OAC AI Feedback Tuning Pipeline.

Download oac-feedback-phase1.zip from the GitHub release. Save it anywhere convenient — Downloads or Desktop is fine. You will not unzip it.

Sign in at cloud.oracle.com. Open the hamburger menu and go to Developer Services → Resource Manager → Stacks. Set the compartment filter to where you want the stack record to live (separate from where the stack's resources will be created — that's set in Step 4).

Click Create stack. On the Create stack page, in order:

1. Under "Choose the origin of the Terraform configuration", select My configuration.
2. Under "Terraform configuration source", select .Zip file (default is Folder).
3. Click Browse and pick oac-feedback-phase1.zip.
4. Leave Custom providers unchecked.
5. Name: oac-feedback-phase1
6. Description: OCI infrastructure for the OAC AI Assistant & AI Agent tuning pipeline
7. Pick a compartment, leave Terraform version at latest, click Next.

Fill these required fields:

On the Review page, make sure Run apply is NOT ticked. Click Create — Resource Manager creates the stack record. From the stack detail page, click Actions → Plan, name it initial-plan, click Plan. Status moves Accepted → In Progress → Succeeded in 30–60 seconds.

Plan: 6 to add, 0 to change, 0 to destroy.

Click Actions → Apply. In the dialog: Apply job name initial-apply, Apply job plan resolution Automatically approve. Click Apply. Watch status move Accepted → In Progress → Succeeded (~2 min).

Apply complete\! Resources: 6 added, 0 changed, 0 destroyed.

On the successful Apply job, switch to the Outputs tab. Find adb_location_uri. It looks like:

https://objectstorage.us-ashburn-1.oraclecloud.com/n/<namespace>/b/oac-feedback-logs/o/oac-ai/

Click the copy icon next to the value. Paste it somewhere you can retrieve it later — a text file, a password manager note, anywhere safe.

Confirm each resource exists:

Download the three files from the release page. Save to a folder on your laptop or straight into Cloud Shell:

• install_oac_feedback_pipeline.sh — one-shot installer (runs in Cloud Shell)
• oac_feedback_pipeline_admin_setup.sql — creates the OACFB DB user
• oac_feedback_pipeline_install.sql — creates the ingestion procedure, views, and scheduler

In the OCI console, click the Cloud Shell icon (top-right, looks like >_). Wait a few seconds for it to initialize. Then upload the three files:

1. In Cloud Shell, click the gear icon (top-right of the shell pane) → Upload.
2. Select all three files you downloaded in Step B-1. Wait for upload confirmation.
3. Verify with ls -l install_oac_feedback_pipeline.sh oac_feedback_pipeline_*.sql

Make the script executable and run it:

chmod +x install_oac_feedback_pipeline.sh
./install_oac_feedback_pipeline.sh

It will prompt you for:

Sign into Database Actions as OACFB with the password you set in B-3. Open SQL Worksheet and run:

-- Check 1: recent ingested rows exist
SELECT COUNT(*) AS rows_ingested
FROM OAC_FEEDBACK_WITH_LSQL
WHERE ts > SYSTIMESTAMP - INTERVAL '1' HOUR;

-- Check 2: no unknown buckets, LSQL correlated
SELECT event_bucket, COUNT(*) 
FROM OAC_FEEDBACK_WITH_LSQL
GROUP BY event_bucket;

The manual path runs the two .sql files directly in Database Actions. It's the same end-state as B-3 — same user, same procedure, same scheduler.

1. Sign into Database Actions as ADMIN (not OACFB — OACFB doesn't exist yet).
2. Open SQL Worksheet, paste the contents of oac_feedback_pipeline_admin_setup.sql, edit the v_oacfb_password line to set your chosen password, run as script. This creates the OACFB user with the right grants.
3. Sign out of ADMIN. Sign back in as OACFB with the password you just set.
4. Open oac_feedback_pipeline_install.sql. Find the line v_uri_prefix := '...'; near the top. Paste the adb_location_uri from Step A-7 between the single quotes.
5. Run the whole script as script (not as statements). It creates: the OCI credential, the SODA collection, the watermark table, the ingestion procedure, the feedback/LSQL views, and the hourly scheduler job.
6. Run the two verify queries from Step B-4 above to confirm the pipeline is live.

In OAC, go to Console → Connections → Create. Pick Oracle Autonomous Data Warehouse or Autonomous Transaction Processing (match your ADB type). Upload the Autonomous wallet, then enter: OACFB as the username, the password you set in Phase B, and the service name from the wallet (usually <dbname>_high or <dbname>_medium). Test the connection and save.

In OAC, go to Data → Create and choose Dataset. Select the connection from C-1, then pick the view OAC_FEEDBACK_WITH_LSQL. Recommended starter visualizations:

• Feedback Sentiment Trend: Timeline of positive / negative / neutral feedback over time
• Top Negative LSQL Patterns: Bar chart of most-criticized LSQL queries
• Feedback by Data Model: Which data models draw the most feedback
• Response Time vs. Satisfaction: Scatter of elapsed_time vs. feedback sentiment

From the dataset, click Create Workbook. Lay out the four starter visualizations from C-2 across two canvases: the first as a daily health check, the second to drill from a thumbs-down to the generated LSQL. Most common symptoms and where to fix them: