research-cache-ops

name	research-cache-ops
description	Operate the research-cache for Perplexity/Context7 query results (especially expensive perplexity_research deep-research calls). Check, inspect, and clear cached research to avoid re-running expensive queries.

Research Cache Operations

Wraps ${CLAUDE_PLUGIN_ROOT}/bin/research-cache for common cache operations. Research is expensive (perplexity_research deep-research — the primary research call — is 30s+ and costs money); cached results should be reused across sessions unless explicitly invalidated.

Operations

• Stats: bin/research-cache stats — entry count and cache size
• Key for query: bin/research-cache key "<query text>" — deterministic SHA-256 key
• Check hit: bin/research-cache has <key> — exit 0 if present
• Fetch: bin/research-cache get <key> — print cached JSON
• Store: bin/research-cache put <key> < result.json — write to cache
• Clear: bin/research-cache clear — wipe all entries

When to use

• Before spawning research-agent: check if the query was already answered
• After a research run: store the result with the computed key so future sessions get a hit
• When invalidating stale research: clear specific keys or the whole cache
• When diagnosing research cost: inspect stats

Integration

The research-agent subagent should call this via Bash before making Perplexity (perplexity_research) / Context7 calls. Pseudocode:

key=$(${CLAUDE_PLUGIN_ROOT}/bin/research-cache key "$query")
if ${CLAUDE_PLUGIN_ROOT}/bin/research-cache has "$key"; then
  ${CLAUDE_PLUGIN_ROOT}/bin/research-cache get "$key"
else
  # run the real query, then:
  echo "$result" | ${CLAUDE_PLUGIN_ROOT}/bin/research-cache put "$key"
fi