| name | web-search |
| description | Queue a web search via Gemini CLI on the host. Free tier, no API cost. Results return via ops.db. |
| version | 1.0.0 |
| author | research |
| tags | ["search","web","gemini","free"] |
| intent | Informed [I18] |
web-search
Queue a web search through Gemini CLI on the host (free tier, google_web_search built in).
When to use
- Any research request requiring current web information
- Replaces direct
web_search tool calls (which burn paid Google API tokens)
- Use for: news lookups, competitor research, documentation searches, fact-checking
Process
- Create an ops.db task via the gateway
tasks_create tool:
{
"agent": "spec-research",
"task": "web-search: <brief description>",
"meta": {
"host_op": "gemini-search",
"query": "<the actual search query — be specific>",
"telegram_chat_id": "<chat_id if results should go to Telegram>"
},
"urgency": "routine"
}
- The host-ops-executor polls every 30s and runs
gemini-task with the query
- Results are stored in the task's
result field in ops.db
- If
telegram_chat_id is set, results also push to Telegram
Cost
Free — uses Gemini CLI free tier (10 RPM, 250 requests/day).
Limitations
- Async: ~30-90s latency (poll interval + search time)
- Free tier rate limits: 10 RPM, 250/day. If exhausted, task blocks with 429 error.
- Results are text summaries, not raw HTML
