Fix Gemini Live to handle bundled server_content fields

[chadbailey59]

Summary

• Gemini 3.x can bundle multiple fields (e.g. model_turn and output_transcription) on the same server_content message
• The previous elif chain only processed the first matching field, silently dropping the rest
• Switched to independent if checks so every field on a message is handled

Test plan

• Test with Gemini Live using transcription enabled to verify both model audio and output transcription are processed from bundled messages
• Verify interruption handling still works correctly
• Confirm tool calls and session resumption updates are unaffected

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 0% with 11 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/pipecat/services/google/gemini_live/llm.py	0.00%	11 Missing ⚠️

Files with missing lines	Coverage Δ
src/pipecat/services/google/gemini_live/llm.py	26.24% <0.00%> (-0.33%)	⬇️

... and 34 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[kompfner]

Haven't dug too deep into this yet, but if this is broadly the case I'd be a bit surprised if the fix were as simple as just changing elif to if (in other words, I'd expect that there'd be some subtle implicit assumptions in our message handling logic about messages being mutually exclusive).

Do we know which server_content fields are allowed together? If it's only model_turn and output_transcription maybe we can be a bit more targeted in our change.

[kompfner]

I'd expect that there'd be some subtle implicit assumptions in our message handling logic about messages being mutually exclusive)

Indeed: #4148

[kompfner]

these tests feel a little too "reach-into-internals-y". pretty much any internal refactoring or name changes will break these.

[kwindla]

I agree. I told @chadbailey59 that I thought we should just cherry-pick the actual fix and not include the test. (I just let Codex write the test it wanted to write, and then committed that change. But I shouldn't have committed the test.) But I think I told Chad that in a DM so it got lost in the shuffle. Apologies.

[kompfner]

Waiting to test/validate this once we can (once our usage of history_config is unblocked)

[kompfner]

Waiting to test/validate this once we can (once our usage of history_config is unblocked)

Gemini 3.1 Flash Live is working for basic conversation-and-tool-call usage (26, 26b, 26e, 26g). We'll continue to iron out the kinks in the others.