Refinement of the finished n8n workflow: Worksection + OpenAI + Telegram (buttons are not working)

I don't know, I would build everything from scratch. The technical specification is high-quality, clear, it's pleasant to deal with such a client, I'm sure. So I will build it from scratch, it will work as the client needs.

Show original

Translate

Good day 👋
I have reviewed the task — there is already a good level of automation here, and it is clear that the workflow is almost ready for production. The problem lies more in the logic of states, callback handling, and the stability of scenarios in n8n.

I can refine the existing workflow without rewriting it from scratch:

• I will fix the Telegram callback_query and the operation of inline buttons;
• I will implement correct logic for Confirm / Regenerate / Copy / Reject;
• I will add protection against resending;
• I will implement state saving through staticData;
… • I will set up a regeneration loop via OpenAI;
• I will add support for voice edits PM → speech-to-text;
• I will implement a reminder flow for Worksection + Telegram;
• I will create anti-spam logic for projects;
• I will check the handling of long messages in Telegram;
• I will tidy up the logs and the security of tokens.

I can also optimize the workflow itself to make it:
— more stable;
— easier to maintain;
— free of "hangs" and duplications;
— ready for scaling to other projects.

I have experience working with:
• n8n
• Telegram Bot API
• OpenAI API
• Worksection API
• state management / automation logic

Message me privately — I will take a look at your JSON workflow and immediately tell you where the problem lies and how best to implement it 🚀

Show original

Translate

Hello.

The scope is clear — three blocks: finalize the Telegram buttons in the existing workflow (Confirm, Redo, Copy, Reject), add reminder logic in Worksection and the internal chat when there are no updates, deduplicate via staticData to avoid duplicate reminders.

Regarding the buttons — I will implement them through separate Telegram CallbackQuery with unique callback_data for each action plus a processed flag in staticData by message_id, so that repeated presses do not duplicate the sending. For the redo with voice, I will catch the voice message from the PM, convert it to text using the Whisper API, and add it to the context for regeneration through OpenAI with the same system prompt as in the existing workflow.

For reminders — check via staticData with a key of the type reminder per task_id per period with the value of worksection_sent_at or telegram_sent_at. On the first run without an update, we write in Worksection, on the next one (if there are also no comments) in the internal Telegram group. The period resets when a new substantive comment appears in Worksection.

Regarding spam protection — staticData in n8n is suitable, but if you want it to be more reliable upon instance restart, you can duplicate it in a separate micro-DB (SQLite or Redis), depending on the infrastructure where n8n is running.
…
One point that will affect the timeline — whether there is currently a working self-hosted n8n or a cloud version, and access to the Worksection API (a token is needed plus a list of task_id we are working with). As soon as I receive the JSON and access — I will deploy it locally in a day, reproduce the current flow, and then follow the scope.

Show original

Translate

Good evening, I am a Python developer and I also work with automation on n8n. I have experience working with n8n and can complete your task according to the specifications. I am ready to start immediately.
Feel free to write - we can discuss the details.

Show original

Translate

Hello! I am ready to fix your workflow and make the TG buttons work smoothly. I will debug the callback_query and staticData processing to eliminate duplicate submissions and clearly return only clean text to the client.

I will implement two-step reminders via the Worksection API and TG chat with spam protection once a day. For the "Redo" button, I will connect Whisper for voice and a cyclic request to OpenAI.

I will provide a verified JSON, test screenshots, and instructions for scaling.

I am ready to discuss the details and start the execution.

Show original

Translate

Pavel, hello! I looked at the technical specifications and the attached JSON workflow. There is no need to rewrite from scratch: there is already a base — Worksection/OpenAI node + a separate Telegram callback handler.

What I see: pmDrafts are already saved in staticData, the approve/rewrite/copy/reject buttons are in place, and voice rewrite through Whisper is also partially available. The main areas for improvement are the idempotency of the buttons, blocking old drafts, Worksection/Telegram reminders, and the state machine against duplicates.

I will do the following:
- approve sends only clean text to the client and does not duplicate;
- reject blocks old buttons;
- copy/rewrite work without service information;
- I will add a Worksection reminder in the absence of comments;
… - I will add a repeated Telegram reminder to the specialist;
- I will save the reminder state in staticData;
- I will add a configuration for the specialist's fields and fallback statuses;
- I will test the scenarios and provide the updated JSON + instructions.

Deadline: 4 days. I will work on top of the current workflow, without rewriting from scratch.

Question: do the callback buttons not reach the handler at all, or does the callback come but the logic inside breaks?

Show original

Translate

Hello! I can implement it, please write to me privately to discuss all the project details. I hope for cooperation!

Show original

Translate

Hello! I have reviewed your project and am ready to start working immediately. I guarantee excellent quality and promptness.

Show original

Translate

Good day! I have analyzed the technical specifications and your JSON file, and I am ready to refine your existing script to meet the new requirements.

What I will specifically do:
Telegram buttons: I will implement the removal of buttons after clicking (to prevent duplicates) and clear the text of service blocks before sending it to the client.

Voice edits: I will integrate the uploading of audio files and their transcription via OpenAI Whisper (whisper-1) directly within the JS code of n8n.

Smart reminders: I will set up the final automation in staticData (the first ping — via the Worksection API add_comment, the repeat after 24 hours — in the specialist's internal Telegram).

… Configuration: I will add all new variables (specialist_name, chats, limits) to your current project array. API keys will remain in $env.

Show original

Translate

Hello! I am Volodymyr, the founder of the engineering agency Vaysed. Here is a brief technical plan on how we will rewrite and optimize your current n8n workflow:
1. State Management
We will utilize the built-in method $getStaticData() within n8n. The script will start to "remember" the history of its previous runs. This will allow for reliable spam protection: the bot will check the saved status and send reminders in Worksection or Telegram strictly once during a specified period, resetting the counter after a comment appears.
2. Isolated Configuration
All hardcoded data (chat IDs, specialist names, task links) will be moved to a single starting configuration node (Set node) or environment variables. You will be able to instantly scale this script to new projects by simply changing the input data, without touching the internal logic of the bot at all.
3. Telegram Button Routing
We will completely rewrite the callback_query handling. When "Confirm" is pressed, the script will strictly filter the JSON object, separating the service information, and will send only the clean final text to the client chat. When "Reject" is pressed, the system will cancel the draft status and deactivate the old buttons.
4. Regeneration Module with Audio Processing
For the "Redo" button, we will set up a waiting node for a response from the PM. If it is a voice message, we will automatically pass it through OpenAI Whisper for accurate transcription. The recognized text will be integrated as an additional prompt and a request will be sent to the LLM for update regeneration.
… 5. Fault-Tolerant Work with the Worksection API
All API calls (checking the latest comments and creating new reminders) will be configured with error handling in mind. If any field is missing, the workflow will not "fail," but will switch to a backup logging branch, while access tokens will never be displayed in open execution logs.

We have already successfully implemented similar fault-tolerant routing and work with complex states (State Machine) in our project RESTART 2026. We have refined deep two-way integration with third-party system APIs in the Prudbay Clothes Support ecosystem, and the mechanisms for preventing duplicate notifications are the foundation of our monitoring system Lumos.
Here is a link to our website, where portfolios and descriptions of implemented infrastructure solutions are collected: https://vaysed.me/
There you can learn more about the mentioned projects. I look forward to your message to agree on the details and start working on the optimization.

Show original

Translate

The task is clear: refine the existing n8n workflow — fix the Telegram buttons (confirm/rework/copy/reject), add voice recognition through OpenAI Whisper for "Rework", implement protection against duplication, and set up reminder logic in Worksection and Telegram when there are no comments.

I work with n8n, Telegram Bot API, and OpenAI — that's the stack. You will provide the JSON workflow, and I will figure it out and deliver it without rewriting from scratch.

Approximately 4–5 days after receiving the JSON and access.

What is the main issue with the buttons right now — do they not respond at all, or is the callback received but the logic does not trigger?

Show original

Translate

Hello. I see that the workflow is already working with Worksection, OpenAI, and Telegram - we just need to refine the inline buttons and add the reminder logic. The buttons "Confirm", "Redo", "Copy", "Reject" are standard callback handlers in n8n, but there are many nuances: deduplication of sending, voice recognition, saving the draft status between iterations. I did something similar for a PM bot on aiogram - there was also a chain editing of content with voice edits through the Whisper API. Question: are the voice edits Telegram voice messages, or are you using some other service for recognition?

Show original

Translate