n8n-nodes-mindrouter 0.5.1 → 0.5.3
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 +43 -15
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -19,36 +19,64 @@ npm install n8n-nodes-mindrouter
|
|
|
19
19
|
## Credentials
|
|
20
20
|
|
|
21
21
|
1. Create an API key in your [MindRouter dashboard](https://mindrouter.io) under **API Keys**.
|
|
22
|
-
2. In n8n, add a **MindRouter API** credential and paste the key.
|
|
22
|
+
2. In n8n, add a **MindRouter API** credential and paste the key. The credential's test button calls `/models`, so a green check means the key works.
|
|
23
23
|
3. Leave the base URL as `https://api.mindrouter.io/v1` unless you run a self-hosted gateway.
|
|
24
24
|
|
|
25
|
+
One credential serves all four nodes.
|
|
26
|
+
|
|
25
27
|
## Nodes
|
|
26
28
|
|
|
27
|
-
### MindRouter Chat Model
|
|
29
|
+
### MindRouter Chat Model — for AI Agents
|
|
30
|
+
|
|
31
|
+
1. Add an **AI Agent** (or **Basic LLM Chain**) node.
|
|
32
|
+
2. Click its **Chat Model** socket → pick **MindRouter Chat Model**.
|
|
33
|
+
3. Choose a model — the dropdown is loaded live from your account, so it always matches your plan.
|
|
34
|
+
4. Optional: temperature, max output tokens, top-p, penalties, timeout.
|
|
35
|
+
|
|
36
|
+
Every agent call then runs through the gateway: guardrails, spend limits, and usage tracking apply, and each call appears on the canvas and in the Logs panel with token counts.
|
|
37
|
+
|
|
38
|
+
### MindRouter Embeddings — for Vector Stores
|
|
39
|
+
|
|
40
|
+
1. On any **Vector Store** node (insert or retrieve), click the **Embeddings** socket.
|
|
41
|
+
2. Pick **MindRouter Embeddings** and choose an embedding model.
|
|
42
|
+
3. Optional: batch size, dimension truncation, newline stripping.
|
|
43
|
+
|
|
44
|
+
Your whole RAG pipeline — indexing and retrieval — then runs on the same key and bill as the chat model.
|
|
28
45
|
|
|
29
|
-
|
|
46
|
+
### MindRouter PII Guard — sanitize text
|
|
30
47
|
|
|
31
|
-
|
|
48
|
+
A regular node (also usable as an AI Agent tool) over MindRouter's PII toolset. Detects Indonesian NIK, NPWP, and phone numbers, credit cards, emails, IBAN, public IPs, API keys and JWTs — and, with **Use NER** enabled, person names and locations in Indonesian and English.
|
|
32
49
|
|
|
33
|
-
|
|
50
|
+
| Operation | What it does | Returns |
|
|
51
|
+
|---|---|---|
|
|
52
|
+
| **Mask** | Irreversibly redacts all findings | `text` with `[REDACTED_*]` markers |
|
|
53
|
+
| **Scan** | Detects without changing the text | `pii_counts`, `secret_findings` |
|
|
54
|
+
| **Tokenize** | Replaces values with placeholders | `text` with `[NIK_1]`-style placeholders + `tokens` map |
|
|
55
|
+
| **Restore** | Puts original values back | `text` restored using the `tokens` map |
|
|
34
56
|
|
|
35
|
-
|
|
57
|
+
Typical pattern — send sanitized text to an external service, then restore:
|
|
36
58
|
|
|
37
|
-
|
|
59
|
+
1. **PII Guard (Tokenize)** on the incoming text: `={{ $json.chatInput }}`
|
|
60
|
+
2. Call any external service with `={{ $json.text }}` (placeholders only — no raw PII leaves).
|
|
61
|
+
3. **PII Guard (Restore)** with `Text` = the service's reply and `Tokens Map` = `={{ $('PII Guard').item.json.tokens }}`.
|
|
38
62
|
|
|
39
|
-
|
|
63
|
+
The map is returned to your workflow and never stored server-side. Calls are metered per your MindRouter plan and logged (counts only, never the text) in the dashboard's guardrail history.
|
|
40
64
|
|
|
41
|
-
|
|
65
|
+
**When to use which:**
|
|
42
66
|
|
|
43
|
-
|
|
67
|
+
- **Tokenize → Restore** — wrap any *external* AI or API call: the service works with placeholders and never sees real values. This is the default whenever text leaves MindRouter.
|
|
68
|
+
- **Mask** — before anything permanent: logs, tickets, spreadsheets, datasets. No later step needs the real values, so keep none.
|
|
69
|
+
- **Scan** — as a gate: branch with an IF node on `pii_counts` to route sensitive messages to approval instead of processing them.
|
|
70
|
+
- Requests through **MindRouter Chat Model** already pass the gateway's server-side guardrails — this node earns its place when text goes anywhere else.
|
|
44
71
|
|
|
45
|
-
|
|
72
|
+
Full guide with example workflows: [mindrouter.io/docs/integrations/n8n](https://mindrouter.io/docs/integrations/n8n).
|
|
46
73
|
|
|
47
|
-
|
|
74
|
+
### MindRouter — direct calls
|
|
48
75
|
|
|
49
|
-
|
|
76
|
+
- **Chat**: send a prompt (plus optional system prompt) or a full OpenAI-format `messages` array. Simplified output (`text`, `model`, `usage`) by default; raw API response when *Simplify Output* is off.
|
|
77
|
+
- **List Models**: one item per model your account can use, including pricing metadata.
|
|
50
78
|
|
|
51
|
-
|
|
79
|
+
Also usable as a tool by AI Agent nodes.
|
|
52
80
|
|
|
53
81
|
## Compatibility
|
|
54
82
|
|
|
@@ -56,7 +84,7 @@ Requires n8n 1.x and Node.js 18.10+.
|
|
|
56
84
|
|
|
57
85
|
## Resources
|
|
58
86
|
|
|
59
|
-
- [MindRouter documentation](https://mindrouter.io/docs)
|
|
87
|
+
- [MindRouter documentation](https://mindrouter.io/docs) — the [integrations page](https://mindrouter.io/docs/integrations) covers this node
|
|
60
88
|
- [n8n community nodes documentation](https://docs.n8n.io/integrations/community-nodes/)
|
|
61
89
|
|
|
62
90
|
## License
|
package/package.json
CHANGED