Add Claude Explain feature for AI-powered code explanations
This PR introduces Claude Explain, a new feature that provides AI-powered explanations of compiler output directly within Compiler Explorer.
Key features:
Claude Explain functionality:
- New explain view pane
- Explains compiler output with full context of source code and compilation output
- Configurable audience level and explanation type
- Response caching to improve performance and reduce API calls
- Usage statistics display showing requests used and token counts
User experience:
- Consent flow on first use explaining data handling and privacy
- AI disclaimer banner warning about potential LLM inaccuracies
- Respects "no-ai" directive in source code for users who don't want AI processing
Privacy and security:
- Data sent to Anthropic's Claude API as documented in privacy policy
- No data used for model training
- Clear consent required before first use
- Support for opting out via "no-ai" directive
The feature is marked as beta and can be enabled via configuration.
Co-authored-by: Claude <noreply@anthropic.com>
Takes the bones of shell-split and makes it handle lots of situations
without throwing exceptions, and using modern-ish typescript/javascript
stuff.
Adds lots more tests too.
Removes unneeded package dependencies.
Changes from before: escaping a pipe character `moo \\# hello` would
drop the `# hello` part (?) that is, I updated the test that did:
```ts
expect(splitArguments('hello \\#veryfancy etc')).toEqual(['hello', '\\']);
```
to be the more sane expectation of `['hello', '\\#veryfancy']`.
Everything else I could think of (unterminated strings, escapes etc) I
kept the same. And added tests for.
Hopefully addresses #7201 without breaking any reasonable use of the
code (though @partouf raises fair concerns about this.)
- Uses strings across the board in the UI part (no functional change
here from before; all state is the same as it was before).
- Sends _arrays_ in the POST, using the same `splitArguments` code as
the backend.
- Backend _still supports_ strings (though doesn't advertise), also
using same `splitArguments`.
- Moves `splitArguments` into common utils, and rephrases to avoid
unnecessary use of underscore and ES2021+ code.
Tested locally:
- with both old and new client code (ran new backend and old webcode to
show sending strings still works)
- with creating and removing tool windows (checked with `ldd` locally)
- with various strings on the client `moo foo "this is bad" #moo` and
even "error" things like `this is "badger` (with a missing close quote).
All works as you'd expect
Happy to break the "move the splitArguments" code into a separate PR if
that'd be easier to review separately.
Fixes#7195
Long numbers are sometimes hard to read in the numeric tooltip. To
improve readability, this PR adds language-specific digit separators to
the numeric tooltip.
Decimal numbers are grouped into chunks of three digits while
hexadecimal numbers are grouped into chunks of length four.
The digit separator is language-specific and chosen so that the number
is a valid token in the source language.
[Examples](https://godbolt.org/z/s86cMbjeK):
* for C++, hovering the number `8583909746840200552` shows this tooltip:
`8'583'909'746'840'200'552 = 0x7720'2C6F'6C6C'6568 =
6.5188685003648344e+265`
* for Python, hovering the number `-12345678` shows this tooltip:
`-123_456_789 = 0xFFFF_FFFF_F8A4_32EB = -2.66427945e+34f`
For languages that don’t have a `digitSeparator` set, the tooltip is not
changed.
This PR refactors some common utilities out of lib/ and into shared/ and
eliminates some use of underscore.js, as well as general type
improvements done along the way.
