test(e2e): add deepl conformance cases for ai-proxy

[123123213weqw]

Ⅰ. Describe what this PR did

Add a non-streaming e2e conformance case for the deepl provider of the ai-proxy wasm plugin. The case sends an OpenAI-format chat completion request (model Free) and verifies that ai-proxy transforms it to the DeepL /v2/translate request shape (text array + target_lang), routes to the Free host api-free.deepl.com, and converts the DeepL translation response back into an OpenAI chat.completion (with detected_source_language mapped to message.name).

It reuses the same conformance scaffolding as the existing openai/hunyuan/etc. cases:

• New Ingress wasmplugin-ai-proxy-deepl (host api-free.deepl.com) pointing at llm-mock-service.
• New matchRule with type: deepl, apiTokens: [fake_token], targetLang: EN (required by deeplProviderInitializer.ValidateConfig).
• A single non-streaming test case asserting the OpenAI-format response, ignoring the volatile created and usage fields.

Ⅱ. Does this pull request fix one issue?

fixes #1723

Ⅲ. Why don't you add test cases (unit test/integration test)?

This PR itself adds the e2e conformance test case for the deepl provider.

Ⅳ. Describe how to verify it

Run the conformance suite against a cluster with the llm-mock (latest image, which ships the deepl mock) service and the ai-proxy wasm plugin built:

go test ./test/e2e/conformance/tests/... -run TestGoWasmAiProxy

The deepl case 1: non-streaming request case should pass.

Ⅴ. Special notes for reviews

• Only a non-streaming case is added; DeepL's /v2/translate is non-streaming, so the deepl provider has no streaming path.
• The request body uses model: "Free" (the deepl provider requires Free or Pro; Free selects the api-free.deepl.com host).
• targetLang: EN is mandatory in the provider config and is therefore set in the matchRule.
• JsonBodyIgnoreFields: ["created", "usage"] excludes the timestamp (time.Now()) and the always-null usage (deepl does not populate token usage) from the comparison.

Ⅵ. AI Coding Tool Usage Checklist (if applicable)

Please check all applicable items:

• For new standalone features (e.g., new wasm plugin or golang-filter plugin):

  • I have created a design/ directory in the plugin folder
  • I have added the design document to the design/ directory
  • I have included the AI Coding summary below

• For regular updates/changes (not new plugins):

  • I have provided the prompts/instructions I gave to the AI Coding tool below
  • I have included the AI Coding summary below

AI Coding Prompts (for regular updates)

AI Coding Summary

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@1547826). Learn more about missing BASE report.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4121   +/-   ##
=======================================
  Coverage        ?   50.70%           
=======================================
  Files           ?       88           
  Lines           ?    13398           
  Branches        ?        0           
=======================================
  Hits            ?     6794           
  Misses          ?     6156           
  Partials        ?      448           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[CH3CHO]

2026-07-11T02:49:35.9073788Z     --- FAIL: TestHigressConformanceTests/WasmPluginAiProxy (387.23s)
2026-07-11T02:49:35.9075247Z         --- FAIL: TestHigressConformanceTests/WasmPluginAiProxy/WasmPlugins_ai-proxy (387.09s)

The test doesn't pass.

[123123213weqw]

Thanks for flagging the higress-wasmplugin-test (GO) failure. I tracked it down and pushed a fix.

Root cause

The deepl conformance case's expected response body was invalid JSON. The
"finish_reason":null field was misplaced: it ended up outside the single
choice object but inside the choices array:

- {"choices":[{"index":0,...,"message":{...}},"finish_reason":null],...}
+ {"choices":[{"index":0,...,"message":{...},"finish_reason":null}],...}

The e2e HTTP comparator (test/e2e/conformance/utils/http/http.go) parses the
expected body with json.Unmarshal before it even looks at the captured
response. Because the expected body failed to parse
(invalid character ':' after array element), the case failed immediately with
failed to unmarshall ExpectedResponse body ... — regardless of what the
gateway/mock returned. That is exactly what surfaced as the GO wasmplugin-test
failure.

Fix

Moved "finish_reason":null back inside the choice object so the document is
well-formed (test/e2e/conformance/tests/go-wasm-ai-proxy.go).

Verification

• Reproduced the deepl provider's responseDeepl2OpenAI success branch and
  marshaled its chatCompletionResponse: the output is
  {"choices":[{"index":0,"message":{"name":"EN","role":"assistant","content":"..."},"finish_reason":null,"logprobs":null}],"created":...,"model":"Free","object":"chat.completion","usage":null}.
• The corrected expected body now parses as valid JSON and is structurally
  equal to that output (after the existing JsonBodyIgnoreFields: ["created","usage"]),
  confirmed both via a focused provider test and an independent JSON parse/equality check.
• gofmt -e reports no syntax errors on the edited file.

The provider/YAML wiring (Ingress api-free.deepl.com → llm-mock-service and
the deepl provider config with targetLang: EN) is unchanged; only the
malformed expected-response literal was corrected.