#125 Support multiple LLM providers per member

[Anonymous]

Originally created by: kumaakh
Originally owned by: yashrajsapra

Summary

Currently each member supports a single LLM provider (none, Claude, or Gemini). This should be expanded to a list of providers per member.

Goal

Allow a member to be registered with multiple LLM providers (e.g. both Gemini and Claude), so the PM can drive which specific model is used at each stage of a larger workflow.

Context

Raised via community FAQ: https://github.com/Apra-Labs/apra-fleet/discussions/123

[Anonymous]

Originally posted by: kumaakh

Technical direction: The cleanest model is a providers array in the member registry replacing the current single provider field.

1. Schema (src/types.ts): change provider?: string to providers?: string[]. Backward compat: single-element array when reading old registry entries.
2. Register_member / update_member: accept providers (array) alongside the legacy provider (string, maps to [provider]).
3. execute_prompt dispatch: add optional provider param — if specified, use that provider's strategy; if omitted, use the first provider in the member's list (current behavior preserved).
4. provision_llm_auth: already takes member_name — extend to accept optional provider param to target a specific provider's auth on a multi-provider member.
5. member_detail / fleet_status: display all providers.

This allows PM to dispatch a Gemini pass for reasoning followed by a Claude pass for code generation on the same member without re-registering anything.

[Anonymous]

• assigned_to: yashrajsapra