[Feature]: X-SMG-Target-Worker for Prefill vs Decode

[lawrence-harmonic]

Problem Statement

X-SMG-Target-Worker is useful, but it does not work if we have a different number of Prefill and Decode workers and need to specify the target worker index separately.

Proposed Solution

Support something like "X-SMG-Target-Prefill-Worker" or "X-SMG-Target-Decode-Worker". It would be even better if this works with DP-aware.

Alternatives Considered

No response

Feature Area

Routing & Load Balancing

Affected Component(s)

model-gateway

Use Case

In agentic RL we may have a fixed number of clients the RL system may want to load balance manually.

Priority to You

Nice to have

Contribution

• I am willing to contribute this feature (with guidance)
• I am willing to help test this feature
• I can provide more detailed requirements if needed

Additional Context

No response