Memory Pipeline E2E Test

Run the full embed → store → search pipeline:

1. Start port-forward (if not already running):

   bashCopy

   kubectl port-forward svc/qdrant -n vector 6333:6333 &
   sleep 2
   

2. Generate embedding:

   bashCopy

   curl -s http://10.10.10.10:30001/v1/embeddings \
     -H "Content-Type: application/json" \
     -d '{"model":"default","input":"Kaizen memory pipeline test at '"$(date -Iseconds)"'"}'
   

   Verify: 4096 dimensions returned.

3. Store in Qdrant: Use the embedding from step 2 to upsert a point into the episodic collection with a random UUID and timestamp payload.

4. Search: Generate a new embedding for a related query and search the episodic collection. Verify score > 0.5.

5. Report:

   Step	Status	Details
   Embedding	?	dims, latency
   Store	?	collection, point ID
   Search	?	score, matches

6. Cleanup port-forward:

   bashCopy

   kill %1 2>/dev/null