
• A plain library. normalize() is a pure function: text plus lexicon in, corrected text and a replacement list out.

Where it applies

How it works

Three tiers over token windows: exact alias first, then double-metaphone phonetic, then Damerau-Levenshtein fuzzy above a confidence floor. Exact hits win the span; matches never overlap. A stoplist of about 3400 common English words, per-term never lists, and (with the default skipCode) code spans, URLs, emails, paths and glued identifiers are all off limits. That is why zero clean sentences changed in the benchmark. Every replacement reports its reason and confidence.

lexicon normalize --diff "deploy to head sner with cuban eatties"
# stderr:  "head sner" -> "Hetzner" (alias, 1.00)
#          "cuban eatties" -> "Kubernetes" (phonetic, 0.86)
# stdout:  deploy to Hetzner with Kubernetes

The rules in full, including every guard, are in docs/MATCHING.md.

Documentation

The full index is in docs/, grouped by task: get started, use it with your client, understand how it works, contribute, internals. The three pages most people need:

Writing an agent that installs this for someone? docs/AGENTS.md is written to you. Changing the code? Start at CONTRIBUTING.md and docs/ARCHITECTURE.md.

Also at the root: SECURITY.md, CODE_OF_CONDUCT.md, CHANGELOG.md.

Downloads

Every GitHub release attaches the browser extension for Chrome/Edge/Brave and for Firefox, LexiconBar.app.zip for macOS, the npm tarball for offline installs, and SHA256SUMS. The Homebrew formula lives in ashlrai/homebrew-tap. After 0.5.6 is published, npm i -g github:ashlrai/lexicon#v0.5.6 installs that tag straight from GitHub and builds on install.

Roadmap and non-goals

Non-goals: this is not a dictation app, and there are no hosted accounts and no sync service. It is a file.

• Chrome Web Store and Firefox AMO listings for the extension. Today it installs from the release zip.
• Notarized macOS app. LexiconBar is ad-hoc signed, so the first launch needs right-click and Open.
• Linux tray app with the same push-to-talk and fix-clipboard actions. The Windows one is built: see docs/WINDOWS-APP.md.
• Non-English phonetics. Double metaphone is tuned for English; names in other languages fall back to fuzzy matching.
• Real-microphone benchmark. The audio corpus is macOS text-to-speech read into whisper.cpp, not recorded speech.

Part of the Ashlr ecosystem

Lexicon is one of the open-source tools from Ashlr, a forward-deployed engineering company: Phantom Secrets, Locus, Ashlr Stack, Ashlr Markdown and the Ashlr plugin.

Contributing

Good first issues are labelled and scoped: a new starter pack, an exporter, an importer, a harvester source. CONTRIBUTING.md has the setup, the test layout and a recipe for each.

Found a name it gets wrong? Open a misheard term issue.

License

MIT. Copyright 2026 AshlrAI, Inc.