Built-in And Custom Forge Guidance

This page explains the public fluid forge domain-guidance flow and where deeper agent customization now fits.

Public user workflow

For end users, the current public entry point is:

fluid forge
fluid forge --domain finance
fluid forge --domain healthcare
fluid forge --domain retail
fluid forge --domain telco

Built-in LLM providers include:

  • OpenAI
  • Anthropic
  • Gemini
  • Ollama

Key Forge flags

fluid forge \
  --llm-provider openai \
  --llm-model gpt-4o-mini \
  --discovery-path ./data \
  --context ./forge-context.json

Useful flags:

  • --domain
  • --llm-provider
  • --llm-model
  • --llm-endpoint
  • --discovery-path
  • --context
  • --memory / --no-memory
  • --save-memory
  • --non-interactive

What changed

Older docs sometimes described:

  • fluid forge --mode copilot
  • fluid forge --mode agent --agent <name>

Those are no longer the public, primary docs path. Current docs lead with fluid forge plus --domain when you want built-in domain guidance.

Built-in domain guidance

The built-in domains are still backed by declarative specs inside forge-cli, but users interact with them through --domain.

Current built-in domains:

  • finance
  • healthcare
  • retail
  • telco

When custom agent work still matters

Contributor-level customization can still matter if you are extending forge-cli itself and want to:

  • add a new built-in domain
  • change domain-specific prompts or defaults
  • alter how domain guidance is sourced from internal specs

That work is implementation detail for CLI contributors, not the primary docs path for day-to-day users.