## What this does

Fixes an edge case in `Models.models_dev_model_to_info` where
`created_at` was built from a blank date when both `release_date` and
`last_updated` were missing.

Before this fix, it could build `" 00:00:00 UTC"`, which later parsed to
the current date, causing incorrect and non-deterministic `created_at`
values in model metadata.

This PR:
- Sets `created_at` only when a non-blank date is present.
- Keeps `created_at` as `nil` when both source dates are missing.
- Adds a regression test covering this nil-date case.

## Type of change

- [x] Bug fix
- [ ] New feature
- [ ] Breaking change
- [ ] Documentation
- [ ] Performance improvement

## Scope check

- [x] I read the [Contributing
Guide](https://github.com/crmne/ruby_llm/blob/main/CONTRIBUTING.md)
- [x] This aligns with RubyLLM's focus on **LLM communication**
- [x] This isn't application-specific logic that belongs in user code
- [x] This benefits most users, not just my specific use case

## Required for new features

<!-- Skip this section for bug fixes and documentation -->

- [ ] I opened an issue **before** writing code and received maintainer
approval
- [ ] Linked issue: #___

**PRs for new features or enhancements without a prior approved issue
will be closed.**

## Quality check

- [x] I ran `overcommit --install` and all hooks pass
- [x] I tested my changes thoroughly
- [ ] For provider changes: Re-recorded VCR cassettes with `bundle exec
rake vcr:record[provider_name]`
- [ ] All tests pass: `bundle exec rspec`
- [ ] I updated documentation if needed
- [x] I didn't modify auto-generated files manually (`models.json`,
`aliases.json`)

## AI-generated code

- [x] I used AI tools to help write this code
- [x] I have reviewed and understand all generated code (required if
above is checked)

## API changes

- [ ] Breaking change
- [ ] New public methods/classes
- [ ] Changed method signatures
- [x] No API changes

## What this does

Sanitizes schema names in `build_schema_payload` by replacing characters
that don't match `[a-zA-Z0-9_-]` with `_`.

Since v1.13.0, `normalize_schema_payload` preserves the `:name` key from
`to_json_schema`. When using `ruby_llm-schema`, the schema name defaults
to the full Ruby class name (e.g. `MyApp::Nested::Schema`) which
contains `::`. OpenAI requires `response_format.json_schema.name` to
match `^[a-zA-Z0-9_-]+$`, so the request is rejected with a
`BadRequestError`.

## Type of change

- [x] Bug fix
- [ ] New feature
- [ ] Breaking change
- [ ] Documentation
- [ ] Performance improvement

## Scope check

- [x] I read the [Contributing
Guide](https://github.com/crmne/ruby_llm/blob/main/CONTRIBUTING.md)
- [x] This aligns with RubyLLM's focus on **LLM communication**
- [x] This isn't application-specific logic that belongs in user code
- [x] This benefits most users, not just my specific use case

## Quality check

- [x] I ran `overcommit --install` and all hooks pass
- [x] I tested my changes thoroughly
- [ ] For provider changes: Re-recorded VCR cassettes with `bundle exec
rake vcr:record[provider_name]`
  - [x] All tests pass: `bundle exec rspec`
- [ ] I updated documentation if needed
- [x] I didn't modify auto-generated files manually (`models.json`,
`aliases.json`)

## AI-generated code

- [x] I used AI tools to help write this code
- [x] I have reviewed and understand all generated code (required if
above is checked)

## API changes

- [ ] Breaking change
- [ ] New public methods/classes
- [ ] Changed method signatures
- [x] No API changes

Fixes #654

---------

Co-authored-by: Alexey Ivanov <[email protected]>
Co-authored-by: Carmine Paolino <[email protected]>

## What this does

Prevents streaming parsers from passing non-Hash SSE payloads into
provider `build_chunk` methods.

`handle_data` can parse valid JSON scalars (like `true`) from `data:`
frames. Those values are truthy and previously passed `if data`, which
crashes provider chunk builders that assume a Hash and call `dig`.

This change narrows the guard in `handle_stream` to `data.is_a?(Hash)`.

## Why this is minimal

- One-line runtime behavior change in shared streaming handler
- No provider-specific branching
- Keeps existing behavior for all normal Hash event payloads

## Tests

- Added `spec/ruby_llm/streaming_spec.rb`
- Verifies non-Hash SSE payloads are ignored without raising
- Verifies Hash SSE payloads still flow to `build_chunk`

Fixes #656