AI Assistant for your business

A smart chat that answers from your documents — and always shows where each answer came from.

What is it

This is a RAG assistant (retrieval-augmented generation). It doesn't make things up — it finds the relevant pieces in your knowledge base and answers from them, with links to the sources. A live example runs right on this site: the chat bubble in the bottom-right corner.

How it works

1

You give us your knowledge base

Documents, FAQs, product descriptions, policies — whatever you already have. No special format required.

2

We build a searchable index

We split the content into pieces and turn them into embeddings — a search index that understands meaning, not just keywords.

3

The assistant retrieves the relevant pieces

For every question it pulls the most relevant fragments from your data.

4

It answers with sources

The reply is grounded in your data and shows which sources it used. Anything tricky is handed off to a human.

What we need from you

  • Your materials: documents, FAQ, website, product info, policies
  • The tone of voice and languages the assistant should reply in
  • The channels where it should live: website, Telegram, WhatsApp, etc.
  • A way to update the knowledge base as your content changes

The data the assistant answers from

Below are the verbatim fragments of the knowledge base — the exact texts the assistant retrieves and reads. Open any fragment, read it, then verify: ask the assistant a question in the chat and compare its answer with the source.

Services5
Automate business processesRead
Together, we will identify a technical solution that simplifies tasks for you and your employees. For instance, we can automate corporate newsletters or regular reports.
Create bots for all platformsRead
We can develop bots for various purposes, such as processing client orders or automatically collecting reviews. The platform is chosen to suit your business, whether it’s Telegram, WhatsApp, or others.
Build websites for various purposes and businessesRead
We’re here to help create an intuitive website for your business that enhances user trust. Whether you need a website built from scratch or based on an existing platform, our editor and designer will craft the content and visual elements perfectly.
Develop apps for smartphones and desktopsRead
We'll identify the ideal solution for your app's purposes. We can design applications for popular operating systems, including iOS, Android, MacOS, and Windows.
Provide technical support for your projectRead
We are open to transferring the code rights to you, especially if you already have IT specialists on your team. We can also provide ongoing project support, ensuring the high-quality operation of the service. You’ll be able to reach out to us with any technical issues.
Portfolio cases13
Helpdesk — Ticketing BotRead
[Telegram API, TypeScript, Node.js, MongoDB, REST API] A Telegram bot integrated with the client's helpdesk system: users create support tickets right in the chat and receive responses back in Telegram. Problem: The client had a large, established product but lacked a convenient channel for creating and managing support requests. They needed a solution that would allow users to contact support via a Telegram bot, automatically create support tickets, and send responses back to the same chat. Solution: We designed and implemented a Telegram bot integrated with the client's helpdesk system: • Receives text messages, images, files, and voice messages from users • Automatically creates corresponding support tickets via API • Tracks ticket status and delivers responses from support staff directly to the user's chat • Supports two-way transfer of both text and media — images, documents, and voice messages are exchanged in both directions The integration was built using APIs and webhooks, ensuring stable data exchange, preserving attachment structures, and providing scalability and data security. Result: • Reduced support request processing time by 40% through automated message intake and routing • Simplified communication — users now receive answers directly in Telegram without switching platforms • Increased customer satisfaction, reflected in a noticeable rise in positive feedback
Request Form BotRead
[Telegram API, Node.js, TypeScript, Google Sheets] A Telegram bot that guides users through multi-section forms and automatically saves structured responses to Google Sheets. Problem: The client needed to collect structured information from users through multi-section forms, automatically save responses to Google Sheets, and notify the team — all without manual data entry or follow-up. Solution: Built a Telegram bot that guides users through multi-section forms based on their selected scenario. All collected data is automatically saved to Google Sheets with proper formatting. The bot sends auto-notifications to the team for each completed form and supports reporting on submissions. Designed to handle multiple form types without code changes.
Internal CRM Sync BotRead
[Telegram API, TypeScript, Node.js, MongoDB] A Telegram bot that verifies group members against the client's CRM and automatically removes users without active subscriptions. Problem: The client needed to automatically verify Telegram group members against their internal database and remove users without active subscriptions. Manual checks were slow, inconsistent, and didn't scale across multiple groups. Solution: Built a Telegram bot that automatically checks group members against the client's internal CRM database. Users without an active subscription are automatically removed. The bot also supports segmented broadcasts by criteria — sending targeted messages only to relevant user groups. All actions are logged for audit.
VPN Client — macOS, Android, iOS, WindowsRead
[TypeScript, Java, Swift, Electron, Cross-platform] A branded, feature-enhanced VPN client with Xray protocol support, deployed across all major platforms. Problem: The client needed a branded, feature-enhanced VPN client beyond off-the-shelf solutions, deployed across all major platforms. Solution: Forked an open-source VPN project and delivered a complete rebranding (icons, colors, copy). Implemented Xray protocol support, custom routing rules, and advanced features. Built for Android (Java), iOS/macOS (Swift), and Windows/macOS (Electron) with a shared TypeScript core. Stack: TypeScript, JavaScript, Java (Android), Swift (iOS/macOS), Electron, Cordova
Data Monitoring ServiceRead
[Java, Selenium, PostgreSQL, Apache POI, Web Scraping] A backend aggregator that collects economic indicators from multiple websites daily — fully automated, with zero manual intervention. Problem: The client needed a daily aggregated summary of economic indicators from multiple websites. Manual collection was slow, inconsistent, and couldn't scale. Solution: Built a backend aggregator service: Selenium for web scraping → KPI calculation → filtering → structured API export. Fully automated daily runs with zero manual intervention.
Corporate LMSRead
[Moodle, NGINX, LMS, Custom Analytics] A unified corporate training platform built on Moodle with custom interface theming and an extended analytics service. Problem: The organization lacked a unified platform for employee training and development tracking across departments. Solution: Deployed and fully configured a Moodle instance: custom interface theming, mobile responsiveness adaptation, and administrator training. Beyond Moodle's native capabilities, developed a custom extended statistics service providing flexible data exports and analytics beyond standard platform features — usage metrics at the department and individual employee level. Stack: Moodle, Nginx, custom analytics service
Corporate Knowledge Base — XWikiRead
[XWiki, Apache Tomcat, Apache Velocity, HTML, CSS] A scalable knowledge platform for one of the largest government organizations with 80+ regional branches and thousands of employees. Problem: One of the largest government organizations with a distributed territorial structure (80+ regional branches, thousands of employees) needed a scalable platform to centrally structure, store, and provide access to critical knowledge and documentation. Solution: Deployed XWiki, customized the interface, configured roles and access, and trained editors. Beyond XWiki's standard capabilities, we developed a separate service for extended statistics collection and analytical report generation, providing usage and engagement metrics at the department and individual employee level.
Workload Report Bot: Excel to BotRead
[TypeScript, Node.js, Telegram API, MongoDB, Google Sheets] A Telegram bot that automates weekly workload reporting for a 50+ employee organization — replacing 10+ hours a week of manual Excel work. Problem: Weekly workload reporting took 10+ hours: managers merged task exports, meeting spreadsheets, and time sheets into separate Excel templates per department or team by hand — slow and error-prone. They needed one repeatable flow in Telegram with feedback before numbers were finalized. Solution: An admin-group bot that walks users through the pipeline: select department or team → upload tasks → parse executors and hours with validation → choose dates (inline calendar) → upload the meetings registry and time-sheet files → output the consolidated workload file plus optional per-employee xlsx files in a ZIP. Parsing errors return an annotated xlsx with highlights and error text — not silent bad data. Roster from Google Sheets or file; templates managed in-bot (MongoDB). Result: The manual merge of three sources into compliant Excel is automated; issues surface at upload time instead of in review meetings. Stack: Node.js, TypeScript, grammY, Koa, MongoDB, Excel pipelines
Paid Community Bot — PayPal, Cards, StarsRead
[Telegram API, TypeScript, Node.js, MongoDB, Payments] A Telegram bot for paid fashion communities that handles the entire subscription lifecycle: payments, channel access, webinars, and certificates. Problem: The client is a fashion stylist running paid Telegram communities: a subscription club, webinar recordings, and certificate programs. All subscriber management was manual — tracking who paid, granting channel access, checking expiry dates, sending reminders, and revoking access. PayPal payments were taken manually outside the bot, then reconciled by hand. Solution: A Telegram bot that handles the entire subscription lifecycle automatically: • Users open the bot, pick a product, choose a payment method, and get instant access after payment • Three integrated payment paths: PayPal, PayAnyWay (bank cards), Tribute (cards + Telegram Stars) • Daily subscription checks: a reminder 3 days before expiry, auto-ban on expiry day • Products, promo codes, and webinars are managed in a Google Spreadsheet and synced on schedule or on demand • Admin work stays in Telegram: broadcast to buyers, sync from Sheets on demand • Full subscriber and subscription dataset imported from the client's materials — no manual re-entry Result: The heavy daily grind around payments and access is automated; the client only edits the sheet when content or promos change. Stack: TypeScript, Grammy, Koa, MongoDB, Railway
Warehouse Management AutomationRead
[Vue.js, TypeScript, Docker, NGINX, Keycloak] A nationwide warehouse automation system for agrochemical tracking: inventory, documents, counterparties, and receipt logging. Built a nationwide warehouse automation system for agrochemical tracking. Full-cycle automation: inventory, documents, counterparties, receipt logging, application zones. Led frontend automation from architecture to delivery. Stack: Vue.js, TypeScript, Nginx, Docker, Keycloak
Internal Process AutomationRead
[Java, Spring Boot, Vue 3, Keycloak, Apache POI] A centralized internal system that automates daily employee processes — record-keeping, data exchange, reporting, and reference information management. Problem: The organization lacked a centralized internal system to automate daily employee processes — record-keeping, data exchange, reporting, and reference information management. Employees relied on scattered tools and manual workflows. Solution: Delivered a full-cycle development project: built new modules and enhanced existing ones with role-based access control. Implemented user-facing forms, notification systems, collaboration sections, and structured directories. Contributed to both backend and frontend development. Key features: • Role-based access for different employee levels • Custom forms for data entry and reporting • Automated notifications and alerts • Collaboration sections for cross-team interaction • Structured directories for reference information • Excel/Word document generation (Apache POI) Result: Daily processes automated into a single system. Employees no longer switched between scattered tools. Role-based access ensured data security and appropriate visibility at each level. Stack: Java, Spring Boot, Vue.js, Apache POI (Excel/Word), Nginx, Keycloak
Enough Habits — App-Like Habit Tracker BotRead
[Telegram API, TypeScript, Node.js, Firestore, Amplitude] A self-compassion-oriented Telegram habit tracker that treats “good enough” as real progress. Publicly available — try it live at t.me/enough_habits_bot. Most habit trackers reward only perfect streaks, fueling burnout and guilt when users miss a day. Enough Habits flips the script by treating “good enough” as real progress, so people see they’ve done enough to rest and recharge — without guilt. What we built — a full-stack TypeScript Telegram bot that turns habit tracking into positive reinforcement: • Onboarding & philosophy: automatic RU/EN language detection, a compassionate welcome, and a 4-step wizard (three starter habits + time-zone selection) • Reminder engine: default morning/evening nudges plus optional per-habit reminders at any exact HH:MM, fully time-zone aware • Daily goal logic: tracks progress toward the “enough” target and sends contextual motivational messages • Habit management: create, edit, archive, or restore habits at any time • Progress & calendar: main list, calendar history, streak counters and totals dashboard • App-like UX: inline keyboards and message-editing flows so it behaves like a lightweight mobile app Data layer on Firestore with real-time sync; Amplitude analytics and Rollbar error tracking; modular CI/CD backend with zero-downtime releases. Result: publicly launched and free to try — “Reminded me to keep my habits on track without feeling pressured.” Stack: TypeScript, Node.js, Telegram API, Firestore, Amplitude, Rollbar. Try it: t.me/enough_habits_bot
Lead Bot + Admin Panel & MailingsRead
[Telegram API, TypeScript, Node.js, MongoDB, Google Sheets] Telegram lead capture with smart routing and automated nurturing for a creative-services firm — plus an admin console for mailings. A creative-services firm needed Telegram lead capture, smart routing, and automated nurturing. I built a full-stack bot that: • greets users, logs handle+source, offers department choice • stores user data and lead requests in MongoDB + Google Sheets • alerts team chat with each inquiry • lets managers mark leads in-progress/closed • tracks UTM-style links • provides an admin console to edit bot messages, send targeted mailings, and generate UTM-tagged links Result: ≈20 hours of managerial time saved monthly; every inquiry auto-logged in a source-tagged Google Sheet for ROI tracking.
How we work5
Getting to know you and your businessRead
We prioritize understanding the problem you aim to solve with software development. At this same stage, we set the project’s price range. You’ll get both a minimum and maximum final cost, allowing you to control your budget and providing us the flexibility to make adjustments and cater to your specific needs. We are always looking for ways to deliver cost-effective and efficient solutions.
Preparing interface mockupsRead
We present interface mockups to you so you can visualize what the final product will look like. Then, we can either develop a prototype or go straight to the final technical solution. We’ll update you at least once a week on how the work is progressing.
Testing the product and gathering your feedbackRead
At this stage, we hand over the product’s first version. You’ll try it from a week up to a month, while we continue refining it until we find the optimal solution.
Transferring rights or providing ongoing supportRead
We can transfer code ownership to you or take on full administration ourselves. This includes ensuring the system’s technical operation, timely server payments, component updates, and prompt resolution of any technical issues.
Staying in touchRead
We are always ready to further develop the project. Your satisfaction with the end result is our top priority.
FAQ10
How much does a project cost?Read
Pricing is individual. After a short free assessment we give you a minimum–maximum cost range and a recommended scope, so you can control the budget.
How do you estimate a project?Read
We start with discovery — understanding the problem you want to solve. At that stage we agree on a min–max price range and the scope of work.
How long does development take?Read
It depends on scope. We show interface mockups first, then build; you get progress updates at least once a week, and typically test the first version from a week up to a month.
Do you transfer the source code?Read
Yes. We can hand over full code ownership to you, or keep maintaining the project ourselves (servers, updates, fixes) — your call.
Do you provide support after launch?Read
Yes. We can provide ongoing support and further development. All projects come with a 6-month warranty on the build.
What technologies do you work with?Read
TypeScript, Node.js, Python, Java, React, Next.js, Vue, PostgreSQL, MongoDB, Docker; Telegram Bot API, OpenAI/Claude API, RAG systems, Google Sheets API, and more.
Can you build an AI assistant on our own data?Read
Yes — that's exactly what this website assistant demonstrates: a RAG assistant grounded in your knowledge base, with sources and human handoff. We integrate it into your site, Telegram, or other channels.
Where are you based and since when?Read
NekaSoft is a studio led by founder Aleksei Nekrasov, based in Belgrade, Serbia, working since 2019 with 100+ delivered projects.
What languages do you work in?Read
We communicate and build in English, Russian, and Serbian.
How do we get started?Read
Leave a request through the form on the site or message Aleksei on Telegram (@Nekadid). We'll do a free assessment and propose a scope and price range.
Company facts1
About NekaSoftRead
NekaSoft is a software development studio ("Friendly Tech Solutions for Business"), led by founder Aleksei Nekrasov (Telegram: @Nekadid), based in Belgrade, Serbia, working since 2019, with a team of developers and specialists and 100+ delivered projects for small and medium businesses. Contacts — Telegram: https://t.me/Nekadid, email: nekasoft.office@gmail.com, Upwork agency: https://www.upwork.com/agencies/2061816205163704592/, Upwork founder: https://www.upwork.com/freelancers/nekasoft.

Verify it yourself

Read a fragment? Now ask the assistant about the same thing in the chat at the bottom-right — and compare its answer with the source text. Under each answer it shows exactly which fragments it used.

Add your own data

Paste your own text — rules, product info, an FAQ. The assistant adds it to the knowledge base and answers from it too, alongside ours. It goes through the exact same RAG pipeline (embeddings + retrieval), and your fragments show up as sources marked “your data”.

Nothing is saved: your data lives for 1 hour in this session only, then it's gone.

Want an assistant like this for your business?