
How Mau Saver Works: From a Link to a File
Inside Mau Saver: platform workers, Kafka, Telegram file_id caching, media delivery and common failure points.
One chat, several different jobs
For the user, the process looks simple: send a link and receive media in Telegram. The system has to recognize the source, find a suitable format, obtain the file and deliver it. Each stage has its own failure modes, from unavailable content to Telegram limits.
This overview is based on the Mau Saver repository code reviewed on 11 November 2026. Individual platform paths differ, and an option in the code does not by itself prove it is enabled in the current deployment. The diagram is an overview, not an inventory of every service.
From link to reply
- Message and handler selection. The bot receives a link and routes it to the relevant platform logic. Sources return different structures: a single video, photos, an album or audio.
- Repeat request check. In a shared handler used for some platforms, the bot first looks for a stored Telegram file_id associated with the link. If a suitable entry exists, it can resend the media without downloading it from the source again. Otherwise, the request moves to the downloader.
- Retrieval and processing. Separate services handle particular platforms. For example, the YouTube code uses yt-dlp and FFmpeg to retrieve and prepare media; not every file necessarily needs transcoding.
- Delivery and result. The bot sends the suitable file to chat and, in some flows, stores the resulting file_id for reuse. When a source or delivery is unavailable, the user needs an error outcome rather than an unexplained wait.
Overview: for a repeat request, some handlers can reuse a stored Telegram file_id.
Section flow screenshotWhy separate handlers?
Platforms have different link formats, errors and access rules. The repository contains separate download services for several sources and separate Kafka topics for some requests. This lets us change source-specific logic independently and organize worker capacity by path.
Separation does not guarantee total isolation: network, disks, database, broker and Telegram may still be shared dependencies. A shared bottleneck can affect several sources, so investigation needs both the platform service and the common delivery path.
How file_id changes a repeat request
After a successful send, Telegram can return a file_id. For some media, Mau Saver code stores the relationship between link, type and identifier in the database, alongside a short-lived in-process cache. On a subsequent match the bot can ask Telegram to resend the file using file_id. [1]
That avoids another source download for the applicable cached path, but does not mean every URL is cached or will work forever. Media type, source changes and storage or delivery errors matter. A file_id is specific to its bot; file_unique_id cannot replace it for resending. [1]
Queues, waiting and retries
The code defines Kafka request and response topics for individual platforms. A request gets an identifier used to match the response to the waiting operation, which has a timeout. This separates processing from message intake, but raises questions about queue delay, unavailable workers and repeated events.
Repository documentation describes a risk during long YouTube processing: a lengthy transcode can exceed Kafka’s poll interval and the expected response can be lost. The lesson is to align timeouts with heavy work, rather than treating a progress message as proof of completion.
A separate Rust delivery worker is also present in the code, but the repository documents a feature flag and staged rollout. Without checking deployment configuration, we cannot claim it handles all current traffic.
Telegram limits and large files
Delivery method matters as well as file size. Telegram documents a 50 MB multipart upload limit for other media on the standard Bot API, while a local Bot API Server supports uploads up to 2,000 MB. Different rules apply when resending by file_id. These are API capabilities, not a promise that Mau Saver can process every file of that size. [1]
Large files take longer to transfer and consume storage and bandwidth. Check the limits of the active configuration, available disk space, temporary-file cleanup and interrupted delivery separately.
What can fail and what to measure
| Symptom | Investigate | Useful signal |
|---|---|---|
| Source returns no media | Content availability and changes to the source page or response. | Errors by platform and type. |
| Request waits too long | Worker load, queue, timeouts and processing time. | Queue wait separately from execution time. |
| File ready but not delivered | Telegram response, size and delivery retries. | Delivery latency and errors/retries. |
| Repeat requests do not speed up | Cache key, file_id availability and resend outcome. | Cache hit rate and cached-delivery failures. |
What it takes to operate the pipeline
The budget is more than the bot server: include video-processing compute, temporary storage, network traffic, broker, database, monitoring and engineering time. Actual costs depend on file size, repeated links, workload and configuration. We do not publish unverified current costs here.
Only save media you have rights to use, and account for source rules and chat privacy. User guides: add the bot to a group and process a YouTube link. For the other side of the problem, see bot development costs.
Sources
Learn more about Mau Saver
The product page describes the bot’s features and how to get started.
Related Articles

How to Add Mau Saver to a Telegram Group and Save Media in Chat
A practical guide to adding Mau Saver to a Telegram group and downloading videos, audio, and posts directly in the chat.

How to Save YouTube Videos in Telegram with Mau Saver
Send a YouTube link to Mau Saver and receive video, audio, or a playlist in Telegram. A simple guide for everyday media workflows.

Telegram Bot Development Cost in 2026: Is It Free? Real Budgets
The Telegram Bot API is free, building a business bot is not. Real 2026 budget ranges for CRM, payments, Mini Apps and admin panels, plus how to scope without hidden costs.