[heydryft]

Summary

Adds Runcrate as a gateway provider. Runcrate serves open models (DeepSeek, GLM, Qwen, Kimi, gpt-oss and others) behind one OpenAI-compatible API at https://api.runcrate.ai/v1, under their makers' vendor/model ids.

The change follows the Opper addition (#869) file for file:

• raven/providers/registry.py: a ProviderSpec at the end of the resale gateway block, reached through OpenAI's driver, told apart by runcrate.ai in the address, OPENAI_API_KEY as the driver's variable, no default model and no shortlist. Model ids carry their maker (deepseek-ai/DeepSeek-V4.1-Flash), so strip_model_prefix stays off. Runcrate's own GET /v1/models fills the picker once a key is entered.
• raven/config/schema.py: a ProvidersConfig.runcrate section.
• raven/cli/onboard_commands.py: a row in the onboarding wizard's key group.
• ui-web/src/assets/providers/runcrate.svg and ProviderMark.tsx: a one-colour mark in currentColor (mono tone, inverted in the dark theme like Opper's and vLLM's), with a LICENSES/README.md entry.
• Counts written elsewhere now say fifty-seven specs, twenty-three of them gateways (ui-web/CONTEXT.md, ProviderSide.tsx, Roles.tsx, the registry docstring and the gateway block comment).
• Tests: the pinned provider set and count, the LiteLLM name tables, NO_SHORTLIST_VENDORS, and tests/data/wire_model_baseline.json regenerated with RAVEN_UPDATE_WIRE_BASELINE=1 (the only change is the new runcrate block).

Keys: https://www.runcrate.ai/dashboard/api-keys?utm_source=Raven&utm_medium=directory&utm_campaign=gpu_pricing

Type

• Fix
• Feature
• Docs
• CI / tooling
• Refactor
• Other

Verification

• uv run --frozen --no-sync pytest -q tests/test_provider_wire_model.py tests/test_provider_catalog.py tests/test_litellm_provider_model_names.py tests/test_rpc_model.py: 967 passed, 7 skipped
• uv run --frozen --no-sync pytest tests -k 'provider or onboard or schema or registry': 4656 passed, 109 skipped
• uv run ruff check raven tests: all checks passed; ruff format --check on the touched Python files: already formatted
• In ui-web (after npm ci): npx vitest run src/components/ProviderMark.test.tsx 15 passed; npm run gen:check matches the contract; npm run type-check clean; eslint on the touched files clean
• Live check against https://api.runcrate.ai/v1 with a real key, through make_provider(Config) with agents.defaults.provider = "runcrate" and providers.runcrate.api_key set (api_base resolved from the spec to https://api.runcrate.ai/v1), a plain prompt and a one-tool prompt per model:

Model	Provider class Raven chose	Plain prompt	Tool call
deepseek-ai/DeepSeek-V4.1-Flash	OpenAIResponsesProvider	pong, stop	get_weather {'city': 'Paris'}
zai-org/GLM-5.3-Flash	LiteLLMProvider	pong, stop	get_weather {'city': 'Paris'}
openai/gpt-oss-120b	OpenAIResponsesProvider	pong, stop	get_weather {'city': 'Paris'}

Not run: the full make ci gate.

• Relevant tests pass locally
• Relevant lint / type checks pass locally
• User-facing docs or screenshots are updated when needed (no user-facing doc lists providers one by one)

Risk

• Security impact considered
• Backward compatibility considered
• Rollback path is clear for risky changes

Additive: one new provider section, nothing existing changes. Rollback is reverting the commit.

Related Issues

N/A

This PR was prepared with AI assistance: an AI coding agent wrote the change and ran the checks above.

Disclosure: I work on Runcrate.

[LivXue]

Windows compatibility review — PR #1143 @ c1dd0ac6cc880b719ccb5c0c2bda3239c2a3b25b

Scope: the full base-to-head diff at this head (14 files, +48/−10): a ProviderSpec for Runcrate reached through OpenAI's driver, its config section and onboarding group entry, the provider icon and tone, and the catalog / wire-model-name baselines and counters. Reviewed for Windows behaviour and run on native Windows.

Result: no Windows-specific issue. The change is a registry and config entry plus an SVG icon and copy-count updates; it adds no path, separator, subprocess, signal, socket, permission, or encoding surface. The spec borrows the OpenAI driver like the other resale gateways (env_key="OPENAI_API_KEY", detect_by_base_keyword="runcrate.ai"), so environment and HTTP handling are the shared paths already in use, unchanged.

Verification on native Windows 11, detached worktree D:\tmp\rv-1143 at the reviewed head, offline venv (uv sync --frozen --offline --all-extras --no-extra segmenter --dev), scratch HOME/USERPROFILE, PYTHONUTF8=1, PYTHONPATH pinned:

• uv run --no-sync pytest tests/test_litellm_provider_model_names.py tests/test_provider_catalog.py tests/test_rpc_model.py -q → 617 passed, 8 skipped. No failures, so no base comparison was needed.

Coverage gaps: the ui-web test this PR touches (ProviderMark.test.tsx) could not be run — the shared D:/projects/Raven/ui-web/node_modules is empty (0 entries, an environment issue present before this review) — so the icon-map entry, its mono tone, and the runcrate.svg asset were checked by reading only. The full suite was not rerun.