mirror of
https://github.com/khoaliber/n8nworkflows.xyz.git
synced 2026-04-28 00:29:22 +00:00
creation
This commit is contained in:
nusquama
+341
@@ -0,0 +1,341 @@
|
||||
Email Drafting & News Research Assistant with OpenAI, Gmail, Tavily & Perplexity
|
||||
|
||||
https://n8nworkflows.xyz/workflows/email-drafting---news-research-assistant-with-openai--gmail--tavily---perplexity-7679
|
||||
|
||||
|
||||
# Email Drafting & News Research Assistant with OpenAI, Gmail, Tavily & Perplexity
|
||||
|
||||
### 1. Workflow Overview
|
||||
|
||||
This workflow, titled **"Email Drafting & News Research Assistant with OpenAI, Gmail, Tavily & Perplexity"**, is designed to streamline research and email composition tasks by combining advanced AI reasoning, up-to-date news search, and email automation within a single chat-driven interface.
|
||||
|
||||
**Target Use Cases:**
|
||||
- Users initiate a chat prompt requesting news research and/or email drafting.
|
||||
- The system performs real-time news searches using Tavily and Perplexity tools.
|
||||
- It drafts emails based on AI reasoning and memory of the conversation.
|
||||
- Emails are sent via Gmail integration.
|
||||
- The workflow is suitable for professionals needing quick synthesis of news insights into email communication without switching between apps.
|
||||
|
||||
**Logical Blocks:**
|
||||
|
||||
- **1.1 Input Reception**: Captures user prompts via chat.
|
||||
- **1.2 AI Reasoning and Memory**: Processes and plans responses using OpenAI and stores conversational context.
|
||||
- **1.3 MCP Client Orchestration**: Connects the AI agent to Modular Cognitive Process (MCP) servers for news research and email drafting.
|
||||
- **1.4 MCP Servers (Email & News Research)**: Backend services exposing tools for Gmail email sending and news search (Tavily + Perplexity).
|
||||
- **1.5 External API Calls**: Nodes that interact with Gmail, Tavily, and Perplexity APIs.
|
||||
- **1.6 Informational Sticky Notes**: Documentation nodes embedded for user guidance.
|
||||
|
||||
---
|
||||
|
||||
### 2. Block-by-Block Analysis
|
||||
|
||||
#### 1.1 Input Reception
|
||||
|
||||
- **Overview**: Listens for incoming chat messages that trigger the entire workflow.
|
||||
- **Nodes Involved**:
|
||||
- *When chat message received*
|
||||
|
||||
- **Node Details**:
|
||||
|
||||
- **When chat message received**
|
||||
- Type: `@n8n/n8n-nodes-langchain.chatTrigger`
|
||||
- Role: Entry point for chat-based user input.
|
||||
- Configuration: Default trigger settings to capture all incoming chat messages.
|
||||
- Inputs: None
|
||||
- Outputs: Connected to AI Agent node
|
||||
- Edge Cases: Incoming message format issues or connectivity problems can prevent triggering.
|
||||
|
||||
---
|
||||
|
||||
#### 1.2 AI Reasoning and Memory
|
||||
|
||||
- **Overview**: This block uses OpenAI’s GPT-4 model to understand and plan responses, while maintaining short-term conversational memory to ensure context-aware multi-turn interactions.
|
||||
- **Nodes Involved**:
|
||||
- *AI Agent*
|
||||
- *OpenAI Chat Model*
|
||||
- *Simple Memory*
|
||||
|
||||
- **Node Details**:
|
||||
|
||||
- **AI Agent**
|
||||
- Type: `@n8n/n8n-nodes-langchain.agent`
|
||||
- Role: Central orchestrator interpreting chat input and deciding which tools to invoke.
|
||||
- Configuration: System message sets the assistant as a “helpful email assistant,” instructing it to use attached Email MCP Tool for email-related tasks.
|
||||
- Inputs: From chat trigger
|
||||
- Outputs: Connects to MCP Clients and OpenAI Chat Model
|
||||
- Edge Cases: Misinterpretation of system prompt or tool communication failures.
|
||||
|
||||
- **OpenAI Chat Model**
|
||||
- Type: `@n8n/n8n-nodes-langchain.lmChatOpenAi`
|
||||
- Role: Provides reasoning and planning capabilities using GPT-4.1-mini model variant.
|
||||
- Configuration: Model set to "gpt-4.1-mini" for balanced performance.
|
||||
- Inputs: From AI Agent
|
||||
- Outputs: To AI Agent for further processing
|
||||
- Edge Cases: API rate limits, authentication errors, or model timeout.
|
||||
|
||||
- **Simple Memory**
|
||||
- Type: `@n8n/n8n-nodes-langchain.memoryBufferWindow`
|
||||
- Role: Maintains a windowed buffer of recent conversation messages to keep context.
|
||||
- Configuration: Default buffer window size; stores recent exchanges.
|
||||
- Inputs: From AI Agent (memory integration)
|
||||
- Outputs: Back to AI Agent
|
||||
- Edge Cases: Memory overflow or stale context if conversation is too long.
|
||||
|
||||
---
|
||||
|
||||
#### 1.3 MCP Client Orchestration
|
||||
|
||||
- **Overview**: These nodes represent the AI Agent’s interface to external MCP servers, allowing the agent to delegate research and email composition tasks.
|
||||
- **Nodes Involved**:
|
||||
- *Email MCP Client*
|
||||
- *News MCP Client*
|
||||
|
||||
- **Node Details**:
|
||||
|
||||
- **Email MCP Client**
|
||||
- Type: `@n8n/n8n-nodes-langchain.mcpClientTool`
|
||||
- Role: Connects to the Email MCP Server to handle email-related tasks like composing and sending emails.
|
||||
- Configuration: Endpoint URL placeholder (to be replaced); transport set to HTTP streamable for real-time communication.
|
||||
- Inputs: From AI Agent
|
||||
- Outputs: To AI Agent
|
||||
- Edge Cases: Server unreachable, incorrect endpoint, or transport protocol mismatch.
|
||||
|
||||
- **News MCP Client**
|
||||
- Type: `@n8n/n8n-nodes-langchain.mcpClientTool`
|
||||
- Role: Connects to the News MCP Server to perform news searches and retrieve up-to-date information.
|
||||
- Configuration: Endpoint URL placeholder; uses HTTP streamable transport.
|
||||
- Inputs: From AI Agent
|
||||
- Outputs: To AI Agent
|
||||
- Edge Cases: Same as Email MCP Client.
|
||||
|
||||
---
|
||||
|
||||
#### 1.4 MCP Servers (Email & News Research)
|
||||
|
||||
- **Overview**: Backend MCP servers expose tools allowing the AI Agent to perform specialized tasks: email sending via Gmail and news research via Tavily and Perplexity.
|
||||
- **Nodes Involved**:
|
||||
- *Email MCP Server*
|
||||
- *News MCP Server*
|
||||
|
||||
- **Node Details**:
|
||||
|
||||
- **Email MCP Server**
|
||||
- Type: `@n8n/n8n-nodes-langchain.mcpTrigger`
|
||||
- Role: Exposes Gmail Tool nodes for sending emails based on AI instructions.
|
||||
- Configuration: Path placeholder (to be replaced with public endpoint path).
|
||||
- Inputs: From Email MCP Client tools
|
||||
- Outputs: To Gmail Tool nodes
|
||||
- Edge Cases: Endpoint misconfiguration, credential issues, or Gmail API errors.
|
||||
|
||||
- **News MCP Server**
|
||||
- Type: `@n8n/n8n-nodes-langchain.mcpTrigger`
|
||||
- Role: Exposes Tavily and Perplexity tools to perform news search and answer queries.
|
||||
- Configuration: Path placeholder to be replaced with actual server endpoint.
|
||||
- Inputs: From News MCP Client tools
|
||||
- Outputs: To Tavily and Perplexity nodes
|
||||
- Edge Cases: Connectivity problems, API quota exhaustion, or query formatting errors.
|
||||
|
||||
---
|
||||
|
||||
#### 1.5 External API Calls
|
||||
|
||||
- **Overview**: These nodes interact directly with external services to perform core functionalities like sending emails or querying news.
|
||||
- **Nodes Involved**:
|
||||
- *Send a message in Gmail*
|
||||
- *Send a message in Gmail1*
|
||||
- *Send a message in Gmail2*
|
||||
- *Search in Tavily*
|
||||
- *Search in Tavily1*
|
||||
- *Message a model in Perplexity*
|
||||
|
||||
- **Node Details**:
|
||||
|
||||
- **Send a message in Gmail / Gmail1 / Gmail2**
|
||||
- Type: `n8n-nodes-base.gmailTool`
|
||||
- Role: Send emails via Gmail API with dynamic fields for recipient, subject, and message.
|
||||
- Configuration: Placeholders for "To," "Subject," and "Message" which are filled dynamically by AI.
|
||||
- Inputs: From Email MCP Server (triggered by AI commands)
|
||||
- Outputs: None (final action node)
|
||||
- Credential Requirements: Valid Gmail OAuth2 credentials connected in n8n.
|
||||
- Edge Cases: Authentication failures, quota limits, invalid email addresses, or message formatting issues.
|
||||
|
||||
- **Search in Tavily / Tavily1**
|
||||
- Type: `@tavily/n8n-nodes-tavily.tavilyTool`
|
||||
- Role: Perform search queries on Tavily platform to retrieve latest news or information.
|
||||
- Configuration: Queries are dynamically injected from AI outputs (using `$fromAI` expressions).
|
||||
- Inputs: From News MCP Server
|
||||
- Outputs: Back to News MCP Server
|
||||
- Credential Requirements: Tavily API credentials configured.
|
||||
- Edge Cases: API key invalid, no results found, or malformed queries.
|
||||
|
||||
- **Message a model in Perplexity**
|
||||
- Type: `n8n-nodes-base.perplexityTool`
|
||||
- Role: Sends queries to Perplexity AI for retrieval-augmented generation or simplified answers.
|
||||
- Configuration: Message content and simplify flag dynamically injected via AI expressions.
|
||||
- Inputs: From News MCP Server
|
||||
- Outputs: Back to News MCP Server
|
||||
- Credential Requirements: Perplexity API credentials connected.
|
||||
- Edge Cases: Rate limiting, connection timeouts, or unexpected response format.
|
||||
|
||||
---
|
||||
|
||||
#### 1.6 Informational Sticky Notes
|
||||
|
||||
- **Overview**: Provide embedded documentation, usage instructions, and references for users editing or running the workflow.
|
||||
- **Nodes Involved**:
|
||||
- *Sticky Note*
|
||||
- *Sticky Note1*
|
||||
- *Sticky Note2*
|
||||
- *Sticky Note3*
|
||||
|
||||
- **Node Details**:
|
||||
|
||||
- **Sticky Note** (Large, detailed)
|
||||
- Purpose: Explains the overall workflow, including components, use cases, setup instructions, and best practices.
|
||||
- Location: Top-left corner for easy visibility.
|
||||
|
||||
- **Sticky Note1**
|
||||
- Content: Labels the cluster containing the AI Agent and MCP Clients.
|
||||
|
||||
- **Sticky Note2**
|
||||
- Content: Labels the Email MCP Server nodes.
|
||||
|
||||
- **Sticky Note3**
|
||||
- Content: Labels the News Research MCP Server nodes.
|
||||
|
||||
---
|
||||
|
||||
### 3. Summary Table
|
||||
|
||||
| Node Name | Node Type | Functional Role | Input Node(s) | Output Node(s) | Sticky Note |
|
||||
|---------------------------|--------------------------------------------|--------------------------------|--------------------------|---------------------------|--------------------------------------------------------------------------------------------------------------|
|
||||
| When chat message received | `@n8n/n8n-nodes-langchain.chatTrigger` | Input reception from chat | None | AI Agent | |
|
||||
| AI Agent | `@n8n/n8n-nodes-langchain.agent` | Orchestrates tools via MCP | When chat message received| OpenAI Chat Model, MCP Clients | Sticky Note1: "Agent & MCP Client" |
|
||||
| OpenAI Chat Model | `@n8n/n8n-nodes-langchain.lmChatOpenAi` | AI reasoning and planning | AI Agent | AI Agent | |
|
||||
| Simple Memory | `@n8n/n8n-nodes-langchain.memoryBufferWindow` | Maintains conversation context | AI Agent (memory) | AI Agent | |
|
||||
| Email MCP Client | `@n8n/n8n-nodes-langchain.mcpClientTool` | Connects AI to Email MCP Server| AI Agent | AI Agent | Sticky Note1: "Agent & MCP Client" |
|
||||
| News MCP Client | `@n8n/n8n-nodes-langchain.mcpClientTool` | Connects AI to News MCP Server | AI Agent | AI Agent | Sticky Note1: "Agent & MCP Client" |
|
||||
| Email MCP Server | `@n8n/n8n-nodes-langchain.mcpTrigger` | Exposes Gmail tool for email | Email MCP Client | Gmail Tool nodes | Sticky Note2: "Email MCP Server" |
|
||||
| News MCP Server | `@n8n/n8n-nodes-langchain.mcpTrigger` | Exposes Tavily and Perplexity | News MCP Client | Tavily, Perplexity nodes | Sticky Note3: "News Research MCP Server" |
|
||||
| Send a message in Gmail | `n8n-nodes-base.gmailTool` | Sends email via Gmail | Email MCP Server | None | Sticky Note2: "Email MCP Server" |
|
||||
| Send a message in Gmail1 | `n8n-nodes-base.gmailTool` | Sends email via Gmail | Email MCP Server | None | Sticky Note2: "Email MCP Server" |
|
||||
| Send a message in Gmail2 | `n8n-nodes-base.gmailTool` | Sends email via Gmail | Email MCP Server | None | Sticky Note2: "Email MCP Server" |
|
||||
| Search in Tavily | `@tavily/n8n-nodes-tavily.tavilyTool` | News search query | News MCP Server | News MCP Server | Sticky Note3: "News Research MCP Server" |
|
||||
| Search in Tavily1 | `@tavily/n8n-nodes-tavily.tavilyTool` | News search query | News MCP Server | News MCP Server | Sticky Note3: "News Research MCP Server" |
|
||||
| Message a model in Perplexity | `n8n-nodes-base.perplexityTool` | AI retrieval-augmented answers | News MCP Server | News MCP Server | Sticky Note3: "News Research MCP Server" |
|
||||
| Sticky Note | `n8n-nodes-base.stickyNote` | Documentation and instructions| None | None | Extensive explanation of the workflow, tools, setup, and usage. See detailed description in 1.6. |
|
||||
| Sticky Note1 | `n8n-nodes-base.stickyNote` | Section label | None | None | "Agent & MCP Client" |
|
||||
| Sticky Note2 | `n8n-nodes-base.stickyNote` | Section label | None | None | "Email MCP Server" |
|
||||
| Sticky Note3 | `n8n-nodes-base.stickyNote` | Section label | None | None | "News Research MCP Server" |
|
||||
|
||||
---
|
||||
|
||||
### 4. Reproducing the Workflow from Scratch
|
||||
|
||||
1. **Create the Chat Trigger Node**
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.chatTrigger`
|
||||
- Purpose: To receive user chat messages as input.
|
||||
- Configuration: Default options.
|
||||
|
||||
2. **Create the AI Agent Node**
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.agent`
|
||||
- Configuration:
|
||||
- System Message: `"You are a helpful email assistant.\n\n##Tool\nUse attached Email MCP Tool for emails when asked\n\nUse attached Email MCP Tool for "`
|
||||
- Connect Input from: *When chat message received* node.
|
||||
- Add AI tools and memory as outputs to be connected later.
|
||||
|
||||
3. **Create OpenAI Chat Model Node**
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.lmChatOpenAi`
|
||||
- Configuration:
|
||||
- Model: Select `"gpt-4.1-mini"`
|
||||
- Options: Default
|
||||
- Connect Input from: *AI Agent* (as AI language model).
|
||||
- Connect Output to: *AI Agent*.
|
||||
|
||||
4. **Create Simple Memory Node**
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.memoryBufferWindow`
|
||||
- Configuration: Default buffer size for recent messages.
|
||||
- Connect Input and Output to *AI Agent* memory ports.
|
||||
|
||||
5. **Create MCP Clients for Email and News**
|
||||
- **Email MCP Client**:
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.mcpClientTool`
|
||||
- Set `endpointUrl` to your public Email MCP Server endpoint (replace placeholder).
|
||||
- Set `serverTransport` to `"httpStreamable"`.
|
||||
- Connect input/output to *AI Agent* (as AI tool).
|
||||
|
||||
- **News MCP Client**:
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.mcpClientTool`
|
||||
- Set `endpointUrl` to your public News MCP Server endpoint.
|
||||
- Set `serverTransport` to `"httpStreamable"`.
|
||||
- Connect input/output to *AI Agent* (as AI tool).
|
||||
|
||||
6. **Create MCP Servers**
|
||||
- **Email MCP Server**:
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.mcpTrigger`
|
||||
- Configure `path` with a public, unique path (replace placeholder).
|
||||
- Connect input from *Email MCP Client* tools (Gmail nodes).
|
||||
|
||||
- **News MCP Server**:
|
||||
- Node Type: `@n8n/n8n-nodes-langchain.mcpTrigger`
|
||||
- Configure `path` with a public, unique path (replace placeholder).
|
||||
- Connect input from *News MCP Client* tools (Tavily and Perplexity nodes).
|
||||
|
||||
7. **Add Gmail Send Message Nodes (3 variants)**
|
||||
- Node Type: `n8n-nodes-base.gmailTool`
|
||||
- Configure:
|
||||
- `sendTo`, `subject`, and `message` as dynamic fields from AI input.
|
||||
- Connect input from *Email MCP Server* (ai_tool port).
|
||||
- Connect these nodes as outputs of *Email MCP Server*.
|
||||
|
||||
8. **Add Tavily Search Nodes (2 variants)**
|
||||
- Node Type: `@tavily/n8n-nodes-tavily.tavilyTool`
|
||||
- Configure `query` parameter to use AI expressions:
|
||||
- `={{ $fromAI('Query', '', 'string') }}`
|
||||
- Connect input from *News MCP Server* (ai_tool port).
|
||||
- Connect outputs back to *News MCP Server*.
|
||||
|
||||
9. **Add Perplexity Message Node**
|
||||
- Node Type: `n8n-nodes-base.perplexityTool`
|
||||
- Configure:
|
||||
- `messages.message[0].content` set to `={{ $fromAI('message0_Text', '', 'string') }}`
|
||||
- `simplify` set to `={{ $fromAI('Simplify_Output', '', 'boolean') }}`
|
||||
- Connect input from *News MCP Server* (ai_tool port).
|
||||
- Connect output back to *News MCP Server*.
|
||||
|
||||
10. **Add Sticky Notes for Documentation**
|
||||
- Create 4 Sticky Note nodes with the provided content for user guidance and section labeling.
|
||||
- Place them appropriately for clarity.
|
||||
|
||||
11. **Credential Setup**
|
||||
- OpenAI: Connect your OpenAI API key to the *OpenAI Chat Model* node.
|
||||
- Gmail: Connect Gmail OAuth2 credentials to all Gmail Tool nodes.
|
||||
- Tavily: Connect Tavily API credentials to Tavily Tool nodes.
|
||||
- Perplexity: Connect Perplexity API credentials to the Perplexity Tool node.
|
||||
|
||||
12. **Final Connections**
|
||||
- Ensure all connections mirror the logical flow: chat trigger → AI Agent → OpenAI Chat Model & memory → MCP Clients → MCP Servers → tool nodes for email and news → back to MCP Clients → AI Agent → output.
|
||||
- Validate all placeholder fields (emails, endpoints, paths) have correct production values.
|
||||
|
||||
13. **Testing**
|
||||
- Initiate chat messages such as:
|
||||
- “Find today’s top stories on Kubernetes security and draft an intro email to Acme.”
|
||||
- “Summarize the latest AI infra trends and email a 3-bullet update to my team.”
|
||||
- Observe AI responses and email dispatch.
|
||||
|
||||
---
|
||||
|
||||
### 5. General Notes & Resources
|
||||
|
||||
| Note Content | Context or Link |
|
||||
|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------|
|
||||
| Build a chat-first MCP-powered research and outreach assistant combining Tavily, Perplexity, Gmail, and OpenAI with short-term memory for coherent interactions. | Sticky Note content in workflow |
|
||||
| Watch build-along videos for workflows like these on: www.youtube.com/@automatewithmarc | https://www.youtube.com/@automatewithmarc |
|
||||
| Requirements: n8n Cloud or self-hosted instance; OpenAI API key; Tavily, Perplexity, Gmail credentials; publicly reachable MCP endpoints. | Sticky Note content in workflow |
|
||||
| Pro Tip: Use clear action verbs in prompts like “Research X, then email Y with Z takeaways.” | Sticky Note content in workflow |
|
||||
| For safer runs, configure Gmail nodes to send to a test inbox or disable actual send action and only draft emails. | Sticky Note content in workflow |
|
||||
| Add guardrails in the AI Agent’s system message to align communication tone and style with your brand voice. | Sticky Note content in workflow |
|
||||
|
||||
---
|
||||
|
||||
**Disclaimer:** The provided text is exclusively derived from an automated workflow built with n8n, a tool for integration and automation. This processing strictly respects applicable content policies and contains no illegal, offensive, or protected elements. All manipulated data is legal and public.
|
||||
Reference in New Issue
Block a user