feat: add OpenAI-compatible embedding backend

[zh820952173]

Summary

Adds a third embedding backend openai alongside the existing dashscope (default) and local backends. It works with any OpenAI-compatible /embeddings endpoint — OpenAI, Zhipu/GLM (embedding-3), self-hosted vLLM/TEI, etc. — configured via base_url + api_key + model.

Default behavior is unchanged: existing users stay on dashscope with no action required.

Changes

• ultron/core/embeddings.py: EmbeddingService now supports openai. New _init_openai_client() (lazy-imports openai, validates base_url/api_key) and _embed_openai() with robust error handling for call failures, empty data, empty rows, and input/output count mismatches. The openai branch is wired into embed_text / embed_texts / is_available / get_model_info.
• ultron/config.py: new embedding_base_url and embedding_api_key fields. They fall back to the LLM endpoint/key (ULTRON_BASE_URL, ULTRON_API_KEY / OPENAI_API_KEY), so most users need no extra setup.
• ultron/api/sdk/ultron.py: passes the new config through to EmbeddingService.
• Docs & .env.example: document the new backend and env vars (ULTRON_EMBEDDING_BASE_URL, ULTRON_EMBEDDING_API_KEY) in both EN and ZH.
• Tests: tests/core/test_embeddings.py covers the openai backend success and error paths (skipped when openai is not installed).

New environment variables

Variable	Maps to	Fallback
ULTRON_EMBEDDING_BACKEND=openai	embedding_backend	dashscope
ULTRON_EMBEDDING_BASE_URL	embedding_base_url	ULTRON_BASE_URL
ULTRON_EMBEDDING_API_KEY	embedding_api_key	ULTRON_API_KEY / OPENAI_API_KEY

Notes

• One ULTRON_DATA_DIR can only use a single embedding backend/model/dimension; switching requires reset_all() (existing constraint, unchanged).
• Requires the openai package only when this backend is selected.

Test plan

• python -m pytest tests/core/test_embeddings.py
• Manual smoke test against a Zhipu embedding-3 / OpenAI endpoint

[gemini-code-assist]

Code Review

This pull request introduces support for an OpenAI-compatible embedding backend (openai) alongside the existing dashscope and local backends, updating the configuration, documentation, and SDK implementation, and adding corresponding unit tests. Feedback on these changes suggests ensuring that EmbeddingService falls back to environment variables for base_url and api_key when instantiated directly, sorting the returned embedding data by index to handle potential out-of-order responses from some providers, and isolating configuration tests from host environment variables to prevent flakiness.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.