hive-mind memories are never recorded to memory.db

```
The latest commit on this repository is:

Commit: 25af48cd1c01cbcc4ac6d0dc4346e956db913845
Author: rUv ruv@ruv.net
Date: Thursday, November 13, 2025 at 11:32:02 AM (EST)
Message: Merge pull request #871 - 🚀 Release v2.7.34: MCP 2025-11 Compliance & Progressive Disclosure (Alpha)

This is a merge commit that brings in MCP 2025-11 specification compliance updates and progressive disclosure features in version 2.7.34.
```

> what triggers writting to hive-mind memory? I've been running hive-mind a lot and there are never any memories in it. They are all in swarm memory.

⏺ Let me investigate how hive-mind memory is initialized and when it should be used versus swarm memory.

⏺ Search(pattern: "new Memory\(|new CollectiveMemory\(|memory.*initialize|DatabaseManager")
⎿  Found 281 files (ctrl+o to expand)

⏺ Read(src/hive-mind/core/HiveMind.ts)
⎿  Read 542 lines

⏺ Read(src/cli/simple-commands/hive-mind.js)
⎿  Read 200 lines

⏺ Read(src/hive-mind/core/DatabaseManager.ts)
⎿  Read 150 lines

⏺ Search(pattern: "memory\.store|await.*store\(|this\.memory\.store", glob: "src/hive-mind/**/*.{ts,js}", output_mode: "content", head_limit: 30)
⎿  Found 7 lines (ctrl+o to expand)

⏺ Bash(ls -la .hive-mind/ 2>/dev/null || echo "No .hive-mind directory found")
⎿  .rw-r--r--@ 16k jimweller 17 Nov 06:46 memory.db

⏺ Bash(find . -name "*.db" -type f -not -path "./node_modules/*" -not -path "./dist-*/*" 2>/dev/null | head -20)
⎿  ./benchmark/.hive-mind/hive.db
./docker/docker-test/.swarm/memory.db
./docs/reasoningbank/models/domain-expert/memory.db
… +4 lines (ctrl+o to expand)

⏺ Bash(ls -la data/ 2>/dev/null || echo "No data directory")
⎿  No data directory

⏺ Bash(sqlite3 data/hive-mind.db "SELECT name FROM sqlite_master WHERE type='table';" 2>/dev/null || echo "Database doesn't exist or can't be read")
⎿  Database doesn't exist or can't be read

⏺ Perfect! I found the issue. Here's what's happening:

The Problem: Hive-Mind Memory Never Persists

Root Cause:
The hive-mind system expects to write to data/hive-mind.db, but the data/ directory doesn't exist in your project. This causes the DatabaseManager to fall back to in-memory storage (see
src/hive-mind/core/DatabaseManager.ts:75-94).

Evidence:
# Expected location (doesn't exist):
data/hive-mind.db ❌

# What exists instead:
.hive-mind/memory.db ✓ (16KB file, but not used by hive-mind core)
.swarm/memory.db ✓ (This is for swarm orchestrator, NOT hive-mind)

What's happening:
1. HiveMind TypeScript Core (src/hive-mind/core/DatabaseManager.ts:75):
this.dbPath = path.join(process.cwd(), 'data', 'hive-mind.db');
- Tries to create database at data/hive-mind.db
- Fails because data/ directory doesn't exist
- Falls back to in-memory storage (line 92-94)
- All memories are lost when process exits!
2. CLI Commands (src/cli/simple-commands/hive-mind.js:134):
const dbPath = path.join(hiveMindDir, 'hive.db'); // .hive-mind/hive.db
- Creates database at .hive-mind/hive.db
- But the core Memory class doesn't use this database!
3. Swarm Memory (src/swarm/memory.ts:1189):
persistencePath: './swarm-memory', // Defaults to ./swarm-memory or .swarm
- This is completely separate and works fine

The Fix:

Create the data/ directory so hive-mind can persist memories:

mkdir -p data

Then when you run hive-mind commands, the DatabaseManager will successfully create data/hive-mind.db and persist memories.

Verify it works:
# Create data directory
mkdir -p data

# Run a hive-mind command
npx claude-flow hive-mind init

# Check if database was created
ls -la data/hive-mind.db

# Verify tables exist
sqlite3 data/hive-mind.db "SELECT name FROM sqlite_master WHERE type='table';"

Why memories never appeared:
- The hive-mind system was using in-memory fallback storage
- All memories were stored in RAM and lost when the command finished
- The .hive-mind/memory.db file you see is either old/unused or created by a different process

Start with src/hive-mind/core/DatabaseManager.ts and src/cli/simple-commands/hive-mind.js, then compare their database paths and fallback behavior. Run the documented hive-mind init flow and verify that memories persist in the intended database rather than only in memory or swarm storage.

Written by the indexing model from the issue text.