Files
ragflow/conf/models/volcengine.json
Haruko386 94f82acd03 Fix(Go): prevent global state pollution in local model connection check (#14669)
### What problem does this PR solve?

1. **Fix Global State Pollution in Local Providers (Critical Bug):** -
Resolved a severe concurrency and architecture issue in
`model_service.go`. Previously, `ListSupportedModels` would permanently
overwrite the global provider singleton with a localized URL instance
(`driver.NewInstance`). This caused cross-request contamination in
multi-tenant environments.
- Fixed `CheckProviderConnection` for local models (LM Studio, vLLM,
Ollama). It now properly creates a localized driver copy and injects the
`base_url` before testing the connection, entirely eliminating the
false-positive `missing base URL` error without polluting the global
state.
2. **Implement `VolcEngine` Embeddings:** - Fully implemented the
`Encode` method for the `volcengine` provider, enabling text embedding
capabilities for VolcEngine models.
3. **Enhance Region Validation in `SiliconFlow`:** - Added a strict
empty string check (`*apiConfig.Region != ""`) alongside the existing
`nil` check when parsing regions. This ensures that if an empty string
is passed, the system safely falls back to the `"default"` region,
preventing malformed URL requests and `unsupported protocol scheme`
errors.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [x] New Feature (non-breaking change which adds functionality)
2026-05-08 15:54:27 +08:00

32 lines
623 B
JSON

{
"name": "VolcEngine",
"url": {
"default": "https://ark.cn-beijing.volces.com/api/v3"
},
"url_suffix": {
"chat": "chat/completions",
"files": "files",
"embedding": "embeddings/multimodal"
},
"class": "volcengine",
"models": [
{
"name": "doubao-seed-2-0-pro-260215",
"max_tokens": 262144,
"model_types": [
"chat"
],
"thinking": {
"default_value": true,
"clear_thinking": true
}
},
{
"name": "doubao-embedding-vision-250615",
"max_tokens": 131072,
"model_types": [
"embedding"
]
}
]
}