multidim_README.md

Multidim testing

Assumes have access to multidim namespace and the ABCD raw files from HF datasets

Assumes 1 convoset with n playbooks, with n pipelines. Each pipeline being for a different dimension of analysis.

implements this logic (sequential)

Upload file to conversation set.

• hf_api.get_conversation_set
• hf_api.upload_json_file_to_conversation_source

Check on each upload file is loaded and jobs all complete before starting pipelines

• loop until TRIGGER_STATUS_COMPLETED
• • hf_api.describe_trigger

Workout all playbooks to run

• hf_api.list_playbooks

For each playbook (workspace) get the pipelines for that playbook

• hf_api.list_playbook_pipelines

Run them one by one

• • hf_api.trigger_playbook_pipeline

Check for each that pipeline finishes

• • loop until TRIGGER_STATUS_COMPLETED
• • • hf_api.describe_trigger

Download the results filtering by filename as a metadata key immediately

• • hf_api.export_query_conversation_inputs

implements this logic (parallel)

Upload sinlge file to conversation set (skippable)

• hf_api.get_conversation_set
• hf_api.upload_json_file_to_conversation_source

Doesn't check that trigger finished

Workout all playbooks to run

• hf_api.list_playbooks

For each playbook (workspace) get the pipelines for that playbook

• hf_api.list_playbook_pipelines

Run them one by one - without waiting for triggers to finsih

• hf_api.trigger_playbook_pipeline

Download the results filtering by filename as a metadata key immediately for every playbook

• hf_api.export_query_conversation_inputs