REGISTRY OF READY-MADE SOLUTIONS

Period: 2026-Q2. Stack: Python, Aiogram 3 + FSM, Google Gemini Vision, Google Sheets API, FSSP API, Court (Russian) Parser, Playwright, CAPTCHA Solver

End-to-end automation of an accident-recovery law firm: OCR of the accident report → debtor check in bailiff database → court case lookup → ruling monitoring → enforcement-proceedings control. The whole funnel in one Telegram bot and a sheet.

• OCR MODULE: Recognition of accident report photos via Gemini Vision — name, date of birth, and address of the at-fault party are extracted with no manual entry.
• FSSP CHECK: Automatic debtor lookup in the bailiff enforcement database — the lawyer immediately sees who is worth suing and who isn't.
• COURT FINDER: A custom court-database parser locates the civil case by name and region, then links the case number and URL into the spreadsheet row.
• CASE MONITORING: Scheduled crawl of case cards — new hearings, rulings, status changes — instant Telegram alert to the responsible lawyer.
• IP MONITORING: After enforcement proceedings open, the system tracks the case in the bailiff registry — status, amount, bailiff, actual collection. The lawyer sees not just the verdict but the real money.
• CAPTCHA BYPASS: Built-in captcha solver for stable access to state registries with no manual intervention.
• SHEETS AS FUNNEL: Google Sheets is the single funnel: check → court → writ → enforcement. Each lawyer sees their workload and progress.
• TELEGRAM CONTROL: All commands run through the bot — force an FSSP check, kick off case search, toggle monitoring, upload a new report.

Period: 2026-Q2. Stack: Python, FastAPI, Aiogram 3, ChromaDB, SentenceTransformers, Gemini 2.5 Flash Lite, SQLite, RapidFuzz, React + Vite (Telegram Mini App)

A 4-layer cascade AI engine that matches client Excel requests to a 30,951-position ERP catalog: from a free cache and fuzzy search to local vector search and LLM verification. The system self-learns from manager confirmations and gets faster with every new order.

• 4-LAYER CASCADE: Pipeline from cheap to smart — client mapping → SQLite cache → RapidFuzz with numeric filter → ChromaDB vector search → Gemini verification. Each layer passes candidates onward only when it is not confident.
• SELF-LEARNING CACHE: Any manager confirmation (including manual swaps via the "Replace" button) goes straight into SQLite and becomes a GREEN ≤1ms response for all repeat queries.
• NUMERIC GUARD: Regex on diameters / lengths / sizes drops fuzzy candidates with mismatched numbers even at 99% text similarity (fixed a real bug «Duct D100 = Duct D160» at score 95%).
• VECTOR LAYER: Local embedding model + persistent ChromaDB index over all 30,951 items — fully offline, no API keys, tuned for the Russian HVAC equipment domain.
• LLM VERIFIER: Gemini 2.5 Flash Lite checks ONLY YELLOW candidates (0.60 < score < 0.95) — greens from the cache and from the client hard-mapping spend zero tokens.
• AI_VERIFIED BADGE: Rows that passed LLM validation are tagged with an "AI Verified" badge in the Telegram Mini App — the manager always sees who made the call: cache, algorithm, or neural net.
• GLOBAL + CLIENT CACHE: Separate namespaces for shared synonym dictionary and per-client mappings — they don't mix and are reused correctly.
• EXCEL INGEST: The client uploads a price-list as Excel — the bot parses it, runs the cascade, and returns a status-colored result inside the Telegram Mini App.
• REPLACE ACTION: The "Replace" button on a row opens a search across the entire ERP catalog — the manager overrides the algorithmic choice, and it lands in the cache for future requests.

Period: 2026-Q1. Stack: Python, Aiogram, FastAPI, React, Google Sheets API, Vercel, OpenAI Vision

Full online-training cycle: client management automation, training and nutrition diary, progress analytics, trainer communication — all inside Telegram, no third-party apps.

• WORKOUT MODULE: The client sees the day schedule, opens an exercise, and immediately sees the previous weight pulled straight from Google Sheet.
• NUTRITION AI: Food-photo scanning via GPT-4 Vision — the system estimates calories and macros and writes them to the database.
• PROGRESS ANALYTICS: The summary screen computes weekly tonnage, nutrition accuracy, activity, and body-weight dynamics.
• PROFILE SYNC: Edits to the client profile (age, height, weight) save directly to Google Sheets.
• FEEDBACK LOOP: Buttons disable during a request, a green check confirms a successful write — the client always knows the data was accepted.
• COACH CONTACT: A built-in P-2-Coach module for urgent or scheduled trainer contact from any screen.
• MATERIALS ARCHIVE: A Biomechanics Archive section with a direct link to the trainer's Google Sheet.

Period: 2026-Q1. Stack: Python, Aiogram, FastAPI, Cloudflare R2, Google Places API, Admin Panel

End-to-end automation of a cleaning business: order intake, cleaner dispatch, customer base, and photo reporting — all in one Telegram bot and web panel.

• ORDER MANAGEMENT: Cleaning request intake and processing, with automatic assignment based on load.
• DISPATCHER BOT: The cleaner gets the task right in Telegram, confirms acceptance, and marks completion.
• PHOTO REPORTING: Site photos are saved to Cloudflare R2 with presigned URLs — available to the manager at any moment without exposing the bucket publicly.
• ADMIN PANEL: A manager-facing web panel: order history, statuses, customer base, address map via Google Places API.
• NOTIFICATION SYSTEM: 2-hour reminders to assignees, automatic reassignment if no response within 15 minutes.
• R2 PRESIGNED URLS: Secured access to photo reports via time-limited links — security without extra complexity.

Period: 2026-Q1. Stack: Python, OpenAI, Bitrix24 API, SMTP, Remotion, Nano Banana, DOCX Template Engine

An AI assistant for the sales team: takes over routine work on proposals, stock checks, knowledge base, and documents. The rep focuses on the client — the agent handles everything else.

• KNOWLEDGE BASE: Consultations from the internal base — products, materials, policies, training, Durlum — the agent answers instantly.
• KP GENERATOR: Automatic proposal generation only as DOCX from a corporate template, with calculations, roll multiples, currency, and FX rate.
• STOCK CHECKER: Real-time stock and assortment checks against a structured database.
• TIME REMINDERS: The agent sets time-based reminders — meetings, deadlines, callbacks.
• BITRIX24 TASKS: Creating CRM tasks in Bitrix24 via webhook straight from chat — no manual entry.
• DOCUMENT PARSER: Parsing PDF / DOCX / XLSX into structured output — from a client file to a finished analysis in seconds.
• EMAIL DISPATCH: Email sending via SMTP on an explicit manager command — the agent composes the message itself.
• VISUAL MATERIALS: Material visualization from 2 photos via Nano Banana — clear for the client.
• INFOGRAPHICS & VIDEO: Infographic and video-presentation generation via Remotion for the marketing team.
• SALES DOCUMENTS: Preparation of working files, tables, calculations, and summaries on request.
• CONTRACTS: Drafting framework contracts from templates with party details auto-filled.

Period: 2026-Q1. Stack: Python, Aiogram FSM, AML Database, Forum API Sync, Escrow Logic

Full automation of OTC deals. Built-in scam check (blacklist), knowledge base, and cross-posting listings to the forum. Turns a chat into a safe marketplace.

• ESCROW PROTOCOL: Guarantor verification and protection against fakes.
• MARKET MAKER: Auto-publish WTB / WTS.
• SECURITY CORE: Counterparty check against the scammer base before the deal.
• KNOWLEDGE BASE: Integrated guides and anti-scam playbooks.

Period: 2025-Q4. Stack: Python, BSC, CREATE2

Buy a token 1 ms before listing.

Period: 2025-Q4. Stack: Telegram Bot API, OpenSea API

Portfolio summary across 6 chains.

Period: 2025-Q4. Stack: Python, Groq AI, WebSocket

Auto-deploy a token from a tweet in <4 sec.

Period: 2025-Q4. Stack: Python, Lighter API

Profit protection every second.

Period: 2025-Q4. Stack: Farcaster SDK, AI Image Gen

Viral NFT mint based on profile and on-chain data.

Period: 2025-Q3. Stack: Python, Web3.py, Linea RPC

Token rescue in <500 ms.

Period: 2025-Q3. Stack: React, TON Connect, FunC

Safe deals inside Telegram.

Period: 2025-Q2. Stack: GPT-3.5, Google Sheets, ZennoPoster

Live tour parsing and AI request handling.

Period: 2025-Q2. Stack: Python, Pandas, OpenSea API, AsyncQueue

Parse 1000+ wallets and an Excel report.

Period: 2025-Q1. Stack: Python, OpenAI GPT, Web Speech API

Service speed up by 40%, order intake automated, staff load optimized.

Period: 2025-Q1. Stack: Python, Aiogram, Whisper, GPT-4

Voice analysis and structured memory, expense tracking, task and schedule routing.

Period: 2025-Q1. Stack: Python, Aiogram, OpenAI API

Auto-laid-out posts and style generation.

Period: 2024-Q4. Stack: ETH, EIP-7702, Smart Contracts

Asset rescue without gas.

Period: 2024-Q4. Stack: CCXT, OKX API

Auto-balancing across exchanges.