CASE STUDY · SEARCH · TELEGRAM

TechPlanner: engineering standards search without manual browsing

A production-ready search assistant that extracts heterogeneous technical documents, retrieves relevant passages in Russian and English, and controls access through a Telegram subscription workflow.

Build something similar ↗View architecture

TechPlanner administration interface — Selected project material · anonymized where required

4Docker services

2search languages

3free searches

RU / ENdomain analyzers

THE CHALLENGE

What had to remain reliable.

The same engineering term can appear in filenames, folder paths, and the body of dozens of PDF, DOCX, or spreadsheet files. A useful result must highlight the right context while tolerating spelling variation and mixed Russian and English terminology.

THE SOLUTION

How the system was designed.

Document extraction is isolated from retrieval. Apache Tika normalizes source files, OpenSearch applies dedicated RU/EN analyzers and domain synonyms, fields receive explicit weights, fuzzy matching handles typos, and long highlighted fragments explain why each result was selected.

DELIVERED

The working scope.

Ingestion for PDF, DOCX, XLSX, and additional formats with extracted text and searchable metadata.
Russian and English analyzers using normalization, stop words, stemming, and a synonym graph for technical terms.
Weighted search across filename, body, and path with AUTO fuzziness and highlighted context passages.
Telegram flows, WebApp document preview, search logs, limits, and PostgreSQL access state.
Monthly YooKassa billing with idempotent payment creation, status verification, and success or cancellation webhooks.

ARCHITECTURE

The main system boundaries and data flow.

[ Telegram Bot / WebApp ] ---> [ Application Service ] ---> [ PostgreSQL ]
              |                         |
              |                         +----> [ YooKassa webhooks ]
              |
              +----> [ Apache Tika ] ---> [ OpenSearch RU/EN index ]
                                               |
                                    [ weighted search + highlights ]

OUTCOMES

What the architecture enables.

Faster access to a precise standard fragment

Search behavior tuned for bilingual engineering terminology

Payment and access state isolated from retrieval

Containerized deployment with health checks and persistent volumes

STACKPythonAiogram 3OpenSearchApache TikaPostgreSQLYooKassa APIDocker Compose

YOUR PROJECT CAN START WITH A SHORT TECHNICAL REVIEW

Show me the workflow, integration points, and current bottleneck.

Discuss the project ↗