mnemosyne: Command for Claude Code

I will help you consolidate similar or duplicate memories to keep your knowledge base clean and organized.

Usage:

• /memory-consolidate - Find and review consolidation candidates
• /memory-consolidate --auto - Auto-apply all recommendations
• /memory-consolidate <id1> <id2> - Analyze specific pair
• /memory-consolidate --namespace <ns> - Limit to namespace

Instructions for me:

1. Parse the arguments:

   • Extract --auto flag if present (auto-apply mode)
   • Extract --namespace flag if present (otherwise auto-detect)
   • Extract two memory IDs if provided as positional args
   • If both --auto and IDs provided: error

2. Auto-detect namespace (if not specified):

   • Use Bash: git rev-parse --show-toplevel 2>/dev/null
   • Read CLAUDE.md for project name
   • Construct namespace as project:<name>
   • If no project: use null (all)

3. Mode A: Specific pair analysis (if two IDs provided):

   • Call mnemosyne.consolidate with the two IDs:

   {
     "name": "mnemosyne.consolidate",
     "arguments": {
       "memory_ids": ["<id1>", "<id2>"],
       "auto_apply": false
     }
   }
   

   • Display the recommendation:

   Analyzing consolidation of two memories:
   
   Memory A [<importance>/10]: <summary>
   Created: <date>
   <content preview (200 chars)>
   
   Memory B [<importance>/10]: <summary>
   Created: <date>
   <content preview (200 chars)>
   
   ---
   
   LLM Recommendation: <MERGE|SUPERSEDE|KEEP_BOTH>
   
   <if MERGE>:
   Reason: Memories contain very similar information
   Action: Merge into Memory <A|B> (higher importance), archive the other
   New content will combine both perspectives
   
   <if SUPERSEDE>:
   Reason: Memory <A|B> contains updated/more accurate information
   Action: Keep Memory <kept>, mark Memory <superseded> as superseded
   
   <if KEEP_BOTH>:
   Reason: Memories are distinct and should be maintained separately
   Action: No consolidation needed
   
   ---
   
   Apply this recommendation? [y/N]:
   (Or run with /memory-consolidate --auto <id1> <id2> to apply automatically)
   

4. Mode B: Find candidates (default mode):

   • Call mnemosyne.consolidate without IDs:

   {
     "name": "mnemosyne.consolidate",
     "arguments": {
       "namespace": "<namespace or null>",
       "auto_apply": <--auto flag value>
     }
   }
   

   • If candidates found, display each:

   🔍 Scanning for consolidation candidates in <namespace>...
   
   Found <N> candidate pairs:
   
   <for each pair>:
   
   <number>. <MERGE|SUPERSEDE|KEEP_BOTH> Recommended
      Memory A [<imp>/10]: "<summary>"
      Created: <date>, Tags: <tags>
   
      Memory B [<imp>/10]: "<summary>"
      Created: <date>, Tags: <tags>
   
      Similarity: <high|medium|low>
      Reason: <LLM reasoning>
   
      <if --auto>:
      ✓ Applied: <action taken>
   
      <if not --auto>:
      [View details: /memory-consolidate <idA> <idB>]
   
   ---
   

   • Summary at end:

   Summary:
   - Total pairs analyzed: <N>
   - Merge recommended: <N>
   - Supersede recommended: <N>
   - Keep both: <N>
   
   <if --auto>:
   - Actions applied: <N>
   - Memories archived: <N>
   
   <if not --auto>:
   To apply recommendations:
   - Review each pair: /memory-consolidate <id1> <id2>
   - Auto-apply all: /memory-consolidate --auto
   

5. Interactive confirmation (if not --auto):

   • After showing recommendation for a specific pair
   • Ask user: "Apply this recommendation? [y/N]: "
   • If 'y': Call consolidate again with auto_apply: true
   • If 'N' or anything else: Do nothing, exit

6. Format the detailed pair view: When showing specific pair details, include:

   • Full content (not just preview)
   • All tags and keywords
   • Related files and entities
   • Link information
   • Access statistics

7. Error handling:

   • If MCP server not available: "Error: Mnemosyne MCP server not running"
   • If API key not configured: "Error: Consolidation requires LLM. Configure API key with 'mnemosyne config set-key'"
   • If invalid memory IDs: "Error: Invalid memory ID(s). Use /memory-list to see available memories"
   • If no candidates found: "No consolidation candidates found. Your memory base is well-organized!"
   • If both --auto and IDs provided: "Error: Cannot use --auto with specific memory IDs"