superlocalmemory 3.6.13 → 3.6.14

package/skills/slm-recall/SKILL.md

@@ -1,343 +0,0 @@
----
-name: slm-recall
-description: Search SuperLocalMemory for relevant facts, decisions, and past context. Use when the user asks to recall, search, find, or retrieve stored information. Invokes 5-channel retrieval with LightGBM reranking via MCP.
-version: "3.4.23"
-license: AGPL-3.0-or-later
-compatibility: "SuperLocalMemory v3.4.22 — MCP (preferred) or CLI fallback"
-attribution:
-  creator: Varun Pratap Bhardwaj
-  role: Solution Architect & Creator
-  project: SuperLocalMemory v3
----
-
-# SuperLocalMemory: Recall
-
-Search and retrieve memories using semantic similarity, knowledge graph relationships, and full-text search.
-
-## Usage
-
-```bash
-slm recall "<query>" [--limit N] [--min-score 0.0-1.0] [--tags tag1,tag2] [--project name]
-```
-
-## Examples
-
-### Example 1: Basic Search
-```bash
-slm recall "FastAPI"
-```
-
-**Output:**
-```
-🔍 Search Results (3 found)
-
-[ID: 42] Score: 0.85
-We use FastAPI for REST APIs
-Tags: python, backend, api
-Project: myapp
-Created: 2026-02-05 14:23
-
-[ID: 38] Score: 0.72
-FastAPI is faster than Flask for high-throughput APIs
-Tags: performance, python
-Project: default
-Created: 2026-02-01 09:15
-
-[ID: 29] Score: 0.68
-Async endpoints in FastAPI improve concurrency
-Tags: async, fastapi, python
-Project: myapp
-Created: 2026-01-28 11:42
-```
-
-### Example 2: Limited Results
-```bash
-slm recall "authentication" --limit 3
-```
-
-**Returns:** Top 3 most relevant results
-
-### Example 3: Minimum Relevance Score
-```bash
-slm recall "React hooks" --min-score 0.7
-```
-
-**Only returns results with relevance score ≥ 0.7**
-
-### Example 4: Filter by Tags
-```bash
-slm recall "database" --tags postgresql,performance
-```
-
-**Only searches memories tagged with specified tags**
-
-### Example 5: Filter by Project
-```bash
-slm recall "API design" --project myapp
-```
-
-**Only searches memories in specified project**
-
-## Arguments
-
-| Argument | Type | Required | Default | Description |
-|----------|------|----------|---------|-------------|
-| `<query>` | string | Yes | - | Search query |
-| `--limit` | integer | No | 10 | Max results to return |
-| `--min-score` | float | No | 0.3 | Minimum relevance (0.0-1-tags` | string | No | None | Filter by tags (comma-separated) |
-| `--project` | string | No | None | Filter by project |
-
-## Search Methods Used
-
-SuperLocalMemory uses **3 search methods simultaneously** and merges results:
-
-### 1. Semantic Search (TF-IDF)
-- Converts query to vector
-- Finds similar content vectors
-- Best for: Conceptual matches
-
-**Example:**
-```
-Query: "authentication patterns"
-Matches: "JWT tokens", "OAuth flow", "session management"
-```
-
-### 2. Knowledge Graph Traversal
-- Finds related memories via graph edges
-- Discovers connected concepts
-- Best for: Related information
-
-**Example:**
-```
-Query: "FastAPI"
-Graph finds: "REST API" → "JWT auth" → "token refresh"
-```
-
-### 3. Full-Text Search (FTS5)
-- Exact keyword matching
-- Fast for known terms
-- Best for: Specific phrases
-
-**Example:**
-```
-Query: "PostgreSQL 15"
-Finds: Exact mentions of "PostgreSQL 15"
-```
-
-## Relevance Scores
-
-**Score range:** 0.0 - 1.0
-
-| Score | Meaning |
-|-------|---------|
-| **0.9 - 1.0** | Excellent match (almost exact) |
-| **0.7 - 0.9** | Strong match (very relevant) |
-| **0.5 - 0.7** | Good match (related) |
-| **0.3 - 0.5** | Weak match (loosely related) |
-| **< 0.3** | Poor match (filtered out by default) |
-
-**Factors affecting score:**
-- Keyword overlap
-- Semantic similarity
-- Graph distance
-- Recency (newer = slight boost)
-- Importance level
-- Your usage patterns (results improve automatically over time)
-
-## Advanced Usage
-
-### Natural Language (in AI chat)
-
-Most AI assistants automatically invoke this skill when you ask:
-- "What did we decide about..."
-- "Recall information about..."
-- "Search for..."
-- "What do we know about..."
-
-**Example in Cursor/Claude:**
-```
-You: "What did we decide about authentication?"
-AI: [Automatically invokes slm-recall skill]
-Found 3 memories about JWT tokens and OAuth...
-```
-
-### Combined with Other Skills
-
-**1. Recall then remember:**
-```bash
-# Find existing memories
-slm recall "API design"
-
-# Add new related memory
-slm remember "New API versioning strategy: use /v2/ prefix" --tags api,versioning
-```
-
-**2. Recall then build graph:**
-```bash
-# Find memories
-slm recall "performance"
-
-# Rebuild graph to discover new connections
-slm build-graph
-```
-
-### Scripting & Automation
-
-**Find and export:**
-```bash
-# Search and save to file
-slm recall "security" --min-score 0.7 > security-notes.txt
-
-# Count memories matching query
-slm recall "python" --limit 999 | grep "^\\[ID:" | wc -l
-```
-
-**Regular reminders:**
-```bash
-# Daily standup helper (cron job)
-#!/bin/bash
-echo "Yesterday's decisions:"
-slm recall "decided" --limit 5
-
-echo -e "\nCurrent blockers:"
-slm recall "blocked" --tags critical --limit 3
-```
-
-## Troubleshooting
-
-### "No memories found"
-
-**Causes:**
-1. No memories matching query
-2. Min-score too high
-3. Wrong project filter
-
-**Solutions:**
-```bash
-# Lower minimum score
-slm recall "query" --min-score 0.1
-
-# Remove filters
-slm recall "query"  # No project/tag filters
-
-# Check what memories exist
-slm list --limit 20
-```
-
-### "Search too slow"
-
-**Causes:**
-- Large database (10,000+ memories)
-- Complex query
-- Knowledge graph not optimized
-
-**Solutions:**
-```bash
-# Rebuild indexes
-slm build-graph
-
-# Use filters to narrow search
-slm recall "query" --project myapp --tags specific-tag
-
-# Increase min-score (fewer results = faster)
-slm recall "query" --min-score 0.7
-```
-
-### "Results not relevant"
-
-**Causes:**
-- Query too vague
-- Need to add more context
-
-**Solutions:**
-```bash
-# Be more specific
-❌ slm recall "it"
-✅ slm recall "authentication system"
-
-# Use multiple keywords
-✅ slm recall "FastAPI JWT authentication"
-
-# Use tags to filter
-✅ slm recall "performance" --tags database
-```
-
-## Output Formats
-
-### Standard Format (Default)
-```
-🔍 Search Results (3 found)
-
-[ID: 42] Score: 0.85
-Content preview...
-Tags: tag1, tag2
-Project: myapp
-Created: 2026-02-05
-```
-
-### Programmatic Use
-```bash
-# JSON output (for scripts)
-slm recall "query" --format json
-# {"results": [{"id": 42, "content": "...", "score": 0.85}, ...]}
-
-# CSV output
-slm recall "query" --format csv
-# id,content,score,tags,project,created_at
-# 42,"Content...",0.85,"tag1,tag2",myapp,2026-02-05
-```
-
-## Performance Benchmarks
-
-| Database Size | Search Time | Notes |
-|--------------|-------------|-------|
-| 100 memories | ~100ms | Instant |
-| 1,000 memories | ~500ms | Fast |
-| 10,000 memories | ~1.5s | Acceptable |
-| 50,000 memories | ~5s | Consider filtering |
-
-**Optimization tips:**
-- Use `--min-score` to filter early
-- Use `--tags` or `--project` to narrow search
-- Rebuild graph periodically: `slm build-graph`
-
-## Notes
-
-- **Multi-method:** Combines semantic, graph, and keyword search
-- **Ranked results:** Best matches first
-- **Cross-tool:** Same results in Cursor, ChatGPT, Claude, etc.
-- **Privacy:** All search happens locally
-- **Real-time:** Database updates reflected immediately
-
-## Related Commands
-
-- `slm remember "<content>"` - Save a new memory
-- `slm list` - List recent memories (no search)
-- `slm status` - Check memory count and graph stats
-- `slm build-graph` - Optimize search performance
-
----
-
-**Created by:** [Varun Pratap Bhardwaj](https://github.com/varun369) (Solution Architect)
-**Project:** SuperLocalMemory V2
-**License:** MIT (see [LICENSE](../../LICENSE))
-**Repository:** https://github.com/varun369/SuperLocalMemoryV2
-
-*Open source doesn't mean removing credit. Attribution must be preserved per MIT License terms.*
-
-## V3 Update (v3.4.22)
-
-**Preferred**: Use the `recall` MCP tool directly — it uses 5-channel retrieval
-(semantic, BM25, entity-graph, temporal, Hopfield) with LightGBM reranking
-and feeds the learning loop.
-
-```
-recall(query="your query", limit=10, session_id="<current_session>")
-```
-
-**CLI fallback** (also valid, no MCP required):
-```bash
-slm recall "<query>" [--limit N]
-```
-
-Always pass `session_id` when using MCP so engagement signals feed the ranker.

package/skills/slm-remember/SKILL.md

@@ -1,194 +0,0 @@
----
-name: slm-remember
-description: Save content to SuperLocalMemory with intelligent indexing and knowledge graph integration. Use when the user wants to remember information, save context, store coding decisions, or persist knowledge for future sessions. Automatically indexes, graphs, and learns patterns.
-version: "3.4.23"
-license: AGPL-3.0-or-later
-compatibility: "Requires SuperLocalMemory V2 installed at ~/.claude-memory/"
-attribution:
-  creator: Varun Pratap Bhardwaj
-  role: Solution Architect & Original Creator
-  project: SuperLocalMemory V2
----
-
-# SuperLocalMemory: Remember
-
-Save content to your local memory system with automatic indexing, knowledge graph integration, and pattern learning.
-
-## Usage
-
-```bash
-slm remember "<content>" [--tags tag1,tag2] [--project name] [--importance 1-10]
-```
-
-## Examples
-
-### Example 1: Basic Memory
-```bash
-slm remember "We use FastAPI for REST APIs"
-```
-
-**What happens:**
-- Content saved to SQLite database
-- TF-IDF vectors generated for semantic search
-- Entities extracted and added to knowledge graph
-- Pattern learning analyzes for coding preferences
-- Memory ID returned (e.g., 42)
-
-### Example 2: With Tags
-```bash
-slm remember "JWT tokens expire after 24 hours" --tags security,auth,jwt
-```
-
-**Tags help with:**
-- Organization
-- Filtering
-- Related memory discovery
-
-### Example 3: With Project
-```bash
-slm remember "Database uses PostgreSQL 15 with UUID primary keys" --project myapp --tags database,postgresql
-```
-
-**Project isolation:**
-- Separate memories per project
-- Switch profiles with `slm switch-profile`
-- No context bleeding
-
-### Example 4: Important Memory
-```bash
-slm remember "CRITICAL: Production deploy requires approval from @lead" --importance 10 --tags deployment,production
-```
-
-**Importance (1-10):**
-- 1-3: Low priority (notes, ideas)
-- 4-6: Normal (coding patterns, decisions)
-- 7-9: High priority (critical info, warnings)
-- 10: Critical (blockers, security issues)
-
-## Arguments
-
-| Argument | Type | Required | Default | Description |
-|----------|------|----------|---------|-------------|
-| `<content>` | string | Yes | - | The text to remember |
-| `--tags` | string | No | None | Comma-separated tags |
-| `--project` | string | No | "default" | Project name |
-| `--importance` | integer | No | 5 | Priority level (1-10) |
-
-## Output
-
-```
-Memory added with ID: 42
-
-✅ Memory saved successfully
-
-Next steps:
-  • Use `slm recall <query>` to search this memory
-  • Use `slm list` to see recent memories
-```
-
-## What Happens Behind the Scenes
-
-1. **Content Storage:** Saved to SQLite (`~/.claude-memory/memory.db`)
-2. **Semantic Indexing:** TF-IDF vectors generated for similarity search
-3. **Knowledge Graph:** Entities extracted and nodes/edges created
-4. **Pattern Learning:** Analyzes content for coding preferences (frameworks, style, testing)
-5. **Full-Text Index:** FTS5 index updated for fast keyword search
-6. **Timestamp:** Created timestamp recorded
-
-## Advanced Usage
-
-### Natural Language (in AI chat)
-
-Most AI assistants will automatically invoke this skill when you say:
-- "Remember that..."
-- "Save this for later..."
-- "I want to store..."
-- "Keep track of..."
-
-**Example in Cursor/Claude:**
-```
-You: "Remember that we decided to use React hooks over class components"
-AI: [Automatically invokes slm-remember skill]
-✓ Memory saved
-```
-
-### Bulk Import
-
-Save multiple memories from a file:
-```bash
-# From text file (one memory per line)
-while IFS= read -r line; do
-  slm remember "$line" --project bulk-import
-done < memories.txt
-
-# From CSV (content,tags,project)
-while IFS=',' read -r content tags project; do
-  slm remember "$content" --tags "$tags" --project "$project"
-done < memories.csv
-```
-
-### Integration with Git Hooks
-
-**Pre-commit hook** (save commit messages):
-```bash
-#!/bin/bash
-# .git/hooks/post-commit
-
-commit_msg=$(git log -1 --pretty=%B)
-commit_hash=$(git log -1 --pretty=%H)
-
-slm remember "Commit: $commit_msg (${commit_hash:0:7})" \
-  --tags git,commit \
-  --project "$(basename $(git rev-parse --show-toplevel))"
-```
-
-## Error Handling
-
-| Error | Cause | Solution |
-|-------|-------|----------|
-| "Database locked" | Another process accessing DB | Wait or `killall python3` |
-| "Content cannot be empty" | Empty string passed | Provide content |
-| "Invalid importance" | Value not 1-10 | Use number between 1-10 |
-| "Database not found" | SuperLocalMemory not installed | Run `./install.sh` |
-
-## Notes
-
-- **100% local:** Nothing leaves your machine
-- **Cross-tool sync:** All AI tools access same database (Cursor, ChatGPT, Claude, etc.)
-- **Unlimited:** No memory limits, no quotas
-- **Privacy:** Your data stays on your computer
-- **Profiles:** Use `slm switch-profile` for profile isolation
-
-## Related Commands
-
-- `slm recall "<query>"` - Search memories semantically
-- `slm list` - List recent memories
-- `slm status` - Check system health
-- `slm build-graph` - Rebuild knowledge graph
-- `slm switch-profile <name>` - Switch memory profile
-
-## Technical Details
-
-**Database Schema:**
-- Table: `memories`
-- Fields: id, content, tags, project_name, importance, created_at, etc.
-- Indexes: Full-text search (FTS5), TF-IDF vectors, timestamps
-
-**Performance:**
-- Add memory: ~50ms
-- With knowledge graph: ~300ms
-- Large content (10KB): ~1s
-
-**Limits:**
-- Max content size: 1MB
-- Max tags: 50 per memory
-- Max project name: 64 characters
-
----
-
-**Created by:** [Varun Pratap Bhardwaj](https://github.com/varun369) (Solution Architect)
-**Project:** SuperLocalMemory V2
-**License:** MIT (see [LICENSE](../../LICENSE))
-**Repository:** https://github.com/varun369/SuperLocalMemoryV2
-
-*Open source doesn't mean removing credit. Attribution must be preserved per MIT License terms.*