Module providers 

Provider factory — builds embedding and LLM providers from environment config.

Provider priority (LLM and embeddings)

1. Mistral native (MISTRAL_API_KEY set) — calls api.mistral.ai directly. Preferred for the European deployment stack: no Google Cloud dependency, EU-hosted, GDPR-compliant. Set MISTRAL_MODEL to override the model.

2. OpenAI (OPENAI_API_KEY set) — calls OpenAI or any compatible API. Override base URL with OPENAI_API_BASE for Ollama, Together, etc.

3. Anthropic (ANTHROPIC_API_KEY set) — calls api.anthropic.com. LLM only (no embeddings).

4. Vertex AI (VERTEX_AI_PROJECT + credentials set) — Google Cloud. Supports Mistral, Gemini, and Claude via Vertex Model Garden.

5. Mock (fallback) — deterministic hash-based embeddings, echo LLM. Used in tests and when no cloud credentials are present.

Environment variables

Variable	Description	Default
MISTRAL_API_KEY	Native Mistral API key (console.mistral.ai)	—
MISTRAL_MODEL	Mistral model name	mistral-small-latest
MISTRAL_EMBEDDING_MODEL	Mistral embedding model	mistral-embed
OPENAI_API_KEY	OpenAI API key (or compatible)	—
OPENAI_MODEL	OpenAI model name	gpt-4o-mini
OPENAI_API_BASE	OpenAI-compatible base URL	https://api.openai.com/v1
OPENAI_EMBEDDING_MODEL	OpenAI embedding model	text-embedding-3-small
ANTHROPIC_API_KEY	Anthropic API key	—
ANTHROPIC_MODEL	Anthropic model name	claude-sonnet-4-20250514
VERTEX_AI_PROJECT	GCP project ID	required (or GOOGLE_CLOUD_PROJECT)
VERTEX_AI_LOCATION	GCP region	europe-west1 (or GOOGLE_CLOUD_LOCATION)
VERTEX_AI_TOKEN	Static GCP auth token	auto-detect
VERTEX_AI_MODEL	Vertex model name	mistral-small-2503
NOETHER_LLM_PROVIDER	Force: mistral | openai | anthropic | vertex | mock	auto
NOETHER_EMBEDDING_PROVIDER	Force: mistral | openai | vertex | mock	auto

Functions

build_embedding_provider

Build the best available embedding provider based on env config.

build_llm_provider

Build the best available LLM provider based on env config.