AI & Computer Vision

Integration with Opencode AI and vision models for automated proof verification.

AI & Verification

The backend/convex/ai/ directory manages the integration between the CommitT backend and external AI models used for proof verification and task generation.

The Opencode Client

Located in ai/client.ts.

CommitT leverages Opencode AI to analyze image uploads and perform complex semantic validation. The client is instantiated as a singleton using the @opencode-ai/sdk/client package.

import { createOpencodeClient } from "@opencode-ai/sdk/client";

export async function getOpencodeClient() {
  if (!opencodeClient) {
    opencodeClient = createOpencodeClient({
      baseUrl: OPENCODE_BASE_URL,
    });
  }
  return opencodeClient;
}

AI Capabilities

1. Visual Proof Verification

When a user submits a photo to prove they completed a task (e.g., a photo of a gym entrance), the AI layer evaluates the image against the task’s context to determine if it meets the criteria.

2. Task Generation

The AI layer can also be used to automatically generate rule sets, target lists, and blocklists based on natural language inputs from the user (e.g., “Block all distracting apps”).