n8n-nodes-memory 0.5.0 → 0.5.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +101 -0
- package/package.json +1 -1
package/README.md
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
1
|
+
# n8n-nodes-memory
|
|
2
|
+
|
|
3
|
+
Custom memory node for n8n AI Agent with external API support.
|
|
4
|
+
|
|
5
|
+
## The Power: Memory as Workflows
|
|
6
|
+
|
|
7
|
+
Traditional memory nodes are passive storage. This node transforms memory into **event-driven workflows** - each memory operation (get, add, clear) can trigger complete n8n workflows with access to 400+ integrations.
|
|
8
|
+
|
|
9
|
+
### On GET (before AI responds)
|
|
10
|
+
- Fetch context from vector databases (RAG)
|
|
11
|
+
- Load user preferences from CRM
|
|
12
|
+
- Inject real-time data (weather, stocks, news)
|
|
13
|
+
- Summarize old conversations automatically
|
|
14
|
+
- Generate dynamic context with another AI
|
|
15
|
+
- Apply predictive loading for likely questions
|
|
16
|
+
|
|
17
|
+
### On ADD (after each message)
|
|
18
|
+
- Analyze sentiment in real-time
|
|
19
|
+
- Extract entities and intents
|
|
20
|
+
- Update user profiles and CRM records
|
|
21
|
+
- Create tasks/tickets automatically
|
|
22
|
+
- Trigger notifications and alerts
|
|
23
|
+
- Route to different AI agents based on intent
|
|
24
|
+
- Build knowledge graphs from conversations
|
|
25
|
+
- Feed training data pipelines
|
|
26
|
+
|
|
27
|
+
### On CLEAR (end of session)
|
|
28
|
+
- Generate conversation summaries
|
|
29
|
+
- Extract action items
|
|
30
|
+
- Send follow-up emails
|
|
31
|
+
- Archive to compliance systems
|
|
32
|
+
- Update customer journey stages
|
|
33
|
+
|
|
34
|
+
## Use Cases
|
|
35
|
+
|
|
36
|
+
**Self-Evolving AI** - Behavior rules that adapt based on conversation patterns. Detect frustration, switch to empathetic mode.
|
|
37
|
+
|
|
38
|
+
**Multi-Agent Orchestration** - Route messages to specialized agents. One memory feeding an agent swarm.
|
|
39
|
+
|
|
40
|
+
**Memory as API Gateway** - Natural language interface to any system. "Check my order status" triggers lookup workflows.
|
|
41
|
+
|
|
42
|
+
**Semantic Compression** - Compress long exchanges into dense summaries, expand on retrieval. Effectively infinite context window.
|
|
43
|
+
|
|
44
|
+
**Cross-Platform Identity** - Sync across WhatsApp, Telegram, Web. Same AI remembers you everywhere.
|
|
45
|
+
|
|
46
|
+
**Conversation Branching** - Create save points, fork conversations, explore different paths. Git for chat.
|
|
47
|
+
|
|
48
|
+
**Regulatory Firewall** - Check compliance before storing, redact sensitive info based on user role. GDPR/LGPD by design.
|
|
49
|
+
|
|
50
|
+
**Continuous Learning** - Route high-quality exchanges to training datasets. Self-improving system.
|
|
51
|
+
|
|
52
|
+
**Social Graph Memory** - Map relationships mentioned in conversations. Build knowledge graphs of user's world.
|
|
53
|
+
|
|
54
|
+
**Emotional State Machine** - Track emotional journey, maintain consistent AI "mood" across sessions.
|
|
55
|
+
|
|
56
|
+
## Installation
|
|
57
|
+
|
|
58
|
+
```bash
|
|
59
|
+
npm install n8n-nodes-memory
|
|
60
|
+
```
|
|
61
|
+
|
|
62
|
+
Then restart n8n.
|
|
63
|
+
|
|
64
|
+
## Setup
|
|
65
|
+
|
|
66
|
+
1. **Create a Memory API workflow** - A webhook that handles `get`, `add`, and `clear` actions
|
|
67
|
+
2. **Set up your storage** - PostgreSQL, Redis, or any backend
|
|
68
|
+
3. **Connect to AI Agent** - Link the Memory API node to your agent's memory input
|
|
69
|
+
|
|
70
|
+
## API Contract
|
|
71
|
+
|
|
72
|
+
Your webhook must handle POST requests with:
|
|
73
|
+
|
|
74
|
+
| Action | Purpose | Returns |
|
|
75
|
+
|--------|---------|---------|
|
|
76
|
+
| `get` | Retrieve messages for session | `{ messages: [{ type, content }, ...] }` |
|
|
77
|
+
| `add` | Store a message | `{ success: true }` |
|
|
78
|
+
| `clear` | Clear session history | `{ success: true }` |
|
|
79
|
+
|
|
80
|
+
Message types: `human` or `ai`
|
|
81
|
+
|
|
82
|
+
## Parameters
|
|
83
|
+
|
|
84
|
+
| Parameter | Description |
|
|
85
|
+
|-----------|-------------|
|
|
86
|
+
| API URL | Your memory webhook URL |
|
|
87
|
+
| Session ID | Unique conversation identifier |
|
|
88
|
+
| API Key | Optional Bearer token |
|
|
89
|
+
| Context Window Length | Messages to include (default: 10) |
|
|
90
|
+
|
|
91
|
+
## How It Works
|
|
92
|
+
|
|
93
|
+
Implements LangChain's `BaseListChatMessageHistory` with `BufferWindowMemory`. The AI Agent calls your workflows automatically during conversation.
|
|
94
|
+
|
|
95
|
+
## License
|
|
96
|
+
|
|
97
|
+
MIT
|
|
98
|
+
|
|
99
|
+
## Author
|
|
100
|
+
|
|
101
|
+
[Filipe Labs](https://github.com/filipexyz)
|