Hi there, need help to building agent support-sales system, I have some knowledges in n8n and different tools, but need someone more experienced and who implemented it faster

There are different parts briefly:

Deploy a system of agents (bots) on n8n (Cloud) to:

  1. Interact with users via a Telegram bot (commands, responses using predefined prompts) using Telegram webhooks.
  2. Parse PDF/image documents (OCR) to extract passport details (full name, passport number) and save them to a CRM.
  3. Automate workflows for reminders, payment checks (crypto and/or traditional payment APIs), and report generation (“Boss”).
  4. Store documents/files on Google Drive or other external services, and forward them to the required messengers.

2. Environment and Access

  • n8n Cloud
    • The Client (you) creates an account on n8n Cloud and provides the Contractor with the necessary access (login, password, invitation link).
    • A basic or mid-tier plan is expected (to be agreed upon), considering the anticipated volume of automations.
  • CRM
    • The Contractor chooses the CRM (Bitrix24, amoCRM, Pipedrive, or any other) based on simplicity and budget.
    • The Contractor configures fields to store passport data (full name, passport/ID number).
    • Original documents may be stored in the CRM (if supported) or on Google Drive.

3. Telegram Bot (Commands, Webhooks, Prompts, Responses)

  1. Webhook-based Interaction
    • Integrate the Telegram bot via webhooks to receive user messages, commands, and file uploads in real time.
    • The user sends commands (e.g., /start, /help, /upload_document), and the bot responds using predefined texts/prompts provided by the Client.
  2. Language Model / Knowledge Base
    • Any fine-tuning or language model is the Client’s responsibility.
    • The Contractor sets up API calls (if necessary) via n8n nodes to integrate with the model’s service.
  3. Prompt Management
    • The Contractor should describe how to update or add prompts in the future (e.g., external file, spreadsheet, repository).

4. Document Parsing (OCR)

  1. File Types: JPG, PNG, PDF (scanned passport).
  2. Recognition Mechanics (Contractor’s Choice)
    • Possible services: Google Cloud Vision API, Tesseract OCR, Azure Computer Vision, ABBYY Cloud OCR, etc.
    • Main criteria: Ease of integration (existing n8n node or well-documented API) and Scalability (capable of handling more documents as usage grows).
  3. Extraction Results
    • Full Name and Passport Number (or ID number) are mandatory fields.
    • Additional fields (e.g., date of birth, address) may be included if necessary.
  4. Saving to CRM
    • The Contractor implements the logic to create or update CRM records with extracted data.
    • The scanned file can be attached to the CRM if supported, or stored on Google Drive with a reference link in the CRM.

5. Reminders, Payment Checks, and Reports

  1. Reminders
    • Integrate with a calendar (e.g., Google Calendar) or schedule workflows in n8n to send reminders (via Telegram or Email) at specified times (service reminders, deadlines, etc.).
  2. Payments
    • Integrate with crypto wallets (Binance, Coinbase, Whitepay, etc.) or standard payment services (Stripe, PayPal) via their APIs.
    • On successful payment, update the CRM record and notify the customer (Telegram/Email).
  3. Reports (“Boss”)
    • Define metrics needed (number of new requests, statuses, total payments).
    • Generate a text or tabular report and send it to the “boss” (Telegram, Email, or another channel).

6. Fallback Flows, Logging, and Error Notifications

  1. Fallback Flows
    • For each critical workflow (Telegram interaction, OCR parsing, CRM updates, payments), implement fallback paths to handle unexpected scenarios (e.g., missing data, OCR errors, network failures).
    • These fallback flows should gracefully log the issue, notify the relevant party (if needed), and allow users/admins to retry or manually intervene.
  2. n8n Cloud Logs
    • Use the built-in Execution List to track successful/failed runs.
    • Assign relevant Tags or Names to each workflow for clarity.
  3. Error Notifications
    • If an error occurs (e.g., the OCR service is unreachable, or the CRM API fails), n8n should send a notification to the contractor/administrator (Telegram or Email).
    • The logs should indicate exactly which node or step failed.
  4. Backup
    • Confirm whether regular backups of workflows are needed (e.g., exporting JSON or auto-saving to GitHub).

7. CRM Requirements (Contractor’s Discretion)

  1. Simplicity, Budget
    • Choose a CRM with a free or low-cost plan and an open API for easy integration.
  2. Passport Data Fields
    • At least two fields: Full Name and Passport/ID Number.
  3. Duplicate Prevention
    • Configure logic to update existing records if the same user (phone, email, etc.) is found, avoiding duplicates.

8. Testing and Acceptance

  1. Test Scenarios
    • Sending different file types (images, PDFs) to the bot → verifying OCR accuracy and CRM data updates.
    • Creating reminders → checking that notifications are sent on schedule.
    • Simulating transactions (crypto/traditional) → confirming CRM status updates.
    • Generating “Boss” reports → confirming correct data and format.
  2. Acceptance Criteria
    • At least 90% accuracy in recognizing passport data (full name, number).
    • Stable workflow execution without major errors.
    • Straightforward maintenance of prompts and CRM fields.

Hi, I am interested in this project.
If you want, we can have a call and discuss more about it.
Please reply here or leave me email.
[email protected]
Discord: @olaf11071107
Skype: live:.cid.dfc5b4cc2433b18b

All the bests.

Hello,
I’m interested in your project and have worked on similar automation using n8n, Telegram bots, OCR, and CRM integrations. Let’s connect to discuss how I can help. You can reach me at [email protected]

Hi, I am interested in this project.
If you want, we can make a call .
Please reply here or leave me email.
[email protected]

All the bests.