Bug: intelligence.cjs graph-state.json grows to 194MB due to duplicate store entries

[sparkling]

auto-memory-store.json accumulates duplicate entries on every session (4482 entries with only 157 unique IDs). intelligence.cjs buildEdges() processes all duplicates, generating O(n^2) edges between copies of the same entry. This produces 1.3M edges (194MB graph-state.json) for just 157 unique nodes.

The cache-hit check (graphState.nodeCount === store.length) also never matches because it compares deduplicated node count (157) against raw store length (4482), so the graph is needlessly rebuilt on every session start.

Fix

Deduplicate store entries by ID before building nodes and edges. Fix cache-hit check to compare against unique ID count.

Result: 194MB -> 79KB, 1,337,498 edges -> 157.

File

v3/@claude-flow/cli/.claude/helpers/intelligence.cjs

[sparkling]

Fix: #1519

[ruvnet]

Fixed in v3.5.54

Root Cause: auto-memory-store.json accumulated duplicate entries across sessions because entries were appended without deduplication, causing graph-state.json to grow to 194MB and triggering 20s+ hangs during PageRank computation.

Fix (PR #1536):

• Added deduplicateById() helper using Map to keep last entry per unique ID
• Applied in init(): reduces 4482→157 entries before graph building
• Applied in consolidate(): deduplicates before edge/node rebuild, with preDedupCount tracking to ensure deduped store is always written back to disk
• Fixed parseMemoryDir() ID collision by appending -${sIdx} index to memory entry IDs
• Applied to both v3 (v3/@claude-flow/cli/.claude/helpers/intelligence.cjs) and root (.claude/helpers/intelligence.cjs) copies

Result: 194MB graph-state.json → ~79KB. Init completes in <200ms instead of 20s+.

Published: @claude-flow/cli@3.5.54, claude-flow@3.5.54, ruflo@3.5.54