Sókrates Wiki

Deployment and Verification of Agent Endpoints

2 min read

Deployment and Verification of Agent Endpoints

Summary

This process describes the deployment and end-to-end testing workflow for new agent-related API features. It involves synchronizing the feat/agent-endpoints backend branch with the feat/agent-tab frontend branch and verifying the functionality of the /api/agents endpoint.

Details

The deployment process was initiated to resolve a discrepancy between the development environment and the running backend service. During testing, it was observed that the backend service, identified as fastapi-verdmat, was running code from the master branch. This resulted in 404 errors when attempting to access the new /api/agents endpoint and a lack of the unified layers field in the prediction logic.

To facilitate end-to-end verification using Chrome DevTools, the following workflow was established:

  1. Code Synchronization: Changes were pushed to two primary repositories:

    • feat/agent-tab was pushed to the RationallyPrime/Sokrates-AI repository (frontend).
    • feat/agent-endpoints was pushed to the RationallyPrime/residual-modeling repository (backend).

  2. Deployment Strategy: Two options were identified for updating the backend service to the new feature set:

    • Merging the feat/agent-endpoints branch into the master branch.
    • Checking out the feat/agent-endpoints branch directly on the server where the service resides.

  3. Service Management: Once the code is updated on the target environment, the backend service must be refreshed. The specific command for this action is sudo systemctl restart fastapi-verdmat. This indicates that the backend is managed as a systemd service, consistent with the project’s NixOS-based infrastructure.

  4. Verification: After the service restart, Chrome DevTools is used to inspect network traffic and verify that the /api/agents endpoint returns the expected data and that the predict functionality correctly utilizes the unified layers field.

The process concluded with the clean shutdown of the frontend-dev environment while awaiting final confirmation on the deployment method for the backend.

Graph View