Problems running with MLflow 3.9

[ablaom]

Tests fail for me when trying to run MLFlowClient.jl 0.7.0 with MLflow 3.9.

I looked into the failure and got to this point in my reduction of the code triggering a fail:

After successfully creating an MLFlow instance, mlf I get this terminal workflow:

julia> mlf
MLFlow(
    apiroot = "http://127.0.0.1:5000/api", 
    apiversion = 2.0
)

julia> experiment_id = createexperiment(mlf, "bunny rabbit")
"7"

julia> result = MLFlowClient.mlfpost(instance, "runs/create"; experiment_id)
JSON.Object{String, Any} with 1 entry:
  "run" => Object{String, Any}("info"=>Object{String, Any}("run_uuid"=>"0d239f494b65437c9fd…

julia> result["run"]
JSON.Object{String, Any} with 3 entries:
  "info"   => Object{String, Any}("run_uuid"=>"0d239f494b65437c9fd3ace8e9333efa", "experime…
  "data"   => Object{String, Any}("tags"=>Any[Object{String, Any}("key"=>"mlflow.runName", …
  "inputs" => Object{String, Any}()

julia> MLFlowClient.Run(ans)
ERROR: KeyError: key "outputs" not found
Stacktrace:
 [1] (::JSON.var"#7#8"{String})()
   @ JSON ~/.julia/packages/JSON/eypqd/src/object.jl:142
 [2] get
   @ ~/.julia/packages/JSON/eypqd/src/object.jl:137 [inlined]
 [3] getindex
   @ ~/.julia/packages/JSON/eypqd/src/object.jl:142 [inlined]
 [4] Run(data::JSON.Object{String, Any})
   @ MLFlowClient ~/Julia/MLFlowClient/src/types/run.jl:142
 [5] top-level scope
   @ REPL[47]:1

Perhaps, when there is no "output" MLflow excludes the result, rather than returning an empty one. I dunno, as I am not too familiar with this codebase.

[pebeto]

The last version of MLFlowClient.jl (v0.8.0) pairs with mlflow==3.11.1, which means that the entire workflow is compatible. This issue must be fixed