Running XCA has a number of problems for the user:
- generating the config.yaml and assemblies.yaml
- running the commands
- working out what's gone wrong - is it a data problem or a code problem?
An AI agent can replace most of the work the user has to do:
- it can make a good guess at the config.yaml and assemblies.yaml
- it knows how to run the commands
- it can work out what's gone wrong - it can look at the error and work out why it happened
- if it's a data problem it can suggest fixes
- if it's not solvable without code changes then it can exclude the offending crystals and try again
The user would still likely be involved with the process and they would need to confirm actions (I think the assembly should be a dimer - is this right, do you want me to exclude crystal XYZ and try again, ...).
The agent would generate a log file of the actions and decisions that took place (and whether they happened automatically or by user request) so that the process can be inspected and re-run differently if needed.
The key thing needed to do this would be to create a "Run XCA" skill which would inform the AI agent how to handle each aspect of the process.
Generating this is non-trivial but doing so would greatly improve the process.
Running XCA has a number of problems for the user:
An AI agent can replace most of the work the user has to do:
The user would still likely be involved with the process and they would need to confirm actions (I think the assembly should be a dimer - is this right, do you want me to exclude crystal XYZ and try again, ...).
The agent would generate a log file of the actions and decisions that took place (and whether they happened automatically or by user request) so that the process can be inspected and re-run differently if needed.
The key thing needed to do this would be to create a "Run XCA" skill which would inform the AI agent how to handle each aspect of the process.
Generating this is non-trivial but doing so would greatly improve the process.