Planning Issue

[kordless]

Purpose of this ticket is to open more tickets:

• disable endpoint on fail to auth to featurebase Disable endpoint if FeatureBase auth fails #14

• implement a query endpoint for insert/query of data already inserted compared to the new insert Query endpoint #18

• test the user key for OpenAI works before running a pipeline on it Validate user's OpenAI tokens. #15

• rename the table components to "pipeline" Add table object to system #19

• ensure a JSON based schema that allows reordering the model pipeline within reason (i.e. the keys are there for the model to ingest and output to). For example, embeddings may or may not take the keyterms from another model to help with the weights. Pipeline control for users #20

• abstract the box_start to take different types of box deployments (GPU type) Abstract box_start to start new box types. #21

• abstract the document model to take those IPs and pass them to the AI handlers as needed (once started). Abstract box_start to start new box types. #21

• add a Llama2 method to the new box type (so we can run codellama or llama 2) Add Llama 2 box type and model runner. Laminoid#1

• implement a search box to test the submitted data (which also submits that data and runs a pipeline on it first) Query endpoint #18

• implement appengine push from appengine branch only...process we practice Implement push from appengine branch. #39

• account delete deletes the tables/pipelines Delete all user data with user delete function #35

• only run schemer when the schema changes (or eliminate it if possible)

• implement slothANN vector balancing algo, and make it obviously extensible by the user Implement basic vector balancing model. #37

• investigate adding weights to stringsets for timing out keyterms/keywords using timequantums (not sure how this is useful yet) not usefule

• test harness written Create test harnesses #36

• task list pulls from kafka too (dropped as requirement)

• write tasks to a log and expire them out after some time...put labels on what they did/worked/didn't work Dead letter handling #38

• add an idea of a delay to the task queue, for starting boxes, etc. (not needed)

[kordless]

• task list pulls from kafka too

[kordless]

• write tasks to a log and expire them out after some time...put labels on what they did/worked/didn't work

[kordless]

• add an idea of a delay to the task queue, for starting boxes, etc.

[kordless]

• account delete deletes the tables/pipelines

[kordless]

• implement appengine push from appengine branch only...process we practice

[kordless]

All tickets or issues addressed. Closing.