Week 9

(July 29, 2025 – August 04, 2025)

Database Column Fixes

Earlier, the infrastructure was causing issues on running it locally at http://192.168.49.2:32035/ showing errors.

I focused on resolving these missing database columns to make it running. I worked on adding some missing columns that were critical for the Web Agent to function correctly. After making these fixes, I also rebuilt the container images with the latest changes to keep everything up to date. The missing columns are:

• create_sysconfig.sql
• create_uploadtree.sql
• populate_sysconfig.sql
• additional_tables.sql
• create_additional_tables.sql

Scheduler Agent & Dependencies

Our main focus is currently on the Scheduler Agent, which plays a key role in coordinating other agents like Ojo and Nomos. In our weekly call, we discussed ways to strengthen its integration with these dependent agents. The scheduler tracks available agents, and for now, we’ve decided to continue with the curl-based approach, where it pings the master node to check agent status.

Meeting 1:

(July 30, 2025)

Attendees:

• Avinal
• Shaheem Azmal M MD
• Gaurav
• Muhammad Salman

Scheduler and Agent Tracking:

• The scheduler’s role is to keep track of which agents are available.

• We discussed replacing the current curl-based ping to the master node with a shared resource approach (shared volume).

• Curl approach → supports dynamic reloading (scheduler detects automatically).

• Shared volume approach → requires manual or controller-based reloads.

• Final agreement: for now, continue with curl since it’s working, then later explore shared volume.

Focus on Getting Curl Running First:

Avinal asked whether we should continue making curl-based scheduling work. Gaurav suggested to first get it into a running state with curl. Once stable, we can think about alternatives.

Service Issues:

• Avinal pointed out a problem where installing certain dependencies starts services automatically.

• We discussed how this affects the setup, especially with PostgreSQL and other services.

Binary vs. Services Approach:

• I suggested compiling agents as standalone binaries instead of depending on PostgreSQL inside the container.

• This would remove service requirements from each Dockerfile but needs significant rework.

• We also noted and understand by help of Gaurav how environment variables and entrypoint scripts currently generate config files for agents.

Dockerfile Updates:

• Avinal asked me to write a cleaner, optimal and updated Dockerfile.ojo , so he can tested on his end.

• I worked on Dockerfile.ojo and sent updated one after testing that on my end and fully compatible with the current curl-based scheduler setup to Avinal for testing.

Meeting 2:

(July 24, 2025)

In the community call, I explained about fixing the database issues, mainly adding the missing columns needed for the Web agent. I also rebuilt the images with these changes. Right now, our main blocker is still the scheduler agent and dependent agents like ojo and nomos and some other. We also discussed keeping curl-based checks for the agents’ status, and maybe simplifying the setup later, like using a standalone binary instead of relying on PostgreSQL inside the containers.

What’s Next for Week 10

• Get feedback of Dockerfile.ojo which I sent to mentor for testing with curl-based setup.
• Fix the scheduler pod’s database connectivity
• Connect with mentors for planning final phase of work
• Test all pods, agents and the UI in Minikube to confirm everything’s rock-solid.