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
+457
@@ -0,0 +1,457 @@
|
||||
Automate Email & Calendar Management with Gmail, Google Calendar & GPT-4o AI
|
||||
|
||||
https://n8nworkflows.xyz/workflows/automate-email---calendar-management-with-gmail--google-calendar---gpt-4o-ai-4366
|
||||
|
||||
|
||||
# Automate Email & Calendar Management with Gmail, Google Calendar & GPT-4o AI
|
||||
|
||||
---
|
||||
|
||||
# 1. Workflow Overview
|
||||
|
||||
This workflow, titled **"[AOE] Inbox & Calendar Management Agent"**, automates comprehensive email and calendar management by integrating Gmail, Google Calendar, and GPT-4o AI capabilities. It is designed to serve as a personal AI assistant that analyzes emails, drafts replies, manages calendar events, and provides structured insights into inbox and calendar data.
|
||||
|
||||
**Target Use Cases:**
|
||||
|
||||
- Automated processing and classification of incoming Gmail messages.
|
||||
- Generating AI-driven draft replies and email summaries.
|
||||
- Managing Google Calendar events including retrieval and creation.
|
||||
- Maintaining conversational context with memory and vector stores.
|
||||
- Enabling multi-modal triggers including chat inputs and workflow executions.
|
||||
- Supporting autonomous and user-confirmed actions to optimize email and calendar productivity.
|
||||
|
||||
**Logical Blocks:**
|
||||
|
||||
- **1.1 Input Reception & Triggers:** Nodes handling incoming chat messages, manual triggers, Gmail push triggers, and external workflow executions.
|
||||
- **1.2 Email Retrieval & Processing:** Fetching emails and threads, summarizing email content, classifying emails, and labeling them in Gmail.
|
||||
- **1.3 AI Analysis & Memory:** Utilizing OpenAI GPT-4o models for chat processing, embeddings generation, vector storage for historical context, and memory buffers.
|
||||
- **1.4 Email Drafting & Management:** Creating new draft emails or replies, deleting emails upon AI instruction.
|
||||
- **1.5 Calendar Event Handling:** Retrieving calendar events, adding new events, and converting date formats.
|
||||
- **1.6 Utility & Support Nodes:** Code nodes for email summarization, date formatting, and sticky notes for documentation and guidance.
|
||||
|
||||
---
|
||||
|
||||
# 2. Block-by-Block Analysis
|
||||
|
||||
---
|
||||
|
||||
### 2.1 Input Reception & Triggers
|
||||
|
||||
**Overview:**
|
||||
This block captures inputs from various sources, including chat messages, manual triggers, Gmail pushes, and external workflow calls, to initiate the workflow.
|
||||
|
||||
**Nodes Involved:**
|
||||
- When chat message received
|
||||
- When Executed by Another Workflow
|
||||
- Gmail Trigger
|
||||
- When clicking ‘Test workflow’
|
||||
- sessionId-master (NoOp node for passing sessionId)
|
||||
|
||||
**Node Details:**
|
||||
|
||||
- **When chat message received**
|
||||
- Type: LangChain Chat Trigger
|
||||
- Role: Listens for incoming chat messages to start the AI-driven email and calendar management process.
|
||||
- Configuration: Uses a webhook to receive chat messages; no extra options set.
|
||||
- Inputs: External chat message via webhook.
|
||||
- Outputs: Passes to `sessionId-master`.
|
||||
- Potential Failures: Webhook connectivity, malformed chat data.
|
||||
|
||||
- **When Executed by Another Workflow**
|
||||
- Type: Execute Workflow Trigger
|
||||
- Role: Allows this workflow to be triggered programmatically by another workflow with inputs `sessionId` and `chatInput`.
|
||||
- Inputs: Workflow inputs containing session context and chat text.
|
||||
- Outputs: Passes to `sessionId-master`.
|
||||
- Edge Cases: Missing inputs or incorrect data types may cause failures.
|
||||
|
||||
- **Gmail Trigger**
|
||||
- Type: Gmail Trigger
|
||||
- Role: Monitors Gmail inbox for new emails every minute to trigger processing.
|
||||
- Configuration: Polls Gmail inbox with no specific filters (monitors all incoming emails).
|
||||
- Credentials: Gmail OAuth2 required.
|
||||
- Outputs: Passes emails to `Classify Emails` and `Gmail - get recent Threads`.
|
||||
- Failure Modes: OAuth token expiration, API rate limits.
|
||||
|
||||
- **When clicking ‘Test workflow’**
|
||||
- Type: Manual Trigger
|
||||
- Role: Allows manual initiation of the workflow for testing purposes.
|
||||
- Outputs: Passes to `Gmail - get recent Threads`.
|
||||
|
||||
- **sessionId-master (NoOp)**
|
||||
- Type: No Operation (NoOp)
|
||||
- Role: Acts as a pass-through node to hold and forward the session ID for contextual continuity.
|
||||
- Outputs: Connects to the `EMail Agent`.
|
||||
|
||||
---
|
||||
|
||||
### 2.2 Email Retrieval & Processing
|
||||
|
||||
**Overview:**
|
||||
Responsible for retrieving emails and threads from Gmail, summarizing their content, classifying them into categories, and applying Gmail labels accordingly.
|
||||
|
||||
**Nodes Involved:**
|
||||
- Get last emails
|
||||
- Gmail - get recent Threads
|
||||
- Gmail1 (Get specific thread)
|
||||
- Code - Summarize Email Thread as Text
|
||||
- Classify Emails
|
||||
- Gmail - Label as Colleges
|
||||
- Gmail label as kunde
|
||||
|
||||
**Node Details:**
|
||||
|
||||
- **Get last emails**
|
||||
- Type: Gmail Tool
|
||||
- Role: Fetches the latest emails from the inbox.
|
||||
- Configuration: Uses dynamic limit from AI input (`limit`), filters inbox only.
|
||||
- Credentials: Gmail OAuth2.
|
||||
- Outputs: Passes emails to `EMail Agent`.
|
||||
- Failures: API quota, invalid query syntax.
|
||||
|
||||
- **Gmail - get recent Threads**
|
||||
- Type: Gmail (threads resource)
|
||||
- Role: Retrieves recent Gmail threads for deeper email context.
|
||||
- Credentials: Gmail OAuth2.
|
||||
- Outputs: Passes thread IDs to `Gmail1`.
|
||||
|
||||
- **Gmail1**
|
||||
- Type: Gmail (thread get operation)
|
||||
- Role: Retrieves detailed data of a specific email thread by thread ID.
|
||||
- Inputs: Thread ID from `Gmail - get recent Threads`.
|
||||
- Outputs: Passes to `Code - Summarize Email Thread as Text`.
|
||||
|
||||
- **Code - Summarize Email Thread as Text**
|
||||
- Type: Code (JavaScript)
|
||||
- Role: Processes each email message in a thread to create a formatted summary text with date, sender, recipient, subject, and snippet.
|
||||
- Key Code Logic: Converts internal timestamp to ISO date, formats message details, concatenates with separators.
|
||||
- Inputs: Messages array from Gmail1 node.
|
||||
- Outputs: Enriches JSON with `emailSummary`, passes to `Default Data Loader`.
|
||||
- Edge Cases: Missing fields in emails, invalid timestamps.
|
||||
|
||||
- **Classify Emails**
|
||||
- Type: LangChain Text Classifier
|
||||
- Role: Classifies emails into categories like "Kollegen" (colleagues) or "Kunden" (customers) based on subject, sender, and snippet.
|
||||
- Configuration: Uses fallback category "other" if no match.
|
||||
- Categories configured with descriptions and criteria.
|
||||
- Outputs: Based on classification, triggers either `Gmail - Label as Colleges` or `Gmail label as kunde`.
|
||||
- Edge Cases: Ambiguous emails, misclassification.
|
||||
|
||||
- **Gmail - Label as Colleges**
|
||||
- Type: Gmail (add label operation)
|
||||
- Role: Adds a Gmail label for emails identified as from colleagues.
|
||||
- Configuration: Uses label ID `Label_749967004333244217`.
|
||||
- Inputs: Message ID from classification.
|
||||
- Credentials: Gmail OAuth2.
|
||||
|
||||
- **Gmail label as kunde**
|
||||
- Type: Gmail (add label operation)
|
||||
- Role: Adds a Gmail label for emails identified as customers.
|
||||
- Configuration: Uses label ID `Label_4725571417728382593`.
|
||||
- Inputs: Message ID from classification.
|
||||
- Credentials: Gmail OAuth2.
|
||||
|
||||
---
|
||||
|
||||
### 2.3 AI Analysis & Memory
|
||||
|
||||
**Overview:**
|
||||
This block powers the AI-driven analysis of emails and calendar data, maintaining conversational context and enabling research into past conversations via vector stores.
|
||||
|
||||
**Nodes Involved:**
|
||||
- OpenAI Chat Model
|
||||
- OpenAI Chat Model1
|
||||
- OpenAI Chat Model2
|
||||
- Embeddings OpenAI
|
||||
- Embeddings OpenAI1
|
||||
- Threads History Vector Store
|
||||
- Write - Threads History Vector Store
|
||||
- Read- Threads History Vector Store
|
||||
- Research context and infos in previous conversations
|
||||
- Window Buffer Memory
|
||||
- EMail Agent
|
||||
|
||||
**Node Details:**
|
||||
|
||||
- **OpenAI Chat Model**
|
||||
- Type: LangChain OpenAI Chat Model
|
||||
- Role: Processes chat inputs using GPT-4o model for natural language understanding and generation.
|
||||
- Credentials: OpenAI API key.
|
||||
- Inputs: From `sessionId-master`.
|
||||
- Outputs: Feeds into `EMail Agent`.
|
||||
- Edge Cases: API rate limits, model unavailability.
|
||||
|
||||
- **OpenAI Chat Model1**
|
||||
- Similar to above but uses "gpt-4o-mini" model variant for context research.
|
||||
- Inputs: Vector store queries.
|
||||
- Outputs: `Research context and infos in previous conversations`.
|
||||
|
||||
- **OpenAI Chat Model2**
|
||||
- Uses "gpt-4.1-mini" model for email classification assistance.
|
||||
- Inputs: Email subject and snippet text.
|
||||
|
||||
- **Embeddings OpenAI**
|
||||
- Generates vector embeddings for email content to support semantic search.
|
||||
- Inputs: Summarized email data from `Default Data Loader`.
|
||||
- Outputs: Feeds both the read and write vector stores.
|
||||
|
||||
- **Embeddings OpenAI1**
|
||||
- Generates embeddings for thread history data.
|
||||
|
||||
- **Threads History Vector Store**
|
||||
- An in-memory vector store holding embeddings for past email conversations.
|
||||
- Inputs: Populated by `Embeddings OpenAI1`.
|
||||
- Outputs: Queried by `Research context and infos in previous conversations`.
|
||||
|
||||
- **Write - Threads History Vector Store**
|
||||
- Inserts new embeddings into in-memory vector store.
|
||||
- Configuration: Clears store on each run to refresh data.
|
||||
|
||||
- **Read- Threads History Vector Store**
|
||||
- Performs semantic search on stored embeddings with a top-k limit of 100 results.
|
||||
|
||||
- **Research context and infos in previous conversations**
|
||||
- LangChain Vector Store Tool using stored embeddings to answer questions about past emails, improving context-awareness.
|
||||
|
||||
- **Window Buffer Memory**
|
||||
- Maintains a sliding window of recent conversational context with a window length of 10 entries, keyed by session ID.
|
||||
- Inputs: Session ID dynamically resolved.
|
||||
- Outputs: Feeds context into `EMail Agent`.
|
||||
|
||||
- **EMail Agent**
|
||||
- LangChain Agent node acting as the core AI assistant managing emails and calendar.
|
||||
- Configuration: Receives chat input, system prompt defines role and constraints (e.g., friendly draft replies, calendar awareness, confirmation before sending).
|
||||
- Inputs: Chat messages, memory, email and calendar tools.
|
||||
- Outputs: Drives subsequent email and calendar actions.
|
||||
|
||||
---
|
||||
|
||||
### 2.4 Email Drafting & Management
|
||||
|
||||
**Overview:**
|
||||
Handles creation of new email drafts, replies to threads, and deletion of emails as directed by the AI assistant.
|
||||
|
||||
**Nodes Involved:**
|
||||
- Create an Email Draft as response to a thread
|
||||
- Create an New Email Draft
|
||||
- Delete an email
|
||||
- Get an email by MessageID
|
||||
|
||||
**Node Details:**
|
||||
|
||||
- **Create an Email Draft as response to a thread**
|
||||
- Type: Gmail Tool (draft creation)
|
||||
- Role: Creates a draft reply within an existing thread.
|
||||
- Parameters: Receives message text, recipient email, subject, and thread ID dynamically from AI inputs.
|
||||
- Credentials: Gmail OAuth2.
|
||||
- Potential Failures: Invalid thread ID, missing recipient.
|
||||
|
||||
- **Create an New Email Draft**
|
||||
- Type: Gmail Tool (draft creation)
|
||||
- Role: Creates a new outgoing draft email not linked to any thread.
|
||||
- Parameters: Message body, recipient email, and subject from AI inputs.
|
||||
- Credentials: Gmail OAuth2.
|
||||
|
||||
- **Delete an email**
|
||||
- Type: Gmail Tool (delete operation)
|
||||
- Role: Deletes an email by message ID as instructed by the AI agent.
|
||||
- Parameters: Message ID from AI input.
|
||||
- Credentials: Gmail OAuth2.
|
||||
- Edge Cases: Email not found, insufficient permissions.
|
||||
|
||||
- **Get an email by MessageID**
|
||||
- Type: Gmail Tool (get operation)
|
||||
- Role: Fetches a specific email by its message ID, often for context or confirmation.
|
||||
- Parameters: Message ID dynamically provided.
|
||||
- Credentials: Gmail OAuth2.
|
||||
|
||||
---
|
||||
|
||||
### 2.5 Calendar Event Handling
|
||||
|
||||
**Overview:**
|
||||
Manages retrieval of calendar events and the creation of new calendar entries based on AI suggestions.
|
||||
|
||||
**Nodes Involved:**
|
||||
- Get calendar events
|
||||
- Add an calender entry
|
||||
- Determine the name of the day of the week
|
||||
|
||||
**Node Details:**
|
||||
|
||||
- **Get calendar events**
|
||||
- Type: Google Calendar Tool (get all events)
|
||||
- Role: Retrieves calendar events within a specified time range.
|
||||
- Parameters: Start and end datetime from AI inputs, event limit, target calendar email.
|
||||
- Credentials: Google Calendar OAuth2.
|
||||
- Edge Cases: Date format errors, calendar access permission denied.
|
||||
|
||||
- **Add an calender entry**
|
||||
- Type: Google Calendar Tool (create event)
|
||||
- Role: Creates new calendar events or meetings with summary and description.
|
||||
- Parameters: Start and end times, summary, description provided by AI.
|
||||
- Credentials: Google Calendar OAuth2.
|
||||
|
||||
- **Determine the name of the day of the week**
|
||||
- Type: DateTime Tool
|
||||
- Role: Formats a given date into a human-readable weekday name plus date (e.g., "Tuesday 14 06").
|
||||
- Input: Date string from AI.
|
||||
- Output: Formatted date string for display or further processing.
|
||||
|
||||
---
|
||||
|
||||
### 2.6 Utility & Support Nodes
|
||||
|
||||
**Overview:**
|
||||
These nodes provide utility functions, documentation, and placeholders for workflow management.
|
||||
|
||||
**Nodes Involved:**
|
||||
- Sticky Note (multiple instances)
|
||||
- Token Splitter
|
||||
- Default Data Loader
|
||||
|
||||
**Node Details:**
|
||||
|
||||
- **Sticky Notes**
|
||||
- Type: Sticky Note
|
||||
- Role: Provide documentation, section titles, and guidance within the workflow editor for maintainers and users.
|
||||
- Contents include: "Email Sorting Agent", "Email Access Tools", "Calendar Access Tools", "Email Thread Knowledge adder", "Main Inbox Assistance Agent".
|
||||
|
||||
- **Token Splitter**
|
||||
- Type: LangChain Text Splitter (Token-based)
|
||||
- Role: Splits large email summaries into 2000 token chunks for embedding or processing limits.
|
||||
|
||||
- **Default Data Loader**
|
||||
- Type: LangChain Document Loader
|
||||
- Role: Converts JSON email summary data into a format suitable for embedding generation.
|
||||
|
||||
---
|
||||
|
||||
# 3. Summary Table
|
||||
|
||||
| Node Name | Node Type | Functional Role | Input Node(s) | Output Node(s) | Sticky Note |
|
||||
|--------------------------------------|----------------------------------------------|-------------------------------------|----------------------------------|-----------------------------------------------|-------------------------------------------------------------------------------------------------------------|
|
||||
| When chat message received | LangChain Chat Trigger | Trigger: Chat message input | External webhook | sessionId-master | |
|
||||
| When Executed by Another Workflow | Execute Workflow Trigger | Trigger: External workflow call | External workflow | sessionId-master | |
|
||||
| Gmail Trigger | Gmail Trigger | Trigger: Gmail inbox push | Gmail inbox | Classify Emails, Gmail - get recent Threads | |
|
||||
| When clicking ‘Test workflow’ | Manual Trigger | Manual trigger for testing | Manual user | Gmail - get recent Threads | |
|
||||
| sessionId-master | NoOp | Pass session ID | When chat message received, When Executed by Another Workflow | EMail Agent | |
|
||||
| Get last emails | Gmail Tool | Retrieve last emails from inbox | External (from AI input) | EMail Agent | |
|
||||
| Gmail - get recent Threads | Gmail (threads resource) | Retrieve recent threads | Gmail Trigger, When clicking ‘Test workflow’ | Gmail1 | |
|
||||
| Gmail1 | Gmail (thread get) | Get detailed thread data | Gmail - get recent Threads | Code - Summarize Email Thread as Text | |
|
||||
| Code - Summarize Email Thread as Text | Code (JS) | Summarize email thread content | Gmail1 | Default Data Loader | |
|
||||
| Classify Emails | LangChain Text Classifier | Classify emails into categories | Gmail Trigger | Gmail - Label as Colleges, Gmail label as kunde | |
|
||||
| Gmail - Label as Colleges | Gmail (add label) | Label emails from colleagues | Classify Emails | None | |
|
||||
| Gmail label as kunde | Gmail (add label) | Label emails from customers | Classify Emails | None | |
|
||||
| OpenAI Chat Model | LangChain OpenAI Chat Model | AI chat processing | sessionId-master | EMail Agent | |
|
||||
| OpenAI Chat Model1 | LangChain OpenAI Chat Model | AI for research in past conversations | Embeddings OpenAI1 | Research context and infos in previous conversations | |
|
||||
| OpenAI Chat Model2 | LangChain OpenAI Chat Model | AI for email classification | Classify Emails | Classify Emails | |
|
||||
| Embeddings OpenAI | LangChain Embeddings OpenAI | Generate embeddings for emails | Default Data Loader | Read- Threads History Vector Store, Write - Threads History Vector Store | |
|
||||
| Embeddings OpenAI1 | LangChain Embeddings OpenAI | Generate embeddings for thread history | Threads History Vector Store | Threads History Vector Store | |
|
||||
| Threads History Vector Store | LangChain Vector Store In-Memory | Store past conversation embeddings | Embeddings OpenAI1 | Research context and infos in previous conversations | |
|
||||
| Write - Threads History Vector Store | LangChain Vector Store In-Memory | Insert new embeddings | Default Data Loader | None | |
|
||||
| Read- Threads History Vector Store | LangChain Vector Store In-Memory | Search embeddings | Embeddings OpenAI | OpenAI Chat Model1 | |
|
||||
| Research context and infos in previous conversations | LangChain Vector Store Tool | Query past emails for context | OpenAI Chat Model1 | EMail Agent | |
|
||||
| Window Buffer Memory | LangChain Memory Buffer Window | Maintain recent conversation context | sessionId-master | EMail Agent | |
|
||||
| EMail Agent | LangChain Agent | Core AI assistant managing emails and calendar | sessionId-master, OpenAI Chat Model, Window Buffer Memory, Get last emails, Delete an email, Create drafts, Calendar tools | Multiple email and calendar nodes | |
|
||||
| Create an Email Draft as response to a thread | Gmail Tool (draft) | Create draft reply in thread | EMail Agent | EMail Agent | |
|
||||
| Create an New Email Draft | Gmail Tool (draft) | Create new draft email | EMail Agent | EMail Agent | |
|
||||
| Delete an email | Gmail Tool (delete) | Delete email by message ID | EMail Agent | EMail Agent | |
|
||||
| Get an email by MessageID | Gmail Tool (get) | Retrieve email by message ID | EMail Agent | EMail Agent | |
|
||||
| Get calendar events | Google Calendar Tool (get all events) | Retrieve calendar events | EMail Agent | EMail Agent | |
|
||||
| Add an calender entry | Google Calendar Tool (create event) | Create new calendar event | EMail Agent | EMail Agent | |
|
||||
| Determine the name of the day of the week | DateTime Tool | Format date to weekday name | EMail Agent | EMail Agent | |
|
||||
| Token Splitter | LangChain Text Splitter (Token) | Split text into token chunks | Default Data Loader | Embeddings OpenAI | |
|
||||
| Default Data Loader | LangChain Document Loader | Load email summary for embeddings | Code - Summarize Email Thread as Text | Embeddings OpenAI, Write - Threads History Vector Store | |
|
||||
| Sticky Note (multiple nodes) | Sticky Note | Documentation and workflow notes | None | None | See notes in section 5 |
|
||||
|
||||
---
|
||||
|
||||
# 4. Reproducing the Workflow from Scratch
|
||||
|
||||
1. **Create Trigger Nodes:**
|
||||
- Add **When chat message received** node (LangChain chat trigger) with webhook enabled.
|
||||
- Add **When Executed by Another Workflow** (Execute Workflow Trigger) configured to accept `sessionId` and `chatInput` inputs.
|
||||
- Add **Gmail Trigger** node configured to poll inbox every minute without filters, using Gmail OAuth2 credentials.
|
||||
- Add **Manual Trigger** node named "When clicking ‘Test workflow’" for testing.
|
||||
|
||||
2. **Setup Session ID Pass-Through:**
|
||||
- Add a NoOp node named `sessionId-master` to forward session context. Connect outputs of all triggers to this node.
|
||||
|
||||
3. **Connect to AI Chat Model:**
|
||||
- Add **OpenAI Chat Model** node configured with model `gpt-4o` and appropriate OpenAI credentials.
|
||||
- Connect `sessionId-master` output to this node.
|
||||
|
||||
4. **Configure Email Retrieval:**
|
||||
- Add **Get last emails** node (Gmail Tool) with inbox filter and dynamic limit from AI input `limit`. Use Gmail OAuth2 credentials.
|
||||
- Add **Gmail - get recent Threads** node (Gmail threads resource) with Gmail OAuth2 credentials.
|
||||
- Add **Gmail1** node to get details of each thread; connect output of "Gmail - get recent Threads" to this.
|
||||
- Add **Code node** named "Summarize Email Thread as Text" with JS code that formats messages into summary.
|
||||
- Connect output of `Gmail1` to the Code node.
|
||||
|
||||
5. **Prepare Embeddings and Vector Store:**
|
||||
- Add **Default Data Loader** node to load the summarized email text for embedding.
|
||||
- Add **Token Splitter** to chunk text into max 2000 tokens.
|
||||
- Add **Embeddings OpenAI** node to generate embeddings from chunks.
|
||||
- Add **Write - Threads History Vector Store** node configured to clear the store before inserting embeddings.
|
||||
- Add **Threads History Vector Store** (in-memory vector store) to hold embeddings.
|
||||
- Add **Read- Threads History Vector Store** node configured to load top 100 matches.
|
||||
- Connect these nodes in sequence from summarized email text to embedding generation and vector store insertion.
|
||||
|
||||
6. **Configure Email Classification and Labeling:**
|
||||
- Add **Classify Emails** node (LangChain Text Classifier) with categories "Kollegen" and "Kunden" and fallback "other".
|
||||
- Connect `Gmail Trigger` to `Classify Emails`.
|
||||
- Add `Gmail - Label as Colleges` and `Gmail label as kunde` nodes to add corresponding Gmail labels based on classification.
|
||||
- Connect classification outputs to these label nodes.
|
||||
|
||||
7. **Setup Email Drafting and Deletion:**
|
||||
- Add **Create an Email Draft as response to a thread** node with dynamic parameters (`message`, `To_Email`, `Subject`, `thread-ID`) from AI inputs.
|
||||
- Add **Create an New Email Draft** node for new draft emails with similar parameters excluding `thread-ID`.
|
||||
- Add **Delete an email** node to delete emails by message ID from AI inputs.
|
||||
- Add **Get an email by MessageID** node to retrieve specific emails on demand.
|
||||
|
||||
8. **Setup Calendar Management:**
|
||||
- Add **Get calendar events** node with dynamic `timeMin`, `timeMax`, and `limit` from AI inputs. Use Google Calendar OAuth2 credentials.
|
||||
- Add **Add an calender entry** node to create new events with `Start`, `End`, `Summary`, and `Description` from AI inputs.
|
||||
- Add **Determine the name of the day of the week** DateTime node to format dates for display.
|
||||
|
||||
9. **Configure AI Memory:**
|
||||
- Add **Window Buffer Memory** node for conversation context with window length 10, using session ID as key.
|
||||
- Connect this memory to `EMail Agent` node.
|
||||
|
||||
10. **Add Core AI Agent:**
|
||||
- Add **EMail Agent** node (LangChain Agent) with system prompt describing assistant role, constraints, and behavior.
|
||||
- Connect inputs from chat models, email tools (get, create, delete), calendar tools, and memory.
|
||||
|
||||
11. **Add Research Vector Store Tool:**
|
||||
- Add **Research context and infos in previous conversations** (LangChain Vector Store Tool) node to answer questions by querying vector store.
|
||||
- Connect outputs from `OpenAI Chat Model1` and vector store.
|
||||
|
||||
12. **Add Sticky Notes:**
|
||||
- Add sticky notes throughout the workflow for documentation and guidance, replicating the content and positions from the original.
|
||||
|
||||
13. **Credential Setup:**
|
||||
- Configure Gmail OAuth2 credentials with appropriate scopes for reading, labeling, drafting, and deleting emails.
|
||||
- Configure Google Calendar OAuth2 credentials for reading and writing calendar events.
|
||||
- Configure OpenAI API credentials with access to GPT-4o and related models.
|
||||
|
||||
14. **Connect All Nodes According to Logical Flow:**
|
||||
- Ensure all nodes are connected as per the original, particularly with AI inputs feeding tool nodes and outputs feeding back into the agent for autonomous decision-making.
|
||||
|
||||
---
|
||||
|
||||
# 5. General Notes & Resources
|
||||
|
||||
| Note Content | Context or Link |
|
||||
|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------|
|
||||
| **Main Inbox Assistance Agent**: Before using, modify the classifier agent to fit your needs and add Gmail labels accordingly. Also, adjust prompts to your role and company. | Sticky Note near `sessionId-master` node. |
|
||||
| More professional AI agents and resources available at [AOE AI Lab](https://ai-radar.aoe.com/). | Sticky Note near the start of the workflow. |
|
||||
| AI Assistant Prompt defines clear role for the assistant: friendly, professional drafts; calendar awareness; confirmation before sending emails or creating events. | Embedded in `EMail Agent` node parameters. |
|
||||
| Gmail label IDs like `Label_4725571417728382593` and `Label_749967004333244217` must be pre-created in Gmail for correct labeling. | Labeling nodes (`Gmail label as kunde` and `Gmail - Label as Colleges`). |
|
||||
| Calendar timestamps from Google API are in UTC; the assistant converts these to user timezone (+2) for display and scheduling. | System prompt instructions. |
|
||||
|
||||
---
|
||||
|
||||
**Disclaimer:**
|
||||
The provided text is exclusively derived from an automated workflow created with n8n, a workflow automation tool. This processing strictly adheres to existing content policies and contains no illegal, offensive, or protected elements. All handled data is legal and public.
|
||||
|
||||
---
|
||||
Reference in New Issue
Block a user