Fix streamed message model_id latching onto empty string (nil costs on Azure OpenAI)

[bdegomme]

What this does

Fixes streamed messages losing their model_id — and therefore all cost information — when the provider's first stream chunk carries an empty model field.

Azure OpenAI's first SSE chunk (the prompt filter chunk) has "model": "", while subsequent chunks carry the real id:

chunk model_ids: ["", "gpt-5.4-2026-03-05", "gpt-5.4-2026-03-05", ...]

StreamAccumulator#add does @model_id ||= chunk.model_id, and since "" is truthy, the empty string latches for the whole stream. The assembled Message then has model_id: "", Message#model_info raises ModelNotFoundError internally, and message.cost returns nil for every component (total, input, output, …) even though token usage is fully populated. Non-streamed requests against the same endpoint price correctly.

The fix keeps the first non-empty model_id instead:

@model_id = chunk.model_id if @model_id.to_s.empty?

Verified against a live Azure OpenAI deployment: before the change a streamed chat.ask returned cost.total => nil; after, it resolves the dated model and prices identically to the non-streaming path.

Type of change

• Bug fix
• New feature
• Breaking change
• Documentation
• Performance improvement

Scope check

• I read the Contributing Guide
• This aligns with RubyLLM's focus on LLM communication
• This isn't application-specific logic that belongs in user code
• This benefits most users, not just my specific use case

Quality check

• I tested my changes thoroughly
  • All tests pass: bundle exec rspec (stream accumulator, streaming, and protocol specs)
• I updated documentation if needed
• I didn't modify auto-generated files manually (models.json, aliases.json)

AI-generated code

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

API changes

• Breaking change
• New public methods/classes
• Changed method signatures
• No API changes

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.18%. Comparing base (5e252ea) to head (0d149ee).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #830   +/-   ##
=======================================
  Coverage   82.18%   82.18%           
=======================================
  Files         170      170           
  Lines        8029     8029           
  Branches     1347     1348    +1     
=======================================
  Hits         6599     6599           
  Misses        890      890           
  Partials      540      540           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.