OpenAI

OpenAI is the catalog default for both chat and embeddings. You need an API key and little else. This page lists the variables, the default models and how to check the connection.

When to use

You want the highest extraction and answer quality with the least setup.
You accept per-token costs and sending text to OpenAI.

Configure

export EDGEQUAKE_LLM_PROVIDER=openai
export OPENAI_API_KEY=sk-...
export EDGEQUAKE_LLM_MODEL=gpt-5.4-mini
export EDGEQUAKE_EMBEDDING_PROVIDER=openai
export EDGEQUAKE_EMBEDDING_MODEL=text-embedding-3-small

Variable	Purpose	Default
`OPENAI_API_KEY`	Required. Sent as `Authorization: Bearer`.	none
`EDGEQUAKE_LLM_MODEL`	Chat model	catalog default `gpt-4.1-mini`; the quickstart picks `gpt-5.4-mini`
`EDGEQUAKE_EMBEDDING_MODEL`	Embedding model	`text-embedding-3-small` (1536 dimensions)
`OPENAI_BASE_URL`	Optional. Send traffic through an OpenAI-compatible proxy or gateway.	`https://api.openai.com/v1`

Models

Chat: gpt-5.4-mini is the quickstart choice. gpt-5.4-nano and gpt-5.4 are also offered.
Embeddings: text-embedding-3-small returns 1536-dimension vectors. text-embedding-3-large returns 3072.

All model names come from edgequake/models.toml. Changing the embedding model changes the vector size, so rebuild stored embeddings after you change it (Roles).

Verify

Two checks answer two different questions:

Check	Calls OpenAI?	Answers
`GET /health`	No. It only checks that `OPENAI_API_KEY` is set.	Is the key configured?
`POST /api/v1/providers/test`	Yes	Does the key work, and do the models respond?

curl -sS http://127.0.0.1:8080/health

Look for "llm_provider": true under components and your model under providers.llm. Use the probe for a live result (Test a server).

%%{init: {"theme":"base","themeVariables":{"primaryColor":"#E0E7FF","primaryBorderColor":"#6366F1","primaryTextColor":"#1E1B4B","secondaryColor":"#D1FAE5","secondaryBorderColor":"#10B981","secondaryTextColor":"#064E3B","tertiaryColor":"#FEF3C7","tertiaryBorderColor":"#F59E0B","tertiaryTextColor":"#6B7A90","lineColor":"#7A889C","clusterBkg":"rgba(99,102,241,0.07)","clusterBorder":"#A5B4FC","noteBkgColor":"#FEF9C3","noteTextColor":"#422006","textColor":"#6B7A90","titleColor":"#6B7A90","signalColor":"#7A889C","signalTextColor":"#6B7A90","loopTextColor":"#6B7A90","edgeLabelBackground":"#F1F5F9","actorLineColor":"#94A3B8"}}}%%
%% eq-theme:v1
sequenceDiagram
  participant U as Operator
  participant EQ as EdgeQuake
  participant OA as OpenAI API
  U->>EQ: GET /health
  EQ-->>U: key present (no network call)
  U->>EQ: POST /api/v1/providers/test
  EQ->>OA: GET /v1/models
  EQ->>OA: POST /v1/chat/completions
  EQ->>OA: POST /v1/embeddings
  OA-->>EQ: responses
  EQ-->>U: live test result

Troubleshoot

/health says the provider is ready, but extraction fails with 401. The key is set but invalid. /health does not test it. Run the probe.
Probe reports model_not_found. The model name is not available to your key. Check the name in EDGEQUAKE_LLM_MODEL.
Requests go to the wrong host. OPENAI_BASE_URL points at a proxy. Unset it or fix the URL.

Notes

To store the key in the database instead of the environment, create a Connection with api_shape=openai_chat (see Provider security).
For another OpenAI-shaped server, see Generic OpenAI-shaped server.

Related: Roles, Anthropic, Providers overview.