Budget: 25000 UAH Deadline: 35 days

We can create the first working version within 25,000 UAH and 4-5 weeks. I would include a Telegram bot, connection of the language model via API, materials for 5-10 modules as a managed knowledge base, specific behavior rules for Starky, response limitations outside the course, and testing on 15-20 typical questions.

I wouldn't fit a separate full admin panel into this budget. It would be more rational to make a simple update through text files or a spreadsheet so you can change materials and prompts without a developer. If WeStudy allows the insertion of an external widget or has the necessary technical capabilities, we can check this option, but the basic path through Telegram will be more reliable for launch.

Estimated API costs are 5-40 USD per month for the initial launch if we take an economical model at the level of Claude Haiku 4.5 or GPT-4o mini and limit the context to the current module. We will calculate more accurately after determining the number of students, the average number of questions per student, and the desired length of responses.

Questions for accurate assessment:
> How should the bot know the current module of the student - does the student select it manually in Telegram or can WeStudy pass progress?
> In what format will the materials for 5-10 modules be - text, PDF, Google Docs, or other?

What will be needed from you:
> module materials
> description of Starky's character and prohibited topics
> Telegram bot token
> API key of the chosen provider
> access to WeStudy only if we check the widget option

Similar examples from Ingello:
> https://business.ingello.com/fractal - similar in the logic of agency automation and managed behavior rules
> https://business.ingello.com/vorfahr - relevant as an example of AI automation for a practical business scenario
> https://systems-fl.ingello.com/ua - Ingello Systems' approach to system development

Look, there's a nuance here - the main thing is not just to connect the model, but to properly limit it to the course, perform response checks, and not let API costs spiral out of control. Without this, even a cute cat can start philosophizing in the wrong direction =) Expand

Budget: 25000 UAH Deadline: 24 days

Irina, at Starky, the key is not just to connect the LLM, but to prevent the assistant from "guessing" the material while maintaining the character's lively nature. I am implementing a Telegram bot in Python/aiogram with a separate context for the student, RAG search across 5–10 modules, query verification outside the course, and controlled responses when there is insufficient data in the materials.

I will move the materials and system prompt to a managed source so they can be updated without editing the code. I will add a dialogue log, API cost limits, and a set of tests for typical, ambiguous, and external questions. The current module can be determined through code/button in Telegram or synchronized with WeStudy if the platform provides an API or webhook.

Before starting, I will clarify the format of the materials and the way the student should confirm the current module. I will calculate the API cost separately based on the number of students and expected activity, without inflated assumptions.

Budget: 25000 UAH Deadline: 21 days

Having read the brief carefully, I see the task clearly: a Telegram bot with the character of Starky, linked to the content of 5-10 WeStudy modules, with the ability to update materials without involving a developer.

I have commercial experience with TypeScript and NestJS, and I have already built systems with LLM integration (Claude API in particular). For this project, I will create a system prompt that keeps Starky in character, and a RAG-like connection to the course materials, so the bot responds on the topic of the lesson and does not go off track. Administration will be implemented through a simple interface or structured files, so the studio manager can update the content independently.

Testing on 15-20 typical questions from the institution is planned from the beginning, not at the end. I am committed to the deadlines of 4-6 weeks and a budget of up to 25,000 UAH.

If you have any questions or want to discuss details, feel free to write.

Budget: 25000 UAH Deadline: 24 days

Good day, Iryna.

Maintaining the character of Starky throughout all responses requires testing in real scenarios. Before launch, we will test 15–20 typical questions: whether it maintains tone, whether it strays into unrelated topics, whether it responds according to the module. Based on the results, a report with recommendations will be provided.

The motivational layer — reminders and praise for progress — works alongside the responses: Starky supports the student between lessons, not just answering questions. In the portfolio, there is a project for an AI assistant in Telegram with a knowledge base and character traits, similar to this task.

We will prepare the instruction for updating content and character in a simple manner — you will make changes independently. I will assess the API costs after clarifying the details.

Stages:

Each stage is documented, and the code is handed over to you.
You can stop after any stage.

≈7 days: 5500–8500 UAH
Minimum viable AI bot for supporting students without a complex content management system.
- Working Telegram bot with LLM integration
- Customized system prompt for the Starky character
- Course knowledge base loaded directly into the code/prompt
- Basic testing report on 15 questions

≈9 days: 7000–11000 UAH
Full-fledged digital tutor with self-management capabilities for course materials.
- Complete functionality of the Telegram bot according to the specifications
- Simple admin panel for updating module content without a programmer
- Deep customization of the RAG system for accurate responses by modules
- Detailed instruction on administration and prompt management
- Extended testing and optimization of token costs

≈8 days: 6500–9500 UAH
Advanced solution with increased automation and integration capabilities.
- Research and attempt to integrate a chat widget directly into WeStudy
- Setting up a proactive reminder system for motivating students
- Extended support during the first month after launch
- Optimization for several language models to compare response quality Expand

Budget: 25000 UAH Deadline: 21 days

I am the founder of Tyap Lyap AI.

I propose to implement Starky as a managed AI assistant for learning: with separate student sessions, context of the current module, and clear rules for handling out-of-course queries.

I will base it on the Hermes Agent, deployed on a separate server. On top of it, I will implement a Telegram bot and the course logic. The solution will not be tied to a single language model: Claude, OpenAI, or another provider can be changed without completely redesigning the bot.

I will implement:

- A Telegram bot with Starky's character;
- Integration of materials for 5-10 modules;
- Separate student sessions and context preservation;
- Restrictions on responses outside the course;
- Motivational messages and reminders;
- Updating materials and prompts via Google Docs or a simple interface;
- Testing on 15-20 scenarios, report, and instructions.

The current module will be determined through WeStudy, if the platform allows progress tracking, or through a button in the Telegram bot.

Full implementation — up to 21 days. Cost — 20,000 UAH. Estimated expenses for LLM API — 500-1500 UAH per month. The API will be registered under the studio's account.

We are - tyaplyapai.com Expand

Budget: 25000 UAH Deadline: 30 days

I have experience in developing Telegram bots based on LLM with RAG architecture (searching through course materials) and creating system prompts to give the bot a unique character.

How we will implement the project:

Platform: We will build a Telegram bot in Python (aiogram).

Model and character: We will connect Anthropic Claude (3.5 Haiku/Sonnet) or OpenAI (GPT-4o-mini). We will set up the system prompt for the character Starky and strict guardrails.

Knowledge base (5–10 modules): We will implement search through notes using a vector knowledge base.

Simple administration: We will export the prompt and materials into a convenient format (Google Drive / Sheets or a simple file config) so you can edit the content without involving a programmer.

Testing and instruction: We will test the bot with your pool of 15–20 questions and prepare a clear user guide. Expand

Budget: 25000 UAH Deadline: 21 days

Hello, Iryna. Two things that seem to be read differently here, and they define the entire architecture.

First. In Meta, you explicitly state that Starky should also answer off-topic questions (general topics, other subjects). This means that the thematic filter "only answer about the lesson" is not needed; it would actually break this requirement. The task is not to limit the bot to topics but to ensure that the cat remains the same Starky when explaining the lesson and when the student asks something unrelated. I am setting the consistency of character, not blocking.

Second, about the "current module." There is no need for integration with WeStudy or manual guessing about which module the student is in. With 5-10 modules, the entire course simply exists in the bot's context simultaneously, so it can respond to any part of the course regardless of what "Elite" outputs externally. The "current module" becomes an optional focus, not a technical problem.

The rest is within 25,000 UAH: Starky's character as per your description, independent content and prompt updates via Google Document, testing on 15-20 questions, report, and instructions. The Claude model, API on your key, and the exact monthly cost will be shown with figures from the first week of work.

The deadline is 3 weeks from the moment when the module texts and character description are available. One clarification that affects the start: are all 5-10 modules already prepared in text, or will they be added gradually?

Budget: 25000 UAH Deadline: 18 days

Good day

I will create a Telegram bot using Make + Telegram Bot API + LLM API. I will place the module materials and character rules in Airtable or Google Docs so that you can update them independently without a developer. For each student, the bot will consider the current module, respond only within its materials, and gently steer the conversation back to learning if the request goes beyond the established boundaries.

Before launch, we will check the scenarios together on typical student questions, including out-of-course requests, inaccurate phrasing, and motivational messages. The widget capability in WeStudy should be clarified separately regarding its technical limitations; we will keep Telegram as the primary channel.

I will calculate the monthly cost of the LLM API after understanding the number of students and average activity in the chat.

How many students do you plan to enroll in the first cohort, and how are they currently accessing the modules?

Budget: 25000 UAH Deadline: 14 days

Good day! The SDEV team is ready to integrate Starky into your course. We have extensive experience in developing intelligent bots and automation systems, as seen in the cases of AcademyGym and Alhim. We will implement a tutor that will effectively interact with students and support the learning process. We guarantee stability and quality integration with your platform. We are ready to discuss the details.

Budget: 25000 UAH Deadline: 35 days

Good day!

Thank you for the detailed brief. I am ready to implement the AI assistant Starky in the format of a Telegram bot with LLM integration and a course knowledge base (RAG), so that the responses are formed based on the lesson materials.

Within the project, I can perform:

the development of the Telegram bot;
integration with the LLM API;
configuration of the Starky character and system prompt;
connection of course materials;
the ability to update content without changes to the code;
testing and preparation of instructions.

Estimated implementation time: 4–5 weeks.

Development cost: 25,000 UAH (fixed price).

API expenses depend on the number of users and messages, approximately at the start — $5–30 per month.

I am ready to discuss the details and format of work. Expand

Budget: 25000 UAH Deadline: 7 days

Key points — "current module": WeStudy does not provide the student's progress externally, so the bot will not know where the student has stopped. The simplest solution is to have a module code at the end of each lesson: the student enters it into the bot, and then Starky responds strictly based on the notes of that module.

Python + Claude API (caching long prompts reduces dialogue costs and keeps the character's framework more stable). Each module and Starky's prompt is a separate text file; if the text is changed — /reload in the admin panel, without a developer. I catch any deviation from the course with a check before the response, not just by asking in the prompt: otherwise, students bypass it in 5 minutes.

Experience: AI assistant for CHARACTER park — created after the unsuccessful bot of previous performers.

Tokens: approximately 300–600 UAH/month for 30–50 active students.

I am within the budget of 20,000 UAH, ~1 week. 15 minutes for a call — to verify the modules and Starky's character?

Budget: 25000 UAH Deadline: 40 days

Hello! I am Bohdan, a Python developer, originally from Vinnytsia — if needed, we can meet offline. The specifications are well written, so briefly by points.
25,000 UAH fixed for the entire described scope: bot, character Starky, 5-10 modules, administration without a developer, testing on 15-20 questions, report and instructions. Plus one round of prompt adjustment after you test the bot yourself. This is ~60 hours; for the grant, a fixed price is more convenient than hourly — one figure in the report.
Deadline: 3 weeks from the moment I receive the texts of the modules and the description of the character Starky. In a week, you will already have a working bot with one module to feel the tone, rather than waiting blindly.
Model — Claude from Anthropic: it handles Ukrainian well, adheres strictly to character instructions (this is exactly your requirement about the course boundaries) and has prompt caching — the module text is not paid for again for each question. I will make the model a switch in the configuration, you can choose more expensive/cheaper yourself.
API — separately, and you pay directly: you register the key with the studio, the Anthropic bill goes to your card, I have no access to your money, and the bot works independently of me. Approximately $10-25/month with 40-50 active students sending 5-15 messages per week. Without caching, it would be three times more. I will verify the rate on the launch day and show real figures from the first week.
Next, two places where I will speak frankly, not pleasantly.
No one can guarantee that the bot will never respond outside the course — it is a language model, not a word filter. I will create two layers (character instructions plus topic verification before sending) and log such attempts so you can see where it went off track, and we can address it specifically. Anyone who promises "100% within the course" is misleading.
You do not need a vector database, and I do not offer it. With 5-10 modules, it is an unnecessary layer that undermines your requirement for simple administration: after each edit, the text will need to be re-indexed. I place the module text directly in context, and updates look like this — you edit a regular Google document, and the bot picks up the changes automatically. When the course grows to several dozen modules, vector search will become an upgrade on top of the ready-made.
I leave Telegram basic not because it’s simpler, but because it’s more accurate for the goal: you support the student between lessons, and the widget works only when the student is already on the platform. Reminders live in push notifications on the phone. The widget is possible if "Elite" allows inserting custom code — this is a question for WeStudy support, and until there is an answer, I do not promise it and do not include it in the price.
Two questions that change the implementation. What is the age of the students? If there are children — the guardrails are stricter and access to the studio's history of dialogues.

Budget: 25000 UAH Deadline: 7 days

Good day, I have extensive experience in developing Telegram bots and integrating various APIs, including language models (LLM).

I can implement your AI assistant Starky, set the right character for it, and create a convenient admin panel (for example, using Laravel + Filament) for easy updating of course materials and prompts without involving a developer.

Please let me know if the course materials for the bot will be uploaded simply as texts through the admin panel, or if they need to be extracted via API directly from the WeStudy platform?

Let's discuss all the details regarding the budget and token expenses, and start our collaboration, okay? 🙂

Best regards, Denis

Budget: 25000 UAH Deadline: 35 days

We already have a practically ready solution for an AI assistant in Telegram, which can be quickly adapted for Starky and launched within the course. We can discuss the details here now; I'm here on the marketplace.

Regarding the estimate - 25,000 UAH fixed, timeline - 4-5 weeks.
Costs for the language model API are separate - approximately 500-2000 UAH per month with moderate load; we will calculate more accurately after determining the number of students and the average number of messages.

I see the basic scenario as follows - a Telegram bot, connecting Claude or OpenAI, a separate instruction for Starky, searching for answers only from the materials of 5-10 modules, limiting topics outside the course, simple updating of materials through a file or a small admin panel, testing on 15-20 questions.
If WeStudy does not provide convenient access to student progress, we can keep it simple - we will make a selection of the current module in the bot or a simple manual logic.

From you, we need the materials of the modules, a description of Starky's character, rules for allowed and prohibited responses, access to create the Telegram bot, and if necessary, access to WeStudy to check the option with the widget.

I would like to clarify two points:
- How should the bot understand the current module of the student - through WeStudy, manually in the bot, or according to the course progression scenario?
- Approximately how many students will there be at the start, and how many messages do you expect from one student per day?

Relevant examples from Ingello:
- https://business.ingello.com/fractal - AI agents and automation of complex scenarios
- https://business.ingello.com/tts - working with AI assistants and voice scenarios
- https://systems-fl.ingello.com/ua - briefly about our approach to system development and automation Expand

Budget: 25000 UAH Deadline: 15 days

Hello. I have carefully reviewed the task regarding the development of the AI assistant Starky for your online course. For implementation, I propose an architecture based on the Python framework using the Retrieval Augmented Generation (RAG) pattern, which will ensure accurate responses from Starky solely within the course content and optimal use of the LLM API. The system will be designed with a focus on ease of administration of course materials and system prompts, high response speed, and an appropriate level of data security. I have significant experience in creating similar integrations and a ready set of technical solutions for rapid deployment and effective configuration of chatbots, which will significantly accelerate the development process. I suggest discussing all implementation details, final budget, and timelines in private messages.