Skip to content

README.md

Latest commit

 

History

History
68 lines (48 loc) · 1.65 KB

README.md

File metadata and controls

68 lines (48 loc) · 1.65 KB
Raw

Sigma Examples

These examples are ordinary Go programs. They use deterministic sigmatest providers by default, so they compile and run without live credentials.

Run all examples through the repository test suite from the repository root:

mise run go:test

Run one example:

go run ./examples/chat
go run ./examples/stream
go run ./examples/tools
go run ./examples/images
go run ./examples/cancel
go run ./examples/custom-model

Live Text Provider

examples/chat defaults to sigmatest. To run it against OpenAI Chat Completions instead:

SIGMA_EXAMPLE_PROVIDER=openai OPENAI_API_KEY=... go run ./examples/chat

Optional:

SIGMA_EXAMPLE_MODEL=gpt-4o-mini

Live Fireworks Firepass Demo

examples/fireworks contains a live CLI demo for the Fireworks Firepass Kimi router. It requires FIREWORKS_API_KEY.

FIREWORKS_API_KEY=... go run ./examples/fireworks

Local OpenAI-Compatible Endpoint

examples/custom-model registers a custom OpenAI-compatible model and exits without making a network call unless SIGMA_LOCAL_BASE_URL is set.

Examples:

SIGMA_LOCAL_BASE_URL=http://localhost:11434/v1 SIGMA_LOCAL_MODEL=llama3.2 go run ./examples/custom-model
SIGMA_LOCAL_BASE_URL=http://localhost:1234/v1 SIGMA_LOCAL_MODEL=local-model go run ./examples/custom-model

Optional:

SIGMA_LOCAL_API_KEY=local

The custom model demonstrates compatibility overrides for local Ollama/vLLM/LM Studio-style endpoints: unsupported store/developer-role fields, streaming usage, strict tools, reasoning, prompt cache control, and tool-result message quirks are configured explicitly on sigma.OpenAICompletionsCompat.