From hive
Search the web and query LLMs with real-time data via Perplexity API. Use when user needs web research, current information, grounded answers with citations, or wants to search using Perplexity's Sonar models. Triggers on web search, research queries, "ask perplexity", current events, or real-time data needs.
npx claudepluginhub skywalking-dev/hiveThis skill is limited to using the following tools:
Web search + grounded LLM responses via bash/curl. Three APIs: Sonar (chat+search), Search (raw results), Agent (multi-provider).
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
Web search + grounded LLM responses via bash/curl. Three APIs: Sonar (chat+search), Search (raw results), Agent (multi-provider).
PERPLEXITY_API_KEY in .envjq installed# Ask a question (Sonar - default, includes citations)
hive/scripts/perplexity_handler.sh ask "What are the latest developments in AI?"
# Ask with specific model
hive/scripts/perplexity_handler.sh ask "Explain quantum computing" --model sonar-pro
# Search only (raw web results, no LLM summary)
hive/scripts/perplexity_handler.sh search "best python web frameworks 2026"
# Search with filters
hive/scripts/perplexity_handler.sh search "machine learning papers" --domains arxiv.org,scholar.google.com --max-results 5
# Agent API (use third-party models with web search)
hive/scripts/perplexity_handler.sh agent "Compare GPT-5 vs Claude Opus" --model openai/gpt-5.2
# Streaming response
hive/scripts/perplexity_handler.sh ask "Summarize today's tech news" --stream
/chat/completionsBest for: Q&A with web-grounded citations. OpenAI-compatible format.
# Basic
hive/scripts/perplexity_handler.sh ask "query"
# With filters
hive/scripts/perplexity_handler.sh ask "query" --domains example.com --recency month
# Structured JSON output
hive/scripts/perplexity_handler.sh ask "Top 3 AI startups with funding" --json-schema '{"type":"object","properties":{"startups":{"type":"array","items":{"type":"object","properties":{"name":{"type":"string"},"funding":{"type":"string"},"focus":{"type":"string"}},"required":["name","funding","focus"]}}},"required":["startups"]}'
# With system prompt
hive/scripts/perplexity_handler.sh ask "query" --system "You are a research assistant. Be precise and cite sources."
Models: sonar (fast), sonar-pro (thorough), sonar-reasoning (step-by-step), sonar-reasoning-pro (deep reasoning)
/searchBest for: Raw ranked web results without LLM processing.
hive/scripts/perplexity_handler.sh search "query" [options]
--max-results 10 # 1-20, default 10
--domains arxiv.org # comma-separated allowlist
--exclude-domains spam.com # comma-separated denylist (cannot mix with --domains)
--country US # ISO 3166-1 alpha-2
--language en,es # ISO 639-1 codes
/v1/responsesBest for: Multi-provider models (OpenAI, Anthropic, Google) with web search tool.
hive/scripts/perplexity_handler.sh agent "query" [options]
--model openai/gpt-5.2 # or any supported provider/model
--tools web_search # default; also: fetch_url
--max-tokens 1000
--reasoning high # effort level for reasoning models
--stream
hour day week month
{
"success": true,
"data": { ... }
}
Sonar responses include citations array with source URLs when available.
| Error | Cause | Fix |
|---|---|---|
| 401 | Invalid API key | Check PERPLEXITY_API_KEY in .env |
| 429 | Rate limited | Wait and retry |
| 400 | Invalid params | Check model name, filters |