Second Stage Reagent-Model Reagent-Object

[ipendlet]

Moving Forward:

1. Building a reagent-object dictionary from the already existing reagent objects in the data structure (particularly the perovskite-iodide chemistry)
2. Build a reagent-model dictionary which stores potentially useful representations for the reagents.
   1. Build a reagent-object dictionary that can be used to refer to a specific instance of a solution (which after each generation / use of the report functionality will be auto-populated by the code in a given 4-data directory)
3. Structure of the reagent dictionary must include: chemical names (max = 4), chemical ids, amounts (x4), units (x4), bare-bones action/units w/ flexibility, notes
   1. Limit: reagents of type solvent can be used at the end of a sequence, reagents of other types should be temporarily reserved for only a full specification (not mixing specifications).
   2. Limit: need temporary “new reagent form (gsheet? Python loader through xls?” which can convert ratios / mL / g (unit selection) to concentrations or …
4. Need a type label in the main pipeline that signals to the existing ‘write to interface code’ that the reagent preparation needs to be led by the mixed solvent …
   1. Add “instructions lead” to the existing interface which tells the chemist to follow a set of instructions. This is likely going to be needed down the road, but can be started now using the “notes” section. Possibly directly write a complex actions sequence to backend JSON?
5. Report::
   1. Generate a complete data frame representation of the all of the reagent objects and models at runtime. Dump these to a data frame.
      1. Each reagent object should have a model UID which describes the similarity in the ’nominal’ structure of the reagent. These models must be used for the specification (even for complex solvents which will be allotted a unique reagent model).
   2. Generate a complete data frame of chemdf at runtime.
   3. Generate the experimental report using the full dataset (will need to append the extend reagent descriptions and then do feature processing as a true secondary step. The new columns with the reagent extensions will need to be considered in the calculation of mols and such. The existing mmol code will need to be fully extracted until after the parsing is complete. This should break anything in report as these are all done downstream. Will just need to be on the experiment level

Q: How to resolve the specification of a reagent model or reagent object using the existing workflow? These two pieces of information are likely to be divergent. We likely need to create an entire entry for the models and experiment models… that means tracking the specification targets from the start of the run until the end. Alternatively, in the short term, it is best to just track the desired targets from experiments on capture and the actual results on report. This means that the models will be used for defining new experiments, but there will be existing code to ‘build’ the reagent-model dictionary from scratch if needed (based on the data in the perovskite dataset). This is already mostly done for the state-space generator.