Best practices

Session management

Reuse Fluent sessions

Keep the same Fluent session open for multiple operations to improve performance. Only disconnect when you are finished or need a clean solver state.

Clean shutdown

Always call disconnect when finished to free resources.

Check status

Use session_status and solver_status to verify connection and iteration state before mutating operations.

Discover-generate-validate-execute loop

For almost every non-trivial setup change:

1. Discover: Use find_api, get_state, or list_named_objects before writing code.

2. Validate: Use validate_code to catch sandbox violations before execution.

3. Execute: Use run_code to apply the change.

4. Verify: Use summarize_setup or get_state to confirm the result.

Never skip discovery for path-dependent settings (URF, Courant, UTL vs classic tree).

Code execution

Validate before running

Always call validate_code on generated or untrusted code before run_code.

Prefer targeted reads

Use get_state with specific paths rather than reading large subtrees.

Use named-object tools

Use list_named_objects and find_named_object instead of guessing collection paths.

Mesh and diagnostics

Route mesh quality intents correctly

Use mesh_quality for skewness, orthogonal quality, aspect ratio, and mesh.check output. Do not use run_code with invented PyFluent accessors.

Use compare_files for case diffs

compare_files runs in isolated headless sessions and never touches the live workspace session.

Workflow design

Modular workflows

Break complex setup into smaller, verifiable steps.

Error recovery

When run_code fails, read the error, re-discover with get_state, and regenerate rather than retrying blindly.

Offline-first discovery

Use find_api and get_help even when offline; only call live tools when you need current values or mutations.

Performance

Minimize reconnects

Avoid repeated connect / disconnect cycles within a single workflow.

Batch related changes

Combine related run_code snippets when safe. However, smaller steps are preferred for easier debugging.

Compare files sparingly

Using compare_files launches two headless Fluent sessions. Use it for explicit diff requests, not routine inspection.

Common patterns

Boundary condition setup

1. connect and load case/mesh.

2. list_named_objects("setup.boundary_conditions.velocity_inlet") lists inlets.

3. get_state(["setup.boundary_conditions.velocity_inlet.inlet-1"]) reads current values.

4. validate_code → run_code applies the desired settings update.

5. summarize_setup(scope="boundaries") verifies the boundary setup.

Convergence check

1. solver_status checks if Fluent is iterating.

2. simulation_report returns a structured solution report.

3. get_state(["solution.monitors.residual"]) reads residuals.

Case comparison

1. compare_files(path_a="baseline.cas.h5", path_b="modified.cas.h5").

2. Reply with the pre-rendered summary markdown table verbatim.