Reading case file
Editing T-502 specs
Saved reviewed archive
1. Agents
Delegate simulator setup, controlled edits, runs, diagnostics, and review evidence while you focus on the engineering decision.
Read T5_Part1.bkp
Read simulator path registry
Thought 6s
I will create a reviewed simulator workflow that preserves chemistry, builds the process objects step by step, and keeps every native write visible.
Reflux Agent
~/simulation/t5 review gate active
Native artifacts and plain-language evidence travel together.
T5_visible_train.apwz saved
One agent across desktop simulators, terminal workflows, team tools, and review surfaces.
Reading case file
Editing T-502 specs
Saved reviewed archive
reflux agent "compare duty" Thinking 5s Opened T5_Part1.bkp Wrote review note Done.
Start reviewed runs from Slack, Teams, MCP clients, or internal workflow tools.
Explore adapters ->Review cloud or desktop work from a browser before the simulator case is accepted.
Open review path ->T5 visible train - ready
Pressure sweep - review needed
Template pump edit
Reflux learns the case, templates, result paths, and approval rules before it changes anything.
Parallel subagents inspect case files, assumptions, paths, and result tables.
Learn more ->Known simulator paths, stream names, and adapter capabilities stay searchable.
Learn about context ->Read KNOWN_PATHS.md
Found pressure paths
Opened adapter registry
Teach Reflux review gates, naming conventions, simulator limits, and approved templates.
Learn more ->Require review before native writes
Preserve component IDs
Attach stream deltas
From ambiguous requests to converged runs, Reflux keeps the work legible and reviewable.
Clarify assumptions, identify missing specs, and produce an approval-ready plan.
Create or fork the case, apply bounded edits, run the simulator, and keep artifacts attached.
Opened seed_case.bkp
Created T5_visible_train.apwz
Ran convergence check
Attached review note
Use run output, diagnostics, and simulator state to explain failures and propose next moves.
Column pressure spec conflicts with condenser duty limit.
Suggested: bounded pressure sweepAgents can run commands, add context, checkpoint cases, and connect to your trusted tools.
Run builds, probes, smoke tests, and simulator scripts with sandboxed defaults.
Learn more ->$ reflux eval text_to_aspen Smoke run complete. Ready to review?
Track generated cases and evidence so teams can compare or roll back work.
Learn more ->Add simulator-specific capability without losing human review or native file traceability.
Connect Aspen Plus, HYSYS, Aveva Pro II, GWB, and controlled automation paths.
Learn more ->Start from known-good cases and preserve trusted assumptions as studies evolve.
Learn more ->Let other AI clients call Reflux tools while your simulator workflow remains governed.
Learn more ->"The agent moves fast, but the review evidence is the part that makes it usable for real process work."
"Reflux feels less like a macro and more like an engineer preparing a case for another engineer to approve."
"Keeping native files, run logs, and explanation together is the difference between automation and a workflow we can audit."