Feature Request: Different models for coding and testing

[jpbester]

Currently you can only select a single model for coding and testing. It would be good to be able to select a model for coding and one for testing. For instance, choosing Opus for code, but Sonnet for testing.

Quick Analysis:

1. In parallel_orchestrator.py (lines 486-503 for coding, 585-595 for testing):
   • Both agent types receive the same self.model parameter
   • The orchestrator passes --model to spawned subprocesses
2. In client.py (line 251):
   - create_client(model) accepts a single model parameter
3. No differentiation exists between agent types for model selection

The Fix Needed

To support different models per agent type, you'd need:

1. Add separate CLI arguments:
   --coding-model opus
   --testing-model sonnet
2. Update the orchestrator to store both:
   self.coding_model = coding_model
   self.testing_model = testing_model
3. Pass the appropriate model when spawning:

For coding agents (line 501):

if self.coding_model:
cmd.extend(["--model", self.coding_model])

For testing agents (line 594):

if self.testing_model:
cmd.extend(["--model", self.testing_model])