Sends issue comment text to an embeddings API so issue discussions can be stored in a vector lake. Intended to run from an issue_comment workflow and pick up the latest comment text for embedding.

How it works

Reads issue body and comments via the GitHub API.
Builds a request payload using the issue body for embed_text and the latest comment for store_text.
POSTs the payload to the configured API URL with optional token auth.

Inputs

api_url (optional): API endpoint. Default: http://agents-api.servc-agents:3000
api_token (optional): API token for authentication.
route (optional): API route. Default: agent-lake
method (optional): API method. Default: embeddings_insert
debug (optional): Enable debug logging. Default: false

Example workflow

name: Send Tasks to Agent
on:
  issue_comment:
    types:
      - created
      - edited

jobs:
  agent-embed-issue:
    if: contains(github.event.comment.body, 'solution:')
    runs-on: ubuntu-latest
    permissions:
      contents: read
      issues: read
    steps:
      - name: Send task to embedding agent
        uses: https://git.yusufali.ca/actions/embed-issues@main
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        with:
          api_token: ${{ secrets.AGENT_TOKEN }}
          api_url: ${{ secrets.AGENT_API_URL }}
          debug: false

Notes

GITHUB_TOKEN is used to read issue comments.
AGENT_API_URL and AGENT_TOKEN are expected to be configured as repository secrets.