feat(providers): Support DeepInfra as an LLM provider

[ats3v]

What does this PR do?

Adds DeepInfra as a first-class LLM provider, giving users access to 100+ top open source and frontier models (GLM-5, DeepSeek-V3+, Qwen3+, Kimi-K2+, Nemotron, etc.) via a single API key.

DeepInfra is one of the largest open-model inference platforms and a top provider on OpenRouter. This integration lets users connect directly — lower latency, simpler billing, no middleman.

The implementation follows the exact same pattern as HuggingFace, KiloCode, and other API-key providers. Purely additive — no changes to existing provider behavior.

Related Issue

N/A — new provider integration.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

Source:

• hermes_cli/models.py — Add _fetch_deepinfra_catalog, _fetch_deepinfra_models_by_tag, _fetch_deepinfra_pricing; switch to ?filter=true&sort_by=hermes; empty out the static fallback list.
• agent/model_metadata.py — Extend _extract_pricing() to recognize DeepInfra's $/MTok pricing keys.
• tools/tts_tool.py — Register deepinfra in BUILTIN_TTS_PROVIDERS; parametrize _generate_openai_tts with optional creds; thin _generate_deepinfra_tts shim.
• tools/transcription_tools.py — Parametrize _transcribe_openai with optional creds + label; thin _transcribe_deepinfra shim; gate + auto-detect chain.
• hermes_cli/tools_config.py — Add the "DeepInfra TTS" picker row.
• hermes_cli/doctor.py — Allowlist deepinfra in providers_accepting_vendor_slugs (kills a false-positive warning).
• hermes_cli/config.py — Mention deepinfra in the stt.provider doc-string.
• plugins/model-providers/deepinfra/__init__.py (new) — ProviderProfile registration; default_aux_model="stepfun-ai/Step-3.5-Flash".
• plugins/model-providers/deepinfra/plugin.yaml (new) — kind: model-provider manifest.
• plugins/image_gen/deepinfra/__init__.py (new) — DeepInfraImageGenProvider; live catalog discovery via the image-gen tag; SDK-based generate().
• plugins/image_gen/deepinfra/plugin.yaml (new) — kind: backend manifest.

Docs:

• cli-config.yaml.example — Document optional tts.deepinfra / stt.deepinfra / image_gen.deepinfra blocks.

Tests (invariants only — no model-id snapshots):

• tests/hermes_cli/test_api_key_providers.py — Add tag-filter, pricing, profile classes; reshape existing tests for the catalog-as-source-of-truth model.
• tests/plugins/image_gen/test_deepinfra_provider.py (new) — 6 tests covering availability, catalog filter, no-model error, SDK delegation, SDK-exception handling.
• tests/tools/test_tts_deepinfra.py (new) — 5 tests covering registration, offline contract, no-model raise, delegation, dispatch.
• tests/tools/test_transcription_deepinfra.py (new) — 5 tests covering offline contract, gating, no-model error, delegation, dispatch.

How to Test

1. Setup:

   # Get a key at https://deepinfra.com/dash/api_keys
   echo "DEEPINFRA_API_KEY=<your-key>" >> ~/.hermes/.env
   

2. Provider selection:

hermes model # Select "DeepInfra" from the menu

3. Direct chat:

hermes chat --provider deepinfra -m "deepseek-ai/DeepSeek-V3-0324" -q "Hello"

4. Tool calling (agent loop):

hermes chat --provider deepinfra -m "deepseek-ai/DeepSeek-V3-0324" -t terminal -q "List files in the current directory"

5. Auto-detection:

With only DEEPINFRA_API_KEY set (no other provider keys), hermes chat auto-selects DeepInfra.

6. Unit tests:

pytest tests/test_api_key_providers.py -k deepinfra -v

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: Ubuntu 24.04 on WSL2.0

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Screenshots / Logs