Greptile Summary

This PR adds EvoLink as a new channel provider, following the same pattern used by siliconflow, aihubmix, and opencode_go. All required layers — ent schema/enum, migration, GraphQL, backend transformer wiring, frontend zod schema, channel/provider configs, icon, and both i18n locales — are updated in a single coherent change.

• evolink routes through the OpenAI-compatible transformer (openai.PlatformOpenAI) with openAICompatibleDefaultEndpoints, matching the default surface of similar aggregator providers.
• evolink_anthropic routes through anthropic.PlatformDirect with a single AnthropicMessage default endpoint, consistent with aihubmix_anthropic, minimax_anthropic, and others in the same case group.

Confidence Score: 5/5

• Safe to merge — the change is purely additive and does not modify any existing channel paths.
• Every required registration point is covered: ent schema, generated enum constants, migration schema, GraphQL enum, backend transformer switch cases, default endpoint map, frontend zod schema, channel and provider configs, icon component, and both i18n locale files. The two new types correctly reuse the established OpenAI-compatible and Anthropic PlatformDirect transformer paths without touching any existing channel logic. No regressions are possible from purely additive enum and map entries.
• No files require special attention.

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Incoming LLM Request] --> B{Channel Type?}
    B -->|evolink| C[openai.PlatformOpenAI Transformer]
    B -->|evolink_anthropic| D[anthropic.PlatformDirect Transformer]
    C --> E["https://direct.evolink.ai/v1\n/chat/completions, /embeddings,\n/images/*, /videos/*"]
    D --> F["https://direct.evolink.ai\n/v1/messages"]
    E --> G[EvoLink Upstream API]
    F --> G

_{Reviews (2): Last reviewed commit: "feat: add EvoLink as a built-in provider..." | Re-trigger Greptile}

Live validation evidence

Both channel types verified end-to-end against the real EvoLink API (API key omitted; requests run with proxy env vars unset to mirror a clean CI environment).

1) evolink — OpenAI Chat Completions

Request

curl -X POST https://direct.evolink.ai/v1/chat/completions \
  -H "Authorization: Bearer $EVOLINK_API_KEY" -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.2","messages":[{"role":"user","content":"Reply with exactly: AxonHub Evolink OK"}],"max_tokens":50}'

Response — HTTP 200

{
  "id": "chatcmpl-DmgtXzwa4Xb0NkIX4ctF9UthkTYIR",
  "object": "chat.completion",
  "model": "gpt-5.2-2025-12-11",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": { "role": "assistant", "content": "AxonHub Evolink OK" }
    }
  ],
  "usage": { "prompt_tokens": 16, "completion_tokens": 10, "total_tokens": 26 }
}

2) evolink_anthropic — Anthropic Messages

Request

curl -X POST https://direct.evolink.ai/v1/messages \
  -H "Authorization: Bearer $EVOLINK_API_KEY" -H "Content-Type: application/json" \
  -H "anthropic-version: 2023-06-01" \
  -d '{"model":"claude-opus-4-8","max_tokens":50,"messages":[{"role":"user","content":"Reply with exactly: AxonHub Evolink Anthropic OK"}]}'

Response — HTTP 200

{
  "id": "msg_bdrk_01LGvyN8CXvNwTUz6XfnA7yP",
  "type": "message",
  "role": "assistant",
  "model": "claude-opus-4-8",
  "content": [ { "type": "text", "text": "AxonHub Evolink Anthropic OK" } ],
  "stop_reason": "end_turn",
  "usage": { "input_tokens": 30, "output_tokens": 18 }
}

Backend checks

• make generate — regenerated ent + GraphQL ✅
• go build ./... ✅
• go test ./internal/server/biz/ -run "Channel|Endpoint" ✅

Rebased onto the latest unstable to resolve the merge conflict — the branch now coexists cleanly with the newly-merged opencode_go_anthropic channel type (#1791); both the channel enum and the i18n entries keep both types. The ent/GraphQL generated files were regenerated via make generate from the merged schema (not hand-edited). PR is MERGEABLE again.

Re-verified locally after the rebase:

• make generate + go build ./... ✅
• go test ./internal/server/biz/ -run "Channel|Endpoint" ✅

The Test/Build/Lint workflows are currently in action_required (first-time fork-contributor gate) — would a maintainer mind approving the workflow runs when convenient? Happy to address anything they surface. Thanks! 🙏