Is your feature request related to a problem? Please describe.
Yes , the issue is related to a problem: today when we generate a video, we call different models and stitch smaller videos together later on, with some interpolation or other post processing whenever required. When the generation of one of the video fails, we retry http calls to the model (delivered behind an API) though this is not always enough: too many calls to a given platform may fail even after 3 calls.
Now we need a way to retry the overall video production process on the fly or later on.
Describe the solution you'd like
A desired solution would consist of a way to declare the expected steps to build the video, accompanied by a way to mark where we are in the building process so we can continue at the right step anytime.
Isolate state from logic is an envisioned approach were DTO's can be transcribed to JSON or XML, then the logic behind carried out with Async and parallel programming using an acyclic graph
Notifying the current production step directly in the JSON could be an option, or use a state manager whose lifetime would need to last as long as the video producer needs
Describe alternatives you've considered
the current basic alternative is to try more and with exponential backoff, though this is probably less desirable from a UX perspective as well as from a predictable and professional video processing pipeline.
Is your feature request related to a problem? Please describe. Yes , the issue is related to a problem: today when we generate a video, we call different models and stitch smaller videos together later on, with some interpolation or other post processing whenever required. When the generation of one of the video fails, we retry http calls to the model (delivered behind an API) though this is not always enough: too many calls to a given platform may fail even after 3 calls. Now we need a way to retry the overall video production process on the fly or later on.
Describe the solution you'd like A desired solution would consist of a way to declare the expected steps to build the video, accompanied by a way to mark where we are in the building process so we can continue at the right step anytime.
Isolate state from logic is an envisioned approach were DTO's can be transcribed to JSON or XML, then the logic behind carried out with Async and parallel programming using an acyclic graph Notifying the current production step directly in the JSON could be an option, or use a state manager whose lifetime would need to last as long as the video producer needs
Describe alternatives you've considered the current basic alternative is to try more and with exponential backoff, though this is probably less desirable from a UX perspective as well as from a predictable and professional video processing pipeline.
Additional context