feat(embedding): add Ollama provider support for local embedding#644
feat(embedding): add Ollama provider support for local embedding#644qin-ctx merged 1 commit intovolcengine:mainfrom
Conversation
|
|
| if self.provider == "openai": | ||
| if not self.api_key: | ||
| # Allow missing api_key when api_base is set (e.g. local OpenAI-compatible servers) | ||
| if not self.api_key and not self.api_base: |
There was a problem hiding this comment.
[Bug] Config validation now allows provider="openai" with api_base but no api_key. However, OpenAIDenseEmbedder.__init__ (openai_embedders.py:65-66) still hard-requires api_key:
if not self.api_key:
raise ValueError("api_key is required")Also, the OpenAI factory lambda passes cfg.api_key directly (which will be None), unlike the Ollama factory which uses cfg.api_key or "ollama" as a fallback.
This means a user configuring provider="openai" + api_base + no api_key (as shown in Test 7) will pass config validation but crash at runtime when the embedder is created.
Suggested fix: add a fallback in the OpenAI factory lambda (e.g., cfg.api_key or "no-key") and relax the check in OpenAIDenseEmbedder.__init__ to allow missing api_key when api_base is set.
- Add 'ollama' as a supported embedding provider - Ollama runs locally via OpenAI-compatible API, no API key required - Allow OpenAI provider to work without api_key when api_base is set (supports local OpenAI-compatible servers like vLLM, LocalAI) - Add configuration example and tests for Ollama provider This enables fully local embedding deployment without cloud API keys.
Summary
This enables fully local embedding deployment without cloud API keys.