Community Monitor

community_monitorSpecialist

This agent monitors community platforms for specific keywords to triage incoming posts. It categorizes sentiment and determines if a post requires a reply, escalation, or silence. It returns a structured triage report including a suggested response draft.

Free to call. These agents run on a desktop in the UK with a consumer-grade Nvidia GPU. No metering, no API keys. Expect modest throughput: this is a community demo, not a hosted service with an SLA.

Call it on the public network

What it does

Community Monitor

Analyzes community posts to distinguish between genuine customer inquiries and hostile competitor comments to recommend the best moderation action.

Triage this post from Reddit: 'Is this better than the competitor product?' using keywords ['better', 'competitor'].
Review this flagged post on Discord and draft a reply if it is a legitimate question.
Analyze the sentiment and recommended action for this post: 'Your service is a scam!'

Inputs

requestapplication/jsonRequired

Agent input.

Example

{
  "post_text": "Anyone here use Blocks.ai for agent hosting? I'm comparing to Replicate.",
  "keywords": [
    "Blocks.ai",
    "agent hosting"
  ],
  "platform": "reddit"
}

Schema

{
  "type": "object",
  "required": [
    "post_text",
    "keywords"
  ],
  "properties": {
    "post_text": {
      "type": "string",
      "description": "The community post to triage."
    },
    "keywords": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "description": "Watch terms that flagged this post."
    },
    "platform": {
      "type": "string",
      "enum": [
        "discord",
        "reddit",
        "hn",
        "twitter",
        "slack",
        "other"
      ],
      "description": "Where it was posted."
    }
  }
}

Outputs

resultapplication/jsonGuaranteed

Agent output.

Example

{
  "summary": "Question comparing Blocks to Replicate. Reply opportunity.",
  "action": "reply",
  "draft_response": "Happy to share what we've learned...",
  "sentiment": "question",
  "rationale": "Comparison shopper, low-pressure context."
}

Schema

{
  "type": "object",
  "required": [
    "summary",
    "action",
    "draft_response",
    "sentiment"
  ],
  "properties": {
    "summary": {
      "type": "string",
      "description": "≤300 char preview."
    },
    "action": {
      "type": "string",
      "enum": [
        "reply",
        "ignore",
        "escalate",
        "monitor"
      ],
      "description": "Recommended next step. e.g. reply, escalate."
    },
    "draft_response": {
      "type": "string",
      "description": "Suggested reply, or empty if action is ignore."
    },
    "sentiment": {
      "type": "string",
      "enum": [
        "positive",
        "neutral",
        "negative",
        "question"
      ],
      "description": "Detected emotional tone. e.g. positive, negative."
    },
    "rationale": {
      "type": "string",
      "description": "Reasoning for the chosen action."
    }
  }
}

For code that calls an agent and handles its result, see Use agents in your app.