klbr/n8nworkflows.xyz
1
0
Fork 0
mirror of https://github.com/khoaliber/n8nworkflows.xyz.git synced 2026-04-28 00:29:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

creation

Browse Source
This commit is contained in:
nusquama
2025-11-12 18:57:06 +01:00
parent aff3b84d13
commit 6350df3eb9
1 changed files with 279 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
workflows/Analyze BeyondPresence Video Calls with GPT-4o-mini and Google Sheets-4436/readme-4436.md
+279
View File
@@ -0,0 +1,279 @@
Analyze BeyondPresence Video Calls with GPT-4o-mini and Google Sheets
https://n8nworkflows.xyz/workflows/analyze-beyondpresence-video-calls-with-gpt-4o-mini-and-google-sheets-4436
# Analyze BeyondPresence Video Calls with GPT-4o-mini and Google Sheets
### 1. Workflow Overview
This workflow automates the analysis of video calls recorded by BeyondPresence, leveraging GPT-4o-mini AI for conversational insights and storing results in Google Sheets for easy access and reporting.
**Target Use Cases:**
- Post-call analysis for video agent interactions
- Automated extraction of call summaries, sentiments, and action items
- Centralized storage of call analytics data for team review and follow-up tracking
**Logical Blocks:**
- **1.1 Input Reception:** Receives call data via a webhook triggered when a BeyondPresence video call ends.
- **1.2 Initial Validation & Enrichment:** Validates webhook data integrity and enriches with calculated metadata (e.g., call duration, time of day).
- **1.3 AI Processing:** Sends enriched call data to GPT-4o-mini for detailed analysis and extracts structured JSON response.
- **1.4 Data Storage:** Appends analyzed call results into a configured Google Sheets document for record keeping.
- **1.5 Optional Error Handling & Extensions:** Suggested nodes for error management and further integrations such as Slack notifications or CRM updates (documented but inactive).
---
### 2. Block-by-Block Analysis
#### 2.1 Input Reception
**Overview:**
This block captures the BeyondPresence call end event via webhook, acknowledges receipt to the source system, and filters events to process only relevant call-ended types.
**Nodes Involved:**
- BeyondPresence Webhook
- Acknowledge Webhook
- Filter Call Events
**Node Details:**
- **BeyondPresence Webhook**
- Type: Webhook Trigger
- Configuration: Listens on path `/beyondpresence-call-webhook` for HTTP POST events. Raw body parsing disabled.
- Inputs: External BeyondPresence system webhook calls
- Outputs: JSON payload containing call metadata, transcript, and evaluation metrics
- Edge cases: Missing or malformed HTTP payloads; network interruptions; webhook security not detailed here.
- **Acknowledge Webhook**
- Type: Respond to Webhook
- Configuration: Returns HTTP 200 with JSON confirmation including status, message, and timestamp based on current time.
- Inputs: Output from webhook node
- Outputs: HTTP response sent back to webhook caller
- Edge cases: Delay in response could cause webhook retries; ensure response sent before downstream processing.
- **Filter Call Events**
- Type: Switch
- Configuration: Checks if `body.event_type` equals `"call_ended"`. Routes matching events to processing; others to fallback output.
- Inputs: JSON from acknowledge webhook
- Outputs: Valid call ended events forwarded; others ignored or handled separately
- Edge cases: Event type missing or unexpected; configured to continue on fail to avoid blocking workflow.
---
#### 2.2 Initial Validation & Enrichment
**Overview:**
Validates the presence and structure of critical call data fields, calculates call duration and contextual metadata, and identifies presence of action items in messages.
**Nodes Involved:**
- Validate & Enrich Data
**Node Details:**
- **Validate & Enrich Data**
- Type: Code (JavaScript)
- Configuration:
- Checks for required fields: `event_type`, `call_id`, `call_data` (with `userName`, `startedAt`, `endedAt`), and `messages` array.
- Throws error if validation fails, halting processing for that input.
- Calculates:
- Call duration in minutes and formatted string
- Day of week from call start time
- Time of day classification (Early Morning, Morning, Afternoon, Evening, Night)
- Message count
- Boolean flag if messages include action items (keywords like "will do", "follow up")
- Inputs: JSON call data from Filter Call Events node
- Outputs: JSON enriched with validation status and calculated fields
- Edge cases: Missing or malformed timestamps; empty or non-array messages; logic errors in keyword detection; fails gracefully with error messages.
---
#### 2.3 AI Processing
**Overview:**
Uses OpenAI GPT-4o-mini to analyze the enriched call data and generate a structured JSON summary including sentiment, main points, action items, and recommendations.
**Nodes Involved:**
- AI Call Analysis
- Parse AI Response
**Node Details:**
- **AI Call Analysis**
- Type: LangChain OpenAI Node
- Configuration:
- Model: `gpt-4o-mini`
- Temperature: 0.3, Top-p: 0.9 for balanced creativity and determinism
- System message: Detailed instructions on analysis requirements and strict JSON output format
- User message: Injects call data, evaluation, and messages as stringified JSON for context
- Inputs: Validated and enriched JSON from previous node
- Outputs: AI-generated message content with JSON embedded in markdown code block
- Edge cases: API authentication errors; rate limiting; malformed AI output; timeout or incomplete response; dependency on accurate prompt formatting.
- **Parse AI Response**
- Type: Code (JavaScript)
- Configuration:
- Extracts JSON content from AI response within markdown code blocks or directly from content if code block missing
- Cleans JSON string from comments, trailing commas, and fixes common JSON formatting issues before parsing
- Throws informative errors if parsing fails
- Inputs: Raw AI text response
- Outputs: Parsed JSON object conforming to analysis schema
- Edge cases: AI returns incomplete or invalid JSON; unexpected formatting; fallback parsing failures.
---
#### 2.4 Data Storage
**Overview:**
Appends the structured call analysis data into a preconfigured Google Sheets document for archival and reporting.
**Nodes Involved:**
- Save to Google Sheets
**Node Details:**
- **Save to Google Sheets**
- Type: Google Sheets Node
- Configuration:
- Operation: Append row
- Document ID: User must specify by copying provided template sheet and entering new sheet ID
- Sheet Name: Uses first sheet (`gid=0`)
- Columns mapped to analysis fields: Date, Title, Topic, Summary, Follow-ups, Participant, References, Action Items, Main Points, Sentiment info, and more (total ~30 columns)
- Cell format: USER_ENTERED to allow formulas and formatting
- Inputs: Parsed JSON from AI analysis
- Outputs: Confirmation of row append
- Edge cases: Invalid or missing sheet ID; insufficient permissions; API quotas; schema mismatch if template altered; network issues.
---
#### 2.5 Optional Error Handling & Extensions (Documented Notes)
**Overview:**
Provides guidance on optional enhancements like error logging, Slack notifications, CRM integrations, and database backups to extend workflow functionality.
**Nodes Involved:**
- Error Handling Note (sticky note)
- Extension Options (sticky note)
**Node Details:**
- **Error Handling Note**
- Type: Sticky Note
- Content: Suggests adding an error trigger node to capture failures, log to separate sheet, notify admins, and retry failed operations.
- Not connected in current workflow, serves as best-practice advice.
- **Extension Options**
- Type: Sticky Note
- Content: Recommends integrations for Slack notifications on sentiment/action items, database backups for analytics, and CRM updates for call records.
- Not implemented but provides roadmap for future enhancements.
---
### 3. Summary Table
| Node Name | Node Type | Functional Role | Input Node(s) | Output Node(s) | Sticky Note |
|---------------------------|------------------------------|---------------------------------|------------------------------|-----------------------------|-------------------------------------------------------------------------------------------------|
| Workflow Overview | Sticky Note | Overview and instructions | — | — | Describes webhook trigger purpose and setup instructions |
| Processing Steps | Sticky Note | Explains validation and AI steps| — | — | Details data validation, AI analysis, and parsing logic |
| Data Storage | Sticky Note | Google Sheets setup instructions| — | — | Provides quick start guide and link to template sheet |
| BeyondPresence Webhook | Webhook | Receives call end event | External webhook | Acknowledge Webhook | See Workflow Overview note for webhook config instructions |
| Acknowledge Webhook | Respond to Webhook | Sends 200 OK response | BeyondPresence Webhook | Filter Call Events | Same as above |
| Filter Call Events | Switch | Routes only call_ended events | Acknowledge Webhook | Validate & Enrich Data | Validates event type filtering |
| Validate & Enrich Data | Code (JavaScript) | Validates and enriches data | Filter Call Events | AI Call Analysis | Handles data integrity and adds calculated fields |
| AI Call Analysis | LangChain OpenAI | Generates AI summary | Validate & Enrich Data | Parse AI Response | Uses GPT-4o-mini with detailed prompt |
| Parse AI Response | Code (JavaScript) | Parses AI JSON output | AI Call Analysis | Save to Google Sheets | Extracts and cleans JSON from AI response |
| Save to Google Sheets | Google Sheets | Stores analysis results | Parse AI Response | — | Requires Google Sheets ID; maps AI data to columns |
| Error Handling Note | Sticky Note | Suggests error handling | — | — | Advises adding error triggers and logging |
| Extension Options | Sticky Note | Suggests additional integrations| — | — | Lists optional Slack, database, CRM integrations |
---
### 4. Reproducing the Workflow from Scratch
1. **Create BeyondPresence Webhook Node**
- Type: Webhook
- HTTP Method: POST
- Path: `beyondpresence-call-webhook`
- Raw Body: Disabled
- Response Mode: Response Node (to allow separate response node)
2. **Create Acknowledge Webhook Node**
- Type: Respond to Webhook
- Connect input to BeyondPresence Webhook output
- Response Code: 200
- Respond With: JSON
- Response Body:
```json
{
"status": "success",
"message": "Webhook received and processing",
"timestamp": "{{ $now.toISO() }}"
}
```
3. **Create Filter Call Events Node**
- Type: Switch
- Connect input to Acknowledge Webhook output
- Add rule:
- Condition: `$json.body.event_type == "call_ended"` (case sensitive)
- Output key: `call_ended`
- Fallback output: `extra`
- Continue on fail: Enabled
4. **Create Validate & Enrich Data Node**
- Type: Code (JavaScript)
- Connect input to Filter Call Events `call_ended` output
- Paste provided JS code that:
- Validates required fields in `body`
- Calculates call duration, day/time info, message count, action items presence
- Throws error if validation fails
- Continue on fail: Enabled (optional but recommended)
5. **Create AI Call Analysis Node**
- Type: LangChain OpenAI
- Connect input to Validate & Enrich Data output
- Model ID: `gpt-4o-mini`
- Temperature: 0.3
- Top-p: 0.9
- Messages:
- System message with detailed instructions and JSON output format (use provided prompt with dynamic expressions)
- User message embedding call data, evaluation, and messages as JSON strings
- Credential: Configure OpenAI or compatible API credential
6. **Create Parse AI Response Node**
- Type: Code (JavaScript)
- Connect input to AI Call Analysis output
- Paste provided JS code that extracts JSON from the AI response content, cleans and parses it
- Continue on fail: Enabled (optional but recommended)
7. **Create Save to Google Sheets Node**
- Type: Google Sheets
- Connect input to Parse AI Response output
- Operation: Append
- Document ID: Enter your copied Google Sheet ID
- Sheet Name: Use first sheet (`gid=0`)
- Map columns according to JSON fields: Date, Title, Topic, Summary, Follow-ups, Participant, Action Items, Sentiment, etc. (map all fields as per node configuration)
- Credential: Configure Google Sheets OAuth2 credentials
- Cell Format: USER_ENTERED
8. **Add Sticky Notes for Documentation** (optional but recommended)
- Create notes for Workflow Overview, Processing Steps, Data Storage, Error Handling, and Extension Options with the provided content for clarity.
9. **Test and Activate Workflow**
- Deploy the workflow and test with a sample BeyondPresence webhook POST containing a call_ended event with full call data.
- Confirm successful AI analysis and row append in Google Sheets.
---
### 5. General Notes & Resources
| Note Content | Context or Link |
|--------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------|
| Copy the Google Sheets template to your drive before using and update the node with your sheet ID. | https://docs.google.com/spreadsheets/d/1TO6-jkCtoSFNLJObtN0UyklgdUd3ZxEnUaNvUaBjpvo/copy |
| Webhook URL must be configured in BeyondPresence dashboard under Settings → Webhooks to ensure event delivery. | BeyondPresence platform configuration |
| AI prompt enforces strict JSON output for reliable parsing and downstream processing. | Embedded in AI Call Analysis node system message |
| Recommended extensions include Slack notifications, database backups, and CRM integration for enhanced workflows. | See Extension Options sticky note |
| Suggested error handling includes logging errors to a sheet and notifying admins for smooth operational robustness.| See Error Handling Note sticky note |
---
This completes the comprehensive reference document for the "BeyondPresence Video Agent Call Analytics with AI & Google Sheets" workflow, enabling advanced users and AI agents to understand, reproduce, and extend the automation reliably.