
Closed
Posted
Paid on delivery
Interní AI asistent nad daty z Pohoda ERP (SQL Server) Poptávám vývoj/dokončení interní webové aplikace, která umožní zaměstnancům firmy klást v přirozeném jazyce (čeština) dotazy nad daty z účetního systému Pohoda a dostávat odpovědi, reporty a exporty. Technický kontext: Data: Pohoda (Stormware) na MS SQL Server, read-only přístup (db_datareader) Část dat (obchodní zástupce u faktur) dostupná jen přes Pohoda mServer (XML API), ne přímo v SQL AI vrstva: Anthropic Claude API (Messages API, ne ChatGPT/jiný model) s tool-use smyčkou — model si sám generuje a spouští SQL dotazy Backend: Python (Flask/podobné), self-hosted na Windows serveru jako služba (NSSM) Databáze appky: SQLite (uživatelé, konverzace, historie, logy) Funkční požadavky: Webové chatovací rozhraní, multi-user, s přihlašováním Role-based přístup (min. 3 úrovně: admin / vedení / běžný uživatel) — každá role vidí jiný rozsah dat/nástrojů AI agent generuje a spouští read-only SQL dotazy nad Pohoda schématem (desítky tabulek, nutná znalostní báze schématu/gotchas) Kombinace SQL + mServer dat v jedné odpovědi Export výsledků do formátovaného Excelu Sdílený kontext/poznámky editovatelné vedením, viditelné pro appku při každém dotazu Naplánované opakující se dotazy (denní/týdenní reporty bez zásahu uživatele) Admin rozhraní: správa uživatelů, log všech dotazů, přehled chyb, přehled spotřeby/nákladů API Optimalizace nákladů na API (prompt caching apod.) Odhad rozsahu dat: cca 11 středisek, řádově statisíce dokladů ročně, víceleté historické archivy (samostatné databáze po ročnících). Prosím o orientační cenu za vývoj (rozpad na fáze/moduly uvítám) a odhad časové náročnosti. Zkušenost s Pohoda/Stormware API nebo podobnými ERP integracemi výhodou, ne podmínkou.
Project ID: 40667852
101 proposals
Remote project
Active 3 days ago
Set your budget and timeframe
Get paid for your work
Outline your proposal
It's free to sign up and bid on jobs
101 freelancers are bidding on average €63,679 EUR for this job

Hello, I'm Elias, a Miami-based senior software engineer with 20 years of experience. I specialize in building web applications and AI integrations. I've carefully reviewed your project regarding the internal AI assistant for Pohoda ERP. It seems your main goal is to enhance data accessibility and decision-making through an intelligent interface that interacts with your ERP system. I've developed similar systems, focusing on secure architecture, responsive user interfaces, and scalability. My approach would involve: 1) Understanding the business workflows and technical requirements. 2) Designing the user experience and AI architecture. 3) Developing core features and integrating with the SQL Server. 4) Thoroughly testing everything before launch. 5) Deploying and providing post-launch support. Could you please clarify the following questions to help me better understand the project? 1) What specific functionalities do you envision for the AI assistant? 2) Are there particular data integrations or external APIs to consider? 3) How will user permissions and roles be managed within the assistant? I've built similar systems before, and while most projects were completed under NDA, I’d be happy to discuss the architecture and features. Additionally, I can ensure a maintainable codebase for future enhancements.
€75,000 EUR in 35 days
8.1
8.1

Dobrý den, specialisto na integraci AI + ERP {{{ JIŽ JSEM VYTVOŘIL PODOBNÉ A MOHU VÁM HO UKÁZAT }}} Pečlivě jsem si prošel vaše požadavky na interního česky mluvícího asistenta AI, propojeného s Pohoda ERP a MS SQL Serverem. Kombinace přístupu SQL pouze pro čtení, XML API Pohoda mServer, používání nástrojů Claude, oprávnění založených na rolích a plánovaného reportingu vyžaduje pečlivě řízenou architekturu, nikoli jednoduchého chatbota. Mám více než 11 let zkušeností s vývojem full-stacků a AI a mohu řešení vytvořit pomocí Pythonu/Flasku, přístupu MS SQL pouze pro čtení, SQLite pro aplikační data a Anthropic Claude Messages API s řízeným pracovním postupem používání nástrojů. Agent AI bude schopen rozumět otázkám v přirozeném jazyce, generovat validované dotazy SQL pouze pro čtení, bezpečně je spouštět v souladu se schématem Pohoda a v případě potřeby kombinovat výsledky SQL s daty mServer. Také implementuji znalostní databázi schémat, která bude zahrnovat vztahy mezi tabulkami, obchodní pravidla a běžné úskalí dat ERP, abych zlepšil přesnost dotazů. Aplikace bude zahrnovat ověřování více uživatelů, role administrátora/správce/uživatele, sdílený kontext správy, formátované exporty do Excelu, plánované denní/týdenní reporty, historii konverzací, protokoly auditu, monitorování chyb a sledování využití/nákladů API. POSKYTUJI 2 ROKY BEZPLATNÉ PRŮBĚŽNÉ PODPORY A KOMPLETNÍ ZDROJOVÝ KÓD. Děkuji, Christina
€50,000 EUR in 30 days
7.2
7.2

I have strong experience building AI-powered web applications, Python backends, SQL-based data agents, role-based access, and API integrations. I can complete the solution with Claude Messages API and tool-use, ensuring the AI works safely with Pohoda SQL data while also combining mServer/XML data where required. For the SQL agent, I would implement a strict read-only execution layer with query validation, restricted database credentials, allowed-table/tool controls, timeouts, logging, and role-based permissions before any query reaches SQL Server. I can also handle Excel exports, scheduled reports, shared context, admin monitoring, API-cost optimization, and multi-year Pohoda databases. Is the current Flask application already partially developed, or would you like the complete implementation from scratch? Also, do you already have the Claude API/tool-use workflow and Pohoda mServer XML integration in place, or should I develop those components as well?
€50,000 EUR in 7 days
6.4
6.4

Hello!, This is Temo. I see you're looking for an internal AI assistant for Pohoda ERP using data from SQL Server. It’s crucial to build a solution that integrates well with your existing workflow while ensuring efficient data handling. With around 15 years of experience in Python and SQL, I understand the importance of a robust architecture for your AI chatbot. I would focus on using Flask for the web application, allowing for a clean integration with the database and AI functionalities. To kick things off, I'd set up a prototype to prototype the core features and gather user feedback early. Could you please clarify the following questions to help me better understand the project? Q1: What specific functionalities do you want the AI assistant to have? Q2: Are there any particular AI models or integrations you have in mind? Q3: How do you envision the user interface for the web application? I’m committed to delivering a solution that meets your needs.
€75,000 EUR in 459 days
6.6
6.6

Hello, I can build this as a secure internal AI assistant around Pohoda ERP / SQL Server + Claude Messages API, with role-based access and fully read-only ERP querying. Recommended architecture: Python/Flask backend on Windows + NSSM MS SQL Server read-only connector for Pohoda data Pohoda mServer XML API for fields not exposed in SQL Claude tool-use loop for schema-aware SQL generation/execution SQLite for users, conversations, logs, schedules, and API usage Excel export with formatted reports Admin panel for users, errors, audit logs, and cost monitoring Scheduled daily/weekly reports Shared management notes/context injected into every query Prompt caching and context optimization to control API cost I’d implement this in phases: core chat + auth, SQL/mServer integration, RBAC, exports, scheduling/admin, then hardening/testing. Estimated MVP: 4–6 weeks, depending mainly on Pohoda schema complexity and the existing codebase. I can provide a fixed-price breakdown after reviewing the current application and sample database schema.
€75,000 EUR in 7 days
6.2
6.2

Hello!, This is James from Hollywood... I understand the real challenge here is not “just building a chatbot” but creating a reliable internal AI assistant that can safely read Pohoda ERP data from SQL Server, answer business questions accurately, and fit into your workflow without becoming a fragile tool. My approach would be practical and phased: 1. Review your current internal app/data model and define the exact use cases 2. Build a secure data layer on top of Pohoda ERP SQL Server 3. Add an AI/RAG layer with strict context control so answers come from your data, not guesses 4. Create a clean internal web UI for chat, search, filters, and role-based access 5. Test accuracy, permissions, and edge cases before handover I’ve built production systems where the difference between a demo and a real business tool was careful architecture, validation, and keeping the AI grounded in the source system. That’s exactly how I’d handle this. Relevant work: - internal AI assistant for an operations dashboard - ERP-connected reporting portal for a distribution company - AI support tool for structured database lookup - Flask-based admin systems with SQL-backed workflows A few quick questions: 1. Do you already have an internal web app, or should I build the assistant UI from scratch? 2. Which Pohoda data should the assistant read first, and are there restricted tables/modules? Regards, James
€75,000 EUR in 40 days
5.8
5.8

Hello, We can develop and complete your internal Czech-language AI assistant integrating **Pohoda ERP, MS SQL Server, Pohoda mServer XML API, and Anthropic Claude API**. Our team has strong experience with **Python, SQL Server, REST/XML APIs, AI tool-use agents, role-based web applications, Excel exports, scheduled reporting, and self-hosted Windows deployments**. We can cover: * Secure multi-user web chat with role-based access * Claude Messages API + tool-use loop for controlled read-only SQL generation/execution * Pohoda SQL schema knowledge base and query safeguards * Integration with Pohoda mServer for data unavailable directly in SQL * Combined SQL + mServer answers * Excel report/export generation * Management notes/shared context * Scheduled daily/weekly reports * Admin panel with users, query/audit logs, errors and API cost monitoring * Prompt caching and other API cost optimizations * SQLite application database and Windows service deployment **Estimated timeline:** 6–10 weeks depending on the current state of the existing application and API access. **Indicative budget:** €4,000–€8,000 for the complete MVP/production-ready scope, which we can refine after reviewing the existing codebase, Pohoda schema, mServer documentation and exact role permissions. We can start with a short technical assessment and then provide a detailed phase-wise estimate. Best regards, Anju Logical Soft Tech Pvt. Ltd.
€75,000 EUR in 60 days
5.8
5.8

You need an internal AI assistant for Pohoda ERP data, and I will build a web chat interface connecting to your SQL Server data, with a Python backend and using the Anthropic Claude API for natural language queries in Czech. I will use Python with Flask for the backend framework, connecting directly to your MS SQL Server via a read-only db_datareader role to fetch data, and also integrate with the Pohoda mServer XML API to access information not directly in the SQL database, so the AI can answer questions about sales representatives on invoices. The AI layer will use Anthropic Claude's Messages API, implementing a tool-use loop where the model crafts and executes SQL queries against your database, also I will set up SQLite for user management, conversation history, and logging. I will not build a separate front-end application that communicates with the backend via REST APIs as that adds unnecessary complexity and delay; the web chat interface will be built directly into the Flask application, keeping the deployment simpler. What is the specific structure of the XML data provided by the Pohoda mServer for sales representative information, and what are the key fields you'd want the AI to be able to query from that data? For what it is worth, every job I have taken on Freelancer has gone out on time and on budget, 100% on both. I need the SQL Server connection details to start.
€83,793 EUR in 21 days
5.3
5.3

Your AI agent will fail in production if it doesn't handle Pohoda's multi-year archive databases correctly — Claude will generate queries against the wrong schema when users ask for historical data spanning multiple fiscal years. Quick questions - how do you plan to handle mServer XML rate limits when the agent needs obchodní zástupce data for 50+ invoices in one query? And what's your fallback strategy if Claude generates a query that locks tables or times out on 500K+ rows? Here is the architectural approach: - ANTHROPIC CLAUDE TOOL-USE: Build a constrained SQL executor that validates generated queries against a whitelist of allowed tables/columns per role before execution, preventing privilege escalation and runaway queries. - POHODA SCHEMA KNOWLEDGE BASE: Inject table relationships and business logic gotchas (e.g., SklDoklad vs. FaktVyd joins, archive DB naming patterns) into Claude's system prompt using prompt caching to cut API costs by 70%+. - MSERVER + SQL FUSION: Implement async request batching for mServer XML calls so the agent fetches obchodní zástupce data in parallel, then joins it with SQL results in-memory before formatting the Excel export. I've built similar agentic systems for 2 ERP platforms where SQL generation accuracy was mission-critical and API costs had to stay under $200/month at scale. Let's schedule a 20-minute technical call to walk through your Pohoda schema and nail down the role-access matrix before I scope phases and pricing.
€67,500 EUR in 30 days
5.9
5.9

Building on a strong background in web and mobile development, AI automation, and digital operations, I can offer you a comprehensive solution for your internal AI assistant project. With proficiency in Python and experience working with SQL Server, I am well-versed in the technological infrastructure required for this task. I also have a solid understanding of Pohoda/Stormware API integration which is advantageous for this project. For the functionalities you've outlined, such as a multi-user chat interface with different levels of data access and role-based authorization, my AI Chatbot Development skills will enable me to build an app tailored to your specific needs. Additionally, my knowledge in SQLAlchemy will facilitate seamless interaction between SQLite-based database of the app and the MS SQL Server data housed in Pohoda. Overall, I am confident that my technical skills coupled with my ability to comprehend and relate to your business operations will ensure an efficient delivery of the project that aligns perfectly with your objectives. Let's discuss further to formulate a phase-wise approach and an accurate timeline-corresponding estimate for this endeavor.
€50,000 EUR in 70 days
5.4
5.4

With my extensive experience as a Full-Stack Developer, I believe I'm the perfect fit for developing your Internal AI assistant for Pohoda ERP. My skills in Python, Flask, and SQL Server align perfectly with your project requirements. I have a strong command over handling large datasets like that of Pohoda's and possess adequate knowledge regarding SQL schemes and gotchas - a crucial aspect for successfully implementing your project functionalities. Apart from the technical aspects, I can also provide you with a clean and intuitive web interface, multi-user functionality with role-based access, export features to Excel as expected, and an admin interface to manage all users effectively as per their access. I also bring to the table my adeptness at incorporating AI & ML. So, your requirement of using Anthropic Claude API will be very comfortable to handle. Additionally, I'm committed to delivering high-quality solutions within agreed upon timelines. My previous projects involve SaaS platforms development and custom ERP systems integration, making me well-versed with the enhancements that'll positively impact your project's efficiency.
€50,000 EUR in 1 day
5.0
5.0

As a team with over 17 years of comprehensive development experience, we specialize in CRITICAL projects involving APIs, Databases and Web Applications using technologies including PostgreSQL, Python, Software Architecture and SQL. We also boast extensive knowledge in Asp.net, including the SQL Server DATABASE PERFORMANCE optimization in your project's context. With an expert team like ours, your project's native Czech context would be a breeze for us to navigate. Our extensive experience also means that you have access to tried-and-true methods and solutions that can help us accommodate all the unique functionalities you require, such as role-based access and combining Pohoda and mServer data within a single response. Moreover, our adaptable team is well-versed in producing cost-effective outcomes whilst optimizing API resource usage. Through regular and transparent communication throughout the development lifecycle, we ensure on-time delivery and intelligent brainstorming around issues like prompt caching (to optimize API costs) if it arises. Please take note that our prices are competitive while delivering uncompromising quality. Let us prove why we deserve this amazing opportunity to tackle such an intriguing project as yours!
€75,000 EUR in 7 days
4.6
4.6

Hi there, Thank you for sharing the detailed requirements for your internal AI assistant project for Pohoda ERP. I’m excited by the opportunity to collaborate with you on developing a robust, AI-driven web application that empowers your team to interact with business data in natural Czech language. With extensive experience in Python backend development (Flask), SQL Server integrations, and AI chatbot solutions, I have delivered similar projects involving complex ERP systems, role-based access control, and multi-source data aggregation. My background also includes optimizing AI API usage, implementing secure authentication, and building user-friendly admin dashboards for ongoing management and transparency. Your vision of leveraging Anthropic Claude API with a tool-use loop for generating and executing SQL queries aligns well with my expertise in AI integration and prompt engineering. I understand the importance of combining data from both direct SQL and Pohoda mServer’s XML API, ensuring seamless, accurate responses and exports—especially when handling diverse data schemas and large volumes. For this project, I propose a modular development approach: starting with secure multi-user authentication and role management, followed by the AI query engine, integration with both SQL Server and mServer, and finally, building out reporting/export features and the admin interface. Throughout, I’ll ensure robust logging, error handling, and cost optimization strategies (such as prompt/result caching). I’m confident I can deliver a scalable, maintainable solution tailored to your needs, and I look forward to discussing the technical details and next steps with you. Best regards, DemiVision, LLC
€50,000 EUR in 90 days
4.6
4.6

I’ve built and deployed production-grade AI agents with tool-use loops over ERPs—direct SQL access, external APIs, and role-based filtering—so this is routine for me. I’ll start with schema mapping and mServer API reverse-engineering to build a static knowledge base, then implement Flask with SQLite auth and role guards; Anthropic Messages API integration will handle dynamic SQL generation, caching prompts to cut costs; scheduled tasks via APScheduler for reports, Excel exports via openpyxl, shared context notes in SQLite with row-level visibility; admin dashboard tracks queries, errors, and API burn rate. Windows service packaging via NSSM, hardened connection strings, and minimal downtime deployment. I can start immediately. Thanks, Andrii.
€75,000 EUR in 30 days
4.8
4.8

Hello, Rozumím, že potřebujete interní AI asistenta nad Pohoda ERP, který bezpečně kombinuje read only SQL data s Pohoda mServer API a umožní zaměstnancům klást dotazy v češtině. Důležitá je také role based access, Excel exporty, plánované reporty, audit logy a kontrola nákladů Claude API. Navrhl bych modulární Python backend s bezpečným tool use přístupem, oddělenými oprávněními a znalostní vrstvou nad Pohoda schématem. SQL dotazy bych omezil na read only přístup a výsledky z SQL a mServeru sjednotil před vytvořením odpovědi. Postupně bych řešení rozdělil na backend, AI agent, Pohoda integraci, frontend, reporting a administraci. Rád s vámi projdu současný stav aplikace a připravím konkrétní rozpad jednotlivých fází, rozsahu a náročnosti. Best Regards, Fizza
€50,000 EUR in 5 days
4.8
4.8

Dobrý den, váš projekt chápu jako interní AI datovou vrstvu nad Pohoda ERP, nikoli jako běžný chatbot. Nejdůležitější částí bude spolehlivě propojit Claude tool-use, SQL Server a mServer tak, aby uživatel mohl položit otázku česky a systém vrátil správnou, dohledatelnou odpověď bez rizika změny ERP dat. Mám zkušenosti s Python backendy, SQL, LLM agenty, tool/function callingem, strukturovanými výstupy, API integracemi a systémy s role-based přístupem. Stávající Flask/Python směr bych zachoval, pokud audit neodhalí důvod ke změně. Claude bych nepouštěl přímo nad databázi bez kontrolní vrstvy. SQL nástroje bych omezil na read-only operace, povolená schémata/tabulky, limity výsledků a timeouty. Součástí řešení by byla znalostní vrstva popisující Pohoda schéma, vazby, význam polí a známé výjimky. Vývoj bych rozdělil na audit Pohoda dat a mServeru, AI/SQL engine, web/RBAC, kombinaci SQL+mServer, reporty a plánované úlohy, administraci/audit a nakonec bezpečnostní testování a produkční nasazení na Windows Server. Počítám s předáním kompletního zdrojového kódu, dokumentace, deployment postupu a testovací sady pro ověřování přesnosti AI odpovědí.
€68,000 EUR in 102 days
4.4
4.4

Dear Hiring Manager, I'm Wasif Muneer M., a seasoned Full Stack Web Developer with a proven track record of delivering high-quality projects on Freelancer.com. With over 11 years of experience, I've worked on numerous projects, including enterprise web platforms, security monitoring systems, and learning management systems. I'm excited to take on your project, Vývoj Interního AI Asistenta pro Pohoda ERP, and help you create an innovative internal AI assistant that enables employees to query data from the Pohoda ERP system in their native language (Czech). My expertise in Python, SQL, and Flask will enable me to develop a self-hosted web application that meets your functional requirements. I'd like to highlight my experience with Pohoda/Stormware API, which will be essential in integrating the AI assistant with the ERP system. My familiarity with Anthropic Claude API and its Messages API will also ensure a seamless integration with the AI layer. Regarding the project requirements, I'd like to clarify a few points: 1. Will the AI agent be able to handle complex queries or will they be limited to simple read-only SQL queries? 2. Are there any specific performance requirements for the web chat interface, such as latency or concurrency? I estimate the project scope to be approximately 10,000 hours, broken down into several phases and modules. The estimated budget for the project is $75,000, which includes development, testing, and deployment costs. I'd be happy to provide a detailed project plan and timeline upon request. Please feel free to contact me to discuss the project further. Best regards, Wasif Muneer M. Freelancer.com ID: [insert ID]
€75,000 EUR in 7 days
4.5
4.5

Roční archivy Pohody jako oddělené databáze jsou největší past tohoto zadání, agent musí vědět, kterou databázi za který rok oslovit, jinak vrátí prázdné výsledky bez chyby. Postavím Flask backend s Claude Messages API v tool-use smyčce nad read-only SQL + mServer XML, se znalostní bází schématu Pohody, rolemi v SQLite a exportem do formátovaného Excelu. Prompt caching schématu a system promptu srazí náklady nejvíc, u desítek tabulek se to pozná hned. Orientační cenu i odhad času upřesním v chatu, jakmile probereme rozsah. 1) Kolik ročníkových databází je aktivních a chcete dotazy napříč roky, nebo vždy v rámci jednoho? 2) mServer běží už teď jako služba, nebo ho mám nasadit a nakonfigurovat také? 3) Má Claude spouštět naplánované reporty pod servisním účtem, nebo pod identitou zadavatele plánu? Happy to talk details in chat. Shayan
€55,000 EUR in 20 days
4.1
4.1

Dobrý den, Dokážu vytvořit a dokončit tohoto interního AI asistenta pro Pohoda ERP s využitím vaší stávající architektury a Claude Messages API s využitím nástrojů, namísto nahrazování stacku. Osnovy: Český chat v přirozeném jazyce s přihlášením pro více uživatelů Přístup na základě rolí: Administrátor / Správa / Běžný uživatel SQL agent pouze pro čtení s podporou Claude pro Pohoda SQL Server Integrace Pohoda mServer/XML API Kombinované odpovědi SQL + mServer Generování excelových reportů/exportů Poznámky pro správu/sdílený kontext Plánované denní/týdenní reporty Administrační dashboard s uživateli, protokoly dotazů, chybami a náklady na API Okamžité ukládání do mezipaměti a optimalizace dotazů pro kontrolu nákladů Claude Bezpečné nasazení Windows serveru s Flask + SQLite Doporučil bych fázovaný vývoj: Audit → AI/SQL Agent → Integrace ERP/mServer → Reporty/Exporty → Plánování/Administrace → Testování a produkce. Mohu poskytnout podrobný odhad ceny po kontrole stávající kódové základny, schématu databáze a aktuální implementace. Děkuji
€50,000 EUR in 35 days
4.2
4.2

==== Hi - Truong here ==== "INTERNÍ AI ASISTENT NAD POHODA ERP" — potřebujete bezpečný český AI chat, který dokáže spojit read-only SQL data s Pohoda mServer daty a vrátit použitelné reporty. Postavil bych backend kolem Flasku s oddělenou tool-use vrstvou pro Claude, která povolí pouze bezpečné SELECT dotazy a kontroluje SQL před spuštěním. Pro kombinaci SQL + mServer bych výsledky sjednotil do jednoho datového modelu; role, audit logy, plánované reporty a Excel export bych řešil jako samostatné moduly. Je už připravený konkrétní Pohoda SQL schema dokument a seznam funkcí, které musí jednotlivé role v první verzi používat? Looking forward to working with you.
€60,000 EUR in 100 days
4.1
4.1

Brno, Czech Republic
Member since Aug 24, 2026
₹12500-37500 INR
$1500-3000 USD
$10-30 AUD
₹12500-37500 INR
$10-30 USD
₹75000-150000 INR
$250-750 USD
$10-30 CAD
₹37500-75000 INR
₹5000-8000 INR
$30-250 AUD
₹250000-500000 INR
$10-30 USD
$2-8 CAD / hour
₹750-1250 INR / hour
₹12500-37500 INR
₹12500-37500 INR
$30-250 USD
$750-1500 AUD
$15-25 USD / hour