Creation of a Human AI Assistant for Telegram groups

Hello! I am an experienced developer, capable of creating bots of any complexity, ranking TOP-4 in the "Bot Development" category. I also develop websites and API documentation in Python, where I rank TOP-2.

Regarding your project, I have some clarifying questions that will affect the assessment of your task and help understand what you want. Please write to me to clarify all the details!

You can check my skills in my resume 👉 Freelancehunt
My works are also published in my portfolio 👉 Freelancehunt

💻 I am also the CEO of a team consisting of a FrontEnd developer and a BackEnd developer! So if necessary, our team of three can organize a turnkey development for you!

… I look forward to your response, thank you.

Show original

Translate

Good evening!

The development of AI agent systems on RAG is our specialized area, so the specifications read as "familiar." You have chosen the right stack (OpenAI/Claude + RAG + Vector DB + LangGraph + PostgreSQL), as we constantly work with these technologies.

The system consists of two parts, both of which we will implement:

1. Communication analysis (supervisor control):

AI reads correspondence in Telegram groups: determines the essence of the request, open/unresolved questions, urgency, risks;
… controls managers' promises and the fulfillment of agreements;
a structured history for each client + dashboard for the supervisor;
notifications to the supervisor in case of delayed responses or risky situations.

2. First-line AI assistant (for the client):

auto-responses to typical requests via RAG: product availability, prices, and commercial information according to set rules, statuses of invoices/payments/shipments;
confirmation of request receipt + estimated processing time;
escalation: complex questions are automatically forwarded to the responsible manager, the client receives a notification, and the system monitors response speed.

Architecturally: Telegram API → pipeline on LangGraph (analysis + routing) → RAG on top of Vector DB (company knowledge base) → LLM (Claude/GPT) → PostgreSQL (history, tasks, metrics) → dashboard.

We emphasize separately: we pay attention to fact-grounding — the assistant responds strictly based on the knowledge base and rules, does not invent prices or availability. This is critical for first-line support.

The project is serious, so we will proceed in stages: MVP (RAG assistant + escalation) → communication analysis + notifications → dashboard. We will provide an estimated cost and timeline after clarifying the scope of the knowledge base and the number of groups — the final depends on this.

Details about the project and consultation — in private.

Show original

Translate

Hello
I am currently studying automation for business, specifically on n8n. If you are interested, I can help with implementation; I have done a lot before in Python. Please write to me in private messages for clarification and to discuss all the details.
The rate for cost and timelines is still tentative.

Show original

Translate

Hello. I can implement this project. If you're interested, write to me, and we will discuss.

Show original

Translate

Good day!

I have experience in developing AI systems for communication analysis, Telegram bots, CRM integrations, and LLM solutions for automating customer support.

I have reviewed your specifications — essentially, this is not just a bot, but an AI platform for:

• analyzing all communication in Telegram groups
• building contextual understanding of dialogues
• monitoring the work of managers
… • automating first-line support
• routing requests between AI and humans

### How this can be implemented technically:

I would suggest a modular architecture:

**1. Data Collector Layer**
• collecting all messages from Telegram groups
• storing dialogue history in a database
• linking messages to users and threads

**2. AI Analysis Layer**
• classifying messages (request / complaint / task / payment, etc.)
• determining customer intents (intent detection)
• extracting entities (prices, products, dates, issues)
• identifying unresolved questions
• analyzing risks and escalations

**3. Business Logic Layer**
• monitoring SLA of managers' responses
• tracking promises and their fulfillment
• prioritizing requests (low / medium / high / critical)
• automatic reminders and escalations to the manager

**4. AI Response Layer**
• knowledge base (FAQ, products, prices, rules)
• generating responses via LLM
• quick responses to typical requests
• fallback → forwarding to the manager

**5. Routing System**
• automatic determination: AI responds or a human
• forwarding complex cases to the manager
• monitoring the response time of the manager

### Important implementation points:

• using LLM with contextual memory (dialogue history)
• vector knowledge base (for quick responses)
• role system (AI / manager / supervisor)
• logging all AI decisions for quality control
• dashboard for the supervisor (risks, SLA, response quality)

### Result:

You will receive a system that:
• automatically resolves 60–80% of typical requests
• monitors managers and the quality of communication
• does not lose customer requests
• provides the supervisor with a complete picture of dialogues
• significantly reduces the load on support

I am ready to discuss the architecture in more detail and propose an optimal stack for your loads and budget.

Show original

Translate

Good day! We have experience in developing intelligent systems for context analysis in messengers. We implement this through the integration of LLM models with configured RAG for deep understanding of communication history and automation of responses. Such architecture will ensure stable operation of the bot in real-time mode and high accuracy in processing requests in your groups.

Show original

Translate

Stack 1 to 1 our: AI agents on Claude Sonnet 4.6 + GPT-4o, aiogram 3 for Telegram, PostgreSQL with pgvector, FastAPI. Similar has already been done as SaaS and CRM, works in production:

Winbix.AI: multi-tenant SaaS platform for AI agents, analyzes customer communication in Telegram, automatically responds to typical inquiries, escalates complex ones, monitors SLA, maintains history. $2K MRR, 30+ paying clients. Direct analog of your task.
BotFusion AI: SaaS chatbots on 19+ platforms with escalation.
BrandSync AI: production SaaS with GPT-4 + Whisper pipeline.

Communication analysis: Claude Sonnet 4.6 with tool use and structured output. Pipeline for each message: intent classification (product inquiry, price, status, complaint, agreement), entity extraction (promises with deadlines, risk markers, sentiment).
Automatic responses via RAG on pgvector with embeddings from your knowledge base, prices, product statuses. Configurable through admin: tone, templates, allowed topics. For non-standard cases, escalation to a manager via mention + DM with context, automatic confirmation to the client.
Promise control: extract entity "promise" with time deadlines ("we will send it tomorrow by 18:00"), task with reminder, escalation to the supervisor in case of delay.
… Risk monitoring: conversation classification normal/warning/at_risk through sentiment + keyword markers. Real-time alerts to the supervisor.
Admin dashboard: SLA metrics for managers, list of active conversations, client history, promise control.
CRM integrations: webhook to HubSpot, Bitrix24, AmoCRM, Pipedrive.
We implement via Telegram Bot API (if the bot is in groups as admin) or MTProto (Telethon/Pyrogram) for user account integration.
Quentar, a studio with 6 developers (UTC+2). For the project: 1 senior Python + AI engineer, 1 fullstack for admin, 1 QA.
Portfolio: winbixai.com/ru/startups. Loom overview of Winbix.AI with message analysis and escalation flow will be sent privately.
Guarantee: 30 days post-launch support with AI calibration at our expense.
Questions:

How many active Telegram groups with clients?
Is the bot in groups as admin, or is a userbot needed through a personal account?
What CRM is being used?

Show original

Translate

Hello! What specific data source for communication analysis do you plan to use — will it be your own Telegram groups, or do you mean integration with existing chats?

I can provide more details about the timeline and budget in personal correspondence.

Here’s how I propose to implement this project:
1. I will set up the integration of the AI assistant with your Telegram groups to read and analyze messages.
2. I will develop the logic for determining the essence of requests, urgency, and risks, as well as automatic responses to common questions.
3. I will set up a system to monitor the fulfillment of agreements, promises, and the response speed of managers.

… Thank you for considering my proposal. I look forward to the opportunity to collaborate with you!

Show original

Translate