🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Complete the Scaleway discovery contract in every locale.

These sections describe a chat-model allowlist but do not explicitly identify the documented Chat Completions allowlist or fail-closed exclusion of unknown, Responses-only, and other media models. The Korean section also omits the custom-provider requirement for Project-qualified URLs and dedicated deployments.

• docs-site/src/content/docs/guides/providers.md#L269-L273: State the documented Chat Completions allowlist and explicit fail-closed exclusions.
• docs-site/src/content/docs/ja/guides/providers.md#L192-L195: State the documented Chat Completions allowlist and explicit fail-closed exclusions.
• docs-site/src/content/docs/ko/guides/providers.md#L191-L194: Add the documented Chat Completions and fail-closed exclusions, and state that Project-qualified URLs and dedicated deployments require a custom provider.
• docs-site/src/content/docs/ru/guides/providers.md#L202-L206: State the documented Chat Completions allowlist and explicit fail-closed exclusions.
• docs-site/src/content/docs/zh-cn/guides/providers.md#L180-L183: State the documented Chat Completions allowlist and explicit fail-closed exclusions.

As per path instructions, localized provider catalogs must stay synchronized and document authenticated allowlist intersection, bounded discovery, fail-closed exclusions, and provider-specific endpoint scope.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@docs-site/src/content/docs/guides/providers.md` around lines 269 - 273,
Synchronize the Scaleway discovery sections in
docs-site/src/content/docs/guides/providers.md:269-273,
docs-site/src/content/docs/ja/guides/providers.md:192-195,
docs-site/src/content/docs/ko/guides/providers.md:191-194,
docs-site/src/content/docs/ru/guides/providers.md:202-206, and
docs-site/src/content/docs/zh-cn/guides/providers.md:180-183. In each locale,
identify the documented Chat Completions allowlist, authenticated-model
intersection, bounded discovery, and fail-closed exclusion of unknown,
Responses-only, embeddings, transcription, and other media models; additionally
update the Korean section to state that Project-qualified URLs and dedicated
deployments require a custom provider.

Source: Path instructions

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

#!/bin/bash
set -eu

echo '--- registry context ---'
sed -n '540,630p' src/providers/registry.ts

echo '--- test context ---'
sed -n '210,275p' tests/digitalocean-scaleway-provider.test.ts

echo '--- model and route references ---'
rg -n -C 3 'gpt-oss-120b|SCALEWAY_SERVERLESS_CHAT_MODELS|openai-chat|Responses API|responses' src tests

#!/bin/bash
set -eu

echo '--- registry allowlist and nearby definitions ---'
sed -n '560,615p' src/providers/registry.ts

echo '--- exact gpt-oss references ---'
rg -n -C 8 'gpt-oss-120b' src/providers/registry.ts tests/digitalocean-scaleway-provider.test.ts

echo '--- relevant test block ---'
sed -n '220,265p' tests/digitalocean-scaleway-provider.test.ts

echo '--- registry discovery and route construction symbols ---'
rg -n 'SCALEWAY_SERVERLESS_CHAT_MODELS|digitalocean|scaleway|openai-chat|routed|catalog|models' src/providers/registry.ts tests/digitalocean-scaleway-provider.test.ts

#!/bin/bash
set -eu

python3 - <<'PY'
from pathlib import Path

for name in ("src/providers/registry.ts", "tests/digitalocean-scaleway-provider.test.ts"):
    print(f"--- {name}: matching lines with numbers ---")
    for i, line in enumerate(Path(name).read_text().splitlines(), 1):
        if any(term in line for term in ("gpt-oss-120b", "SCALEWAY_SERVERLESS_CHAT_MODELS", "scalewayModels", "openai-chat")):
            start = max(1, i - 6)
            end = i + 10
            lines = Path(name).read_text().splitlines()
            for j in range(start, min(end, len(lines)) + 1):
                print(f"{j}: {lines[j-1]}")
            print()
PY

#!/bin/bash
set -eu

echo '--- Scaleway fixture entry ---'
rg -n -C 5 'gpt-oss-120b' tests/fixtures/scaleway-models.json

echo '--- tool-request test ---'
sed -n '255,315p' tests/digitalocean-scaleway-provider.test.ts

echo '--- OpenAI Chat adapter request construction ---'
rg -n -C 12 'tools|tool_choice|parallel_tool_calls|buildRequest' src/adapters/openai-chat.ts

echo '--- relevant provider contract types and routing metadata ---'
rg -n -C 8 'parallelToolCalls|adapter.*openai-chat|modelDiscovery|filter' src/types.ts src/providers src/codex

#!/bin/bash
set -eu

python3 - <<'PY'
import json
from pathlib import Path

fixture = json.loads(Path("tests/fixtures/scaleway-models.json").read_text())
rows = fixture.get("data", fixture) if isinstance(fixture, dict) else fixture
for row in rows:
    if isinstance(row, dict) and row.get("id") == "gpt-oss-120b":
        print(json.dumps(row, indent=2, sort_keys=True))
PY

Exclude gpt-oss-120b from Scaleway’s tool-capable catalog.

Scaleway does not support tool calling for this model through Chat Completions. The openai-chat adapter sends tools to /chat/completions, so routed agent requests fail. Remove it from SCALEWAY_SERVERLESS_CHAT_MODELS and update tests/digitalocean-scaleway-provider.test.ts to assert that mixed-catalog discovery excludes it.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/providers/registry.ts` around lines 586 - 598, Remove "gpt-oss-120b" from
the SCALEWAY_SERVERLESS_CHAT_MODELS catalog in src/providers/registry.ts. Update
the mixed-catalog discovery assertion in
tests/digitalocean-scaleway-provider.test.ts to verify that this model is
excluded while preserving the existing supported-model assertions.

Source: Path instructions

Advertise image input for Pixtral

When a user selects scaleway/pixtral-12b-2409 for an image request, this allowlist exposes the multimodal model but the Scaleway registry entry provides no modelInputModalities metadata. Because /models supplies only generic OpenAI-shaped fields, catalog normalization consequently marks this row as input_modalities: ["text"], causing Codex App to reject image attachments before they reach the proxy. Add the model's ['text', 'image'] metadata to the canonical registry and cover the gathered catalog row in the provider test.

AGENTS.md reference: src/AGENTS.md:L18-L19

Useful? React with 👍 / 👎.

Disable reasoning controls for non-reasoning models

When Codex routes a request with reasoning.effort to an admitted non-reasoning model such as meta-llama/Meta-Llama-3.1-8B-Instruct, the DigitalOcean entry has neither a conservative reasoningEfforts: [] default nor a noReasoningModels classification. The catalog therefore advertises the full reasoning ladder and openai-chat forwards reasoning_effort, even though this model does not support that control; strict upstream validation can reject the request, while permissive handling silently ignores a user-visible option. Define the docs-backed reasoning models explicitly and suppress the field for the remainder, with a request regression test that supplies a reasoning effort.

AGENTS.md reference: src/AGENTS.md:L18-L19

Useful? React with 👍 / 👎.