litellm: upgrade minimax to MiniMax-M3 default

[octo-patch]

Summary

Promote MiniMax-M3 as the new default for the minimax LiteLLM
alias in both cloud and hybrid modes. M3 is MiniMax's latest
general-purpose chat model (512K context window, 128K max output).
The existing M2.7 mapping is preserved under a minimax-legacy alias
for installs that pinned the old default and need an explicit fallback,
and minimax-fast continues to point at MiniMax-M2.7-highspeed
(M3 does not ship a highspeed variant today).

Diff:

• dream-server/config/litellm/cloud.yaml — add minimax -> openai/MiniMax-M3, rename previous entry to minimax-legacy (openai/MiniMax-M2.7).
• dream-server/config/litellm/hybrid.yaml — same change for hybrid mode.
• dream-server/extensions/services/litellm/README.md — update the cloud alias table to reflect the new default.

No infra/env changes:

• Same MINIMAX_API_KEY env var, secret schema, installer wiring.
• Same https://api.minimax.io/v1 OpenAI-compatible base URL.
• Same compose / extension manifests.

AI Assistance

AI-assisted drafting for the YAML alias rename and README sync. The diff was reviewed by hand, focused validation listed below was run locally, and the human author is accountable for the change.

Release Lane

• Stable hotfix targeting release/2.5.x
• Mainline change targeting main
• Next-minor work targeting the next feature/minor release
• Not sure; reviewer should help classify

Stable hotfix reason:

N/A — additive routing change targeting main.

Changed Surface

• Docs only
• Tests only
• Dashboard UI
• Dashboard API / host agent
• Installer / bootstrap / lifecycle
• Docker Compose / service manifests
• Model routing / Hermes / capabilities
• Network exposure / auth / proxy
• Dependencies / runtime wiring

(Docs touched too — only the litellm extension README's alias table.)

Risk And Validation

• Risk level: Low
• Validation run:
  • git diff --check
  • Markdown/link sanity for docs
  • Focused tests listed below
  • Dashboard lint/test/build
  • Extension audit / compose validation
  • Release-grade fleet or scoped hardware validation
  • Stable-lane patch validation, if targeting release/2.5.x
  • Not required because: change is limited to two LiteLLM YAML files (alias rename + new entry) and one README table; no installer/compose/manifest paths touched, and the same MINIMAX_API_KEY / base URL / OpenAI-compatible protocol are reused.

Commands/results:

# YAML parses + alias table sanity
python3 -c "import yaml; ..."  ->  6 models each, minimax -> openai/MiniMax-M3

# Existing LiteLLM / cloud-mode regression suites
bash dream-server/tests/test-linux-cloud-mode.sh
  -> 15/15 PASS (cloud overlay, gateway inclusion, Hermes default, ...)

bash dream-server/tests/contracts/test-installer-contracts.sh
  -> 34/34 PASS

Operational Change Check

• This is not an operational change.
• This is an operational change and validation is recorded above.
• This is an operational change and validation is intentionally deferred for:

LiteLLM model_list entries are config data consumed at runtime; no installer phase, compose generation, or lifecycle command is touched. Existing cloud/hybrid users keep working (the minimax alias name is unchanged, only the model behind it advances; the prior model is still reachable via minimax-legacy).

Notes For Reviewers

• The naming convention <alias>-legacy mirrors what other LiteLLM gateways use when graduating a default; happy to swap to minimax-m2.7 or similar if you prefer a version-explicit name.
• M3 reuses the existing openai/... LiteLLM prefix because MiniMax's OpenAI-compatible endpoint (api.minimax.io/v1) is unchanged — no SDK swap needed.
• MINIMAX_API_KEY and base URL remain identical, so no .env.example / .env.schema.json / installer changes were necessary.

[Lightheartdevs]

Thanks for keeping the LiteLLM MiniMax aliases current while preserving an explicit legacy route for the old default.

Advisory status: looks merge-ready

What looks good:

• dream-server/config/litellm/cloud.yaml and dream-server/config/litellm/hybrid.yaml keep the public minimax alias stable, move the previous MiniMax-M2.7 mapping to minimax-legacy, and leave minimax-fast on the existing high-speed model.
• The change keeps the same MINIMAX_API_KEY and https://api.minimax.io/v1 wiring, so it does not require installer, env schema, compose, or secret-handling changes.
• dream-server/extensions/services/litellm/README.md is updated with the same alias table, so docs and runtime config agree.

What I checked:

• Static diff review of the two LiteLLM YAML configs and README table.
• Parsed both YAML files locally and confirmed minimax -> openai/MiniMax-M3, minimax-legacy -> openai/MiniMax-M2.7, and minimax-fast -> openai/MiniMax-M2.7-highspeed in both cloud and hybrid configs.
• git diff --check origin/main...origin/pr-1489: passed locally.
• GitHub checks are green, including Ruff, shell lint, dashboard API/frontend, integration-smoke, matrix-smoke, PowerShell lint, secret scan, and tier env validation.
• MiniMax's OpenAI-compatible API docs list MiniMax-M3 as a supported model on the same https://api.minimax.io/v1 base URL.

Thanks again for the tidy config update.