n8n-nodes-memcontext

n8n custom nodes for a self-hosted memcontext API (Flask). Includes:

• Memory Add: add text memories (/init_memory → /import_conversations)
• Memory Search: search memories and get a reply (/init_memory → /chat)
• Memory Add Multimodal: add multimodal memories (video/audio/image, /init_memory → /add_multimodal_memory), supports binary upload or local file path.

Installation

Local / private install

# in package root
npm run build
npm pack  # produces n8n-nodes-memcontext-<version>.tgz

# in n8n custom nodes dir (example: C:\Users\<you>\.n8n\nodes)
npm init -y
npm install /path/to/n8n-nodes-memcontext-<version>.tgz

Restart n8n to load the nodes.

Publish to npm (optional)

# ensure version is bumped
npm login
npm publish   # if using a scope, rename to @your-scope/n8n-nodes-memcontext and add --access public

Target env: npm install n8n-nodes-memcontext, then restart n8n.

Backend requirements

• Your Flask backend running (default http://127.0.0.1:5019).
• Multimodal video converter needs:
  • env vars: LLM_API_KEY, LLM_BASE_URL, LLM_MODEL
  • ffmpeg/ffprobe in PATH
• videorag converter is optional (imagebind/hnswlib, better on Linux/GPU). On Windows/CPU, use video.

Credentials (Memory API)

• Base URL: backend URL (default http://127.0.0.1:5019)
• API Key: optional; if backend requires auth, fill it (e.g. Bearer xxx, passed through as-is)

Nodes

Memory Add

• Writes text memory.
• Params: userId, userInput, agentResponse, timestamp (optional), metaData (JSON).
• Routes: /init_memory → /import_conversations.

Memory Search

• Searches memory and returns reply.
• Params: userId, query.
• Routes: /init_memory → /chat.

Memory Add Multimodal

• Writes multimodal memory (video/audio/image).
• Modes:
  • Upload File (Binary): binary from upstream node; set Binary Property Name.
  • Use Local File Path: absolute path accessible by backend host.
• Converters:
  • video (recommended; requires ffmpeg + Ark SDK + LLM API env vars)
  • videorag (optional; extra deps/models)
• Routes: /init_memory → /add_multimodal_memory.

Example workflow

Manual Trigger
  └─ Memory Add
       └─ Memory Add Multimodal
            └─ Memory Search

• Three nodes share the same Memory API credential and the same userId.
• Base URL recommendation: http://127.0.0.1:5019.
• For Memory Add Multimodal path mode: File Path must be an absolute path reachable by the backend machine; start with converter video.

FAQ

• ECONNREFUSED 127.0.0.1:5019: backend not running or wrong port; curl http://127.0.0.1:5019/ first.
• Video file missing / ingested_rounds=0: path not reachable or ffmpeg/env vars missing; check backend logs.
• Backend 500: inspect python app.py console stack; common causes are deps/env/path issues.
• Community nodes TLS errors: unrelated to node execution; usually external network/proxy; ignore or configure network.

Development build

• npm run build to generate dist/
• npm pack to produce release .tgz
• Before publishing, ensure package.json n8n field points to nodes/credentials under dist.