Advanced
The pages in this folder cover the mechanism underneath the everyday commands: running Forge unattended in CI, the security boundaries, the AI authoring stack, and the runtime reference. The sidebar shows them under Operations & advanced and Contract & runtime reference; this page lists them in the same groups, with the question each one answers.
None of it is required reading. Start at Concepts for the model and CLI reference for a command, and come here when you need the detail.
Operate & deploy
| Page | The question it answers |
|---|---|
| Operating in CI | How do I run the Jenkins pipeline unattended, and which stage gates what? |
| Production troubleshooting | A run failed in production: what do I run first, and what does this symptom mean? |
| Airflow integration | How do I turn a contract into an Airflow DAG instead of hand-writing one? |
| Blueprints | Is there a template for this shape of data product, so I do not start from an empty file? |
| Source-aligned acquisition | How do I declare an ingestion from a source system rather than writing the pipeline that does it? |
| DuckDB sandbox | What may SQL inside a contract read and write on the local engine, and how do I grant more? |
Govern & secure
| Page | The question it answers |
|---|---|
| Governance and compliance | How do I declare access policy, classification and data residency in the contract, and what does the engine do when a binding contradicts them? |
| Network safety | Which outbound fetches are allowed by default, and how do I allow a host I trust? |
| Credential resolver | Where do catalog credentials live when an agent calls a tool, and what happens when none resolve? |
The sidebar puts one more page in this group, Governance, compliance and the business case, which lives outside this folder.
AI & agents
| Page | The question it answers |
|---|---|
| MCP server | How do I let an editor's AI author contracts, and how do I serve one expose of a published contract to an agent read-only? |
| Built-in and custom Forge guidance | What does fluid forge --domain already know, and where does deeper agent customisation fit? |
| Forge discovery | What does fluid forge look at in my workspace before it scaffolds, and how do I point it somewhere else or switch it off? |
| Forge memory | Where does Forge keep what it remembers between runs, and how do I inspect, search or clear it? |
| Authoring Forge tools | How do I add a new tool the agent loop can call? |
Guided fluid forge UX | What happens between typing fluid forge and a file being written? |
| LLM providers | Which providers can a forge data-model run use, and how do I select one? |
| LiteLLM backend | What does --llm-provider <name> resolve to, and what is the routing layer doing underneath? |
| Capability warnings | Why is the CLI warning me that my model does not support tool use, and what should I do about it? |
| Cost tracking | What did that run cost, and how do I override the price table for my org? |
| ChatGPT contract-GPT packet | I am building a custom GPT that drafts FLUID contracts: which files do I upload to it? |
| Agentic primitives | What are the reusable parts of the staged pipeline, and which one do I hook to watch a run? |
The sidebar also lists Forge data model here, which lives outside this folder.
Runtime reference
| Page | The question it answers |
|---|---|
| Environment variables | Which FLUID_* variable controls this, and which values does it accept? |
| Error codes | What are the parts of a catalogued failure (event name, ERR_ code, suggestions, docs link), and which events have curated suggestions? |
| Typed CLI errors | What shape does a CLI error take, and can my CI parse it without a regex? |
| Forge agent errors | A forge run raised RateLimitError or ContextOverflowError: is it retryable, and what do I change? |
| API stability | Which imports are safe to build a plugin on, and what notice do I get before one changes? |
| Contract loading API | How does my Python code load a contract exactly as fluid plan sees it, with fluid_build.api.load_contract? |
The contract schema itself is in the contract reference.
Architecture
| Page | The question it answers |
|---|---|
| Catalog integration: architecture | Why is the catalog integration shaped the way it is, and where does a new adapter plug in? |
| Catalog integration: release notes | What did the catalog-integration and cost-tracking work add, and in which release did it first ship? |