feat(providers): add Mistral AI provider plugin

[merouaneagar]

Summary

Production-grade Mistral AI provider profile (api.mistral.ai) as a
bundled model-provider plugin, addressing long-standing requests
(#42357, #20859, #61160).

Wire-format quirks handled

• reasoning_effort accepts ONLY none/high; a global Hermes medium is
  rejected (HTTP 400 / code 3051). Positive efforts are OMITTED
  rather than clamped to high: with high, Mistral streams
  delta.content as a JSON list (thinking + text parts), which crashes
  the Hermes 0.19 stream parser ("".join(content_parts) ->
  TypeError: sequence item 0: expected str instance, list found) --
  the "Stream interrupted" symptom.
• reasoning_effort is rejected entirely on models without the
  reasoning capability -- even none (HTTP 400 / code 3051). The
  profile gates per family with exact exclusions for
  mistral-medium-2505/2508.
• extra_body.think (Ollama quirk) is never emitted (HTTP 422).
• Vision only on vision-capable models; supports_vision=True
  provider-wide so Hermes routes images correctly.

Included

• plugins/model-providers/mistral/init.py + plugin.yaml
• tests/providers/test_mistral_profile.py -- 12 tests (discovery,
  reasoning mapping contracts, never-emits-think)
• Validated E2E against the live API: 34 chat models (basic, stream,
  tools, reasoning, vision); 32/34 OK -- the 2 failures are
  account-gated (labs-leanstral-*, HTTP 403).

Closes #42357

[teknium1]

Thanks for the careful provider profile and focused coverage.

This automated hermes-sweeper review is closing this under the standing in-tree-provider-integration policy:

• The PR adds a new bundled third-party inference-provider integration at plugins/model-providers/mistral/ (fb94b17effa6).
• AGENTS.md:797-809 requires new third-party product and niche LLM-provider integrations to ship as standalone plugin repositories, rather than adding another in-tree plugin.
• The supported route already exists: providers/__init__.py:163-171 discovers user-installed provider plugins from $HERMES_HOME/plugins/model-providers/.

Please publish this as a standalone Mistral provider plugin users can install into ~/.hermes/plugins/ (or via a pip entry point); it can be promoted in #plugins-skills-and-skins. This is a maintenance/coupling decision, not a judgment on the implementation quality.

---

Closed as not-planned per standing maintainer policy (in-tree-provider-integration). This is a design-direction decision, not a code-quality judgment — see the Contribution Rubric in AGENTS.md for what the project is looking for. If you believe this policy was misapplied to your change, comment here and a maintainer will take a look.

[merouaneagar]

Published as a standalone plugin repository per the in-tree policy: https://github.com/merouaneagar/hermes-mistral-provider

Install with: hermes plugins install merouaneagar/hermes-mistral-provider

Same profile, same 12 tests, validated E2E on 34 chat models. Thanks for the quick review.