CueLight

# Fix list: CueLight

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cuelight, the October 2026 research run, assessed 15 October 2026. Grade D, 48.8 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on CueLight: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 29 out of 100, up to 14.2 more on the total

Why it scored 29: Read as local software. Installs with pip from the Git repository, not from PyPI, with Python 3.10 or later and its dependencies declared in pyproject.toml (10 of 20). The repository holds two tests (streaming matches whole-sequence output, chunk size is enforced), but the only GitHub workflow is the Pages deployment, so no public CI runs them (5 of 25). No issues or pull requests exist, and at one day old there is no record to judge either way (12 of 25). Version 0.1.0 in pyproject.toml with no tags or changelog (2 of 15). Below 1.0, and the README calls it a first iteration (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Security & auth, 50 out of 100, up to 8.8 more on the total

Why it scored 50: Local-software reading. No credential and no open port. The code runs in the caller's process, weights load from safetensors rather than pickle, but `snapshot_download` fetches the main branch of each Hugging Face repository with no pinned revision (20 of 30). It only classifies audio and takes no actions (16 of 20). It returns probabilities, not content from third parties (10). No logging. Per-frame probabilities and a reset counter are visible to the caller (4 of 15). No SECURITY.md, disclosure policy, security.txt or advisories (0 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 3. Schema & documentation, 50 out of 100, up to 8.1 more on the total

Why it scored 50: Library reading. No OpenAPI applies. The public classes are small, with type hints on most signatures and frozen dataclasses for both policies, though `push` takes an untyped chunk (15 of 25). No llms.txt on the Pages site, and docs are the README (0). The README says which model suits which device, when to choose each preset, and states the limits (English only, audio only, Base's CPU cost) (16 of 20). The 320-sample frame is enforced by an assert, while an unknown preset name falls back silently to default thresholds rather than raising (8 of 15). Quick-start snippets and two example scripts. Error behaviour is not documented (8 of 15). A version number in pyproject.toml and `__version__`, with no changelog (3 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 4. Agent ergonomics, 61 out of 100, up to 6.3 more on the total

Why it scored 61: Library reading. Four public names and one call per frame, with an 11 KB README, so the context cost of learning it is low (22 of 25). Output is one probability and one speech probability per frame, and thresholds, minimum silence, timeout and reset behaviour can all be set (14 of 20). The only error is an assert on chunk size, which `python -O` strips, and a mistyped preset gives defaults without a warning (6 of 20). `push` is stateful by design. `clear()` resets between calls and the detector resets itself at each new turn, and the source shows streaming matches whole-clip output to 1e-4 (10 of 20). One required argument with working defaults. Python only, with ONNX step files and a browser demo but no packaged second-language SDK (9 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 5. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: No payment protocol (0). Read with the self-hosted rule. Apache-2.0 code and weights with no hosted or paid option (20). No card or account (20). Weights download anonymously, with no signup (20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 6. Maintenance & community, 46 out of 100, up to 4.7 more on the total

Why it scored 46: Published on 14 October 2026, one day before the check (30). One publication and no tagged releases, given a quarter of the line as for a single release elsewhere (5 of 20). No issues or pull requests yet, so responsiveness cannot be seen (5 of 25). Not on PyPI. The weights are on Hugging Face and the package installs only from Git (3 of 15). Dependencies are current minimums (torch 2.1 or later), with no lockfile, Dependabot or CI (3 of 10). The project is one day old, so these lines reflect the checklist, not a judgement of the maintainer.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 7. Transparency & trust, 54 out of 100, up to 4 more on the total

Made of editorial 67, provenance 40.

Why it scored 54: Apache-2.0 for code and weights, with a NOTICE that lists each training dataset, teacher model and licence, and the conditions passed on from the data (30). Nothing is hosted. The README says the browser demo keeps audio on the device, which the demo code agrees with, and that weights download from Hugging Face on first use. There is no privacy statement as such (22 of 30). No deprecation policy or dated notices (0 of 20). No telemetry found in the package or demo source. The README does not mention telemetry, and huggingface_hub's own defaults are not discussed (15 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: sshah03.github.io, no registry record we could read (0 of 15)
- Status page: not found (0 of 10)
- Changelog: not found (0 of 10)
- security.txt: not found (0 of 10)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- Judgement call on category. Listed in voice agents under `voice.pipeline`, as a component of the speech-to-text, model and text-to-speech chain, because no category covers turn detection on its own. It runs no conversation, so `voice.agent` is left out. `inference.open-weights` is added because the weights are published.
- The lead was right. Initial commit and the Nano, Nano-SSM and Base weights on 14 October 2026 UTC (13 October in the author's time zone). `firstReleased` is 2026-10-14. The project was one day old at the check.
- Graded on the Python library interface, which is what an agent-building pipeline would call. There is no hosted API or local server.
- We did not install or run the package or the demo. Accuracy and CPU figures are the author's, from the README.
- The author's eot-bench results use presets tuned on half of eot-bench, and eot-bench data was used to choose between models. The README discloses both. No deduction, since the claims are qualified.
- The OpenVoiceOS plugin README gives `pip install ovos-eot-plugin-cuelight`, but PyPI had no such project on 15 October 2026.

## Weaknesses

- English only, and the author does not plan other languages soon
- Not on PyPI, no tags, releases, changelog or CI, and one commit at the check
- Benchmarks are self-reported, and the presets were tuned on half of the benchmark they are scored on
- Weights download from Hugging Face main branches with no pinned revision
- An unknown preset name falls back silently to default thresholds

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Feed 320 float32 samples at 16 kHz mono per `push`. Resample first, or the call raises an AssertionError
- Spell the preset as `snappy` or `careful`. Any other name silently uses default thresholds
- Call `clear()` before a new call or session. Pass `reset_on_turn_start=False` if your pipeline already resets per turn
- Pin weights by downloading them to a local folder and passing that path to `from_pretrained`
- Don't use it to identify speakers or clone voices. The training data's conditions forbid it

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.