OpenAI-Compatible Provider

Provider guide · Configuration · 简体中文

Minimal Setup

export SIGIL_OPENAI_COMPATIBLE_API_KEY="sk-..."
sigil

config_version = 2

[agent]
connection = "custom-default"
model = "gpt-4.1"

[connections.custom-default]
label = "Custom endpoint"
provider = "custom"
protocol = "chat_completions"
base_url = "https://api.openai.com/v1"
credential = { source = "environment", name = "SIGIL_OPENAI_COMPATIBLE_API_KEY" }

See openai-compatible.toml for a copyable file.

Authentication

The example binds only this connection to SIGIL_OPENAI_COMPATIBLE_API_KEY. The secure credential store is also available in setup and /config; plaintext api_key connection fields are rejected in V2. organization and project are optional connection options.

Options And Visible Limits

The endpoint and model must support streamed Chat Completions and tool calls. Each custom endpoint owns its exact URL, protocol, credential, and model catalog; Sigil never borrows a model or credential from another connection.

Generic compatible endpoints do not accept image attachments through Sigil, even if a specific service offers its own multimodal extension. DeepSeek-only FIM and strict-tool settings also do not apply here.

Verify

Run sigil doctor and confirm default=custom-default/gpt-4.1, the chat_completions protocol, expected /v1 endpoint, credential source, and readiness.

Common Problems

• 404: point base_url at the compatible /v1 root.
• Authentication: check the bound environment variable or repair this connection in /config.
• Tool calls fail: confirm endpoint and model support streamed tool calls.
• Wrong account: review organization, project, and provider dashboard settings.

Next: Return to Providers.