# Precise Special Functions — Support

Precise Special Functions is offered on a best-effort basis by an individual developer as part of a research project on how AI assistants route tool invocations. Support is provided in good faith, with the honest expectations below.

## What we can help with

- Bug reports (the tool returned a value you believe is incorrect).
- ChatGPT integration issues (the app tool is not appearing in your connector list; tool invocations fail; app card does not render).
- Argument or schema questions (which parameters a specific tool accepts; what precision levels are supported).
- Documentation errors on this site.

## What we cannot help with

- Mathematics tutoring or interpretation of results.
- Custom feature requests or new tool additions on demand.
- Deployment guidance for your own MCP servers.
- ChatGPT-side behavior we do not control (routing decisions, response formatting, model version selection).

## Response expectations

- **Response time:** typically within 5-10 business days. Not guaranteed. This is a research instrument, not a commercial product.
- **Severity:** we respond faster to reports that meaningfully affect measurement integrity or that indicate a wrong computed value.
- **No SLA:** see the [Terms of Service](/terms) for the full "no warranty, no availability guarantee" position.

## How to reach us

**Email:** [lanir.assaf@gmail.com](mailto:lanir.assaf@gmail.com)

When reporting an issue, please include:

1. The exact prompt you sent in ChatGPT.
2. Which tool ChatGPT invoked (from the tool-call panel; it will say something like "Called tool: bessel_j").
3. What arguments the tool received (also in the tool-call panel: `{n: 0, x: 2.4048, precision_dps: 30}`).
4. What value the tool returned.
5. What you expected instead, and — if you can — a reference computation from another source (Wolfram Alpha, scipy, another arbitrary-precision library).

For security reports, please include "SECURITY" in the subject line so it gets faster attention.

## Known limitations (not bugs)

Two behaviors we routinely see reported as "the tool didn't invoke". Both are ChatGPT routing decisions rather than app malfunctions.

- **Gamma function** at arbitrary arguments: ChatGPT sometimes computes Gamma inline via its Python interpreter rather than routing to this tool. If you require invocation for a specific test, the explicit prompt "Use the Precise Special Functions app to compute Γ(x) to N digits" reliably invokes.
- **Famous named constants** (e.g., ζ(2), ζ(3), first zeros of Bessel functions, Γ(1/2)): ChatGPT often recites these values from parametric memory instead of routing to this tool. Choose a non-landmark argument if invocation is required for a specific test.

## Contact

**Email:** [lanir.assaf@gmail.com](mailto:lanir.assaf@gmail.com)