4.5 Upstream Analysis

4.5.1 BioLine supplies Flows dynamically

Upstream quality control, host removal, taxonomic profiling, and functional processing run through BioLine Flow versions. The execution workflow discovers candidates with fkit flow list and checks their real interfaces with fkit flow inspect. The plan records BioLine-returned Flow/Version IDs, ports, parameters, file groups, and defaults; a local name is never substituted for an ID.

4.5.2 Inputs and bindings

Inputs must be readable by BioLine. The workflow uses fkit ls to obtain the file list and builds bindings from the inspect response. If files, ports, required parameters, or authoritative documentation are missing, record the concrete gap and stop in document_gap or plan_failed instead of filling it from assumptions.

4.5.3 Submission, polling, and materialization

Create the BioLine Job only after the user confirms the plan. Store the real Job ID and every polling response. Only a successful status permits downloading and finalizing outputs into stage with an input/output record. Downstream analysis must not read upstream data that has not been successfully finalized.

4.5.4 Downstream handoff and reproducibility

Retain the upstream plan, Flow version, Job, and input bindings with the task. Then select an exact downstream Tool NAME:VERSION and retain its summary/config snapshot or hash, generated script, on-demand pulled image identity, actual command, logs, and outputs. If the Flow version, inputs, or interface cannot be confirmed, preserve the failed phase and evidence rather than silently selecting another object.