Pipeline Mode & LLM Access

Pipeline modes

S2S has two pipeline modes, set via pipelineMode in .s2s/config/runtime.json.

Chat-native mode (default)

{ "pipelineMode": "chat-native" }

In chat-native mode (the default), s2s stage <stage> outputs a structured context package to the active chat session. The chat AI reads the package, generates the required artifact, writes it to .s2s/artifacts/, and runs s2s stage <stage> --submit to record completion. S2S never calls an LLM directly. No llm.json is created or required.

Standalone mode

{ "pipelineMode": "standalone" }

In standalone mode, s2s stage <stage> calls the LLM API directly and produces the artifact autonomously. Use this for headless CI pipelines or environments without an interactive chat session.

Standalone mode requires .s2s/config/llm.json. Use s2s config edit to configure it — selecting standalone mode will prompt for provider, model, and API key settings.

---

LLM access modes (standalone only)

The following modes are only relevant when pipelineMode: 'standalone'. In chat-native mode these settings have no effect.

API mode (mode: "api")

Use direct provider SDK access with API keys.

{
  "mode": "api",
  "provider": "anthropic",
  "model": "claude-sonnet-4-5-20250929",
  "apiKeyEnvVar": "ANTHROPIC_API_KEY"
}

Notes:

• Supported providers: anthropic, openai.
• Requires an API key in the configured env var.

OpenAI-compatible mode (mode: "openai_compatible")

Use any OpenAI-compatible endpoint.

{
  "mode": "openai_compatible",
  "provider": "openai",
  "model": "gpt-5.4",
  "baseURL": "https://api.openai.com/v1",
  "apiKeyEnvVar": "OPENAI_API_KEY"
}

Notes:

• Requires API key for the configured endpoint.
• Useful for gateways or self-hosted OpenAI-compatible services.

---

Recommendation

Use s2s config edit to switch between modes interactively. s2s doctor validates that the required credentials are present for the selected mode.