Automated Call Analysis System for Managers
1. Project Description
A simple system needs to be developed that will automatically transcribe calls from managers, analyze them for errors, and send ready-made recommendations via a Telegram bot.
2. How the system should work:
1️⃣ The manager sends an audio file to the Telegram bot.
2️⃣ The system transcribes voice to text (Whisper, locally).
3️⃣ Dialogue analysis is performed:
- Searching for key errors in the conversation.
- Performing comparison with 15 sales books (ChromaDB).
- Generating tips and phrases for improvement (ChatGPT API).
4️⃣ The bot sends the manager a report with the transcription, errors, and tips.
5️⃣ The system self-learns – over time it should consider new errors, update recommendations, and improve analysis accuracy.
3. Main Development Requirements
✔ Work on a local PC (Windows 11), without expensive servers and clouds.
✔ Python – preferred programming language.
✔ Use Whisper for transcription (local operation).
✔ Use ChromaDB for searching recommendations in books.
✔ ChatGPT API only for generating tips.
✔ The Telegram bot must work automatically, without manual intervention.
✔ Self-learning system – the ability to add new errors, improve analysis, and personalize recommendations.
4. Main Work Stages
🔹 Stage 1: Development of the system core
- Installation and configuration of Whisper (locally).
- Development of a module for audio to text conversion.
- Configuration of ChromaDB and loading 15 sales books.
- Setting up a request in ChatGPT API for generating recommendations.
🔹 Stage 2: Self-learning of the system
- Adding a feedback mechanism from managers (e.g., accuracy assessment of analysis).
- The ability to update the error database and recommendations manually or automatically.
- Integration of new data into ChromaDB for improved analysis.
🔹 Stage 3: Creating the Telegram bot
- Receiving audio files via Telegram.
- Integration with Whisper and ChromaDB.
- Sending completed reports back to Telegram.
🔹 Stage 4: Testing and debugging
- Checking the accuracy of transcription.
- Correctness of error and recommendation searches.
- Optimizing performance speed.
5. Minimum Requirements for the Performer
✅ Experience with Python.
✅ Experience with Whisper and ChromaDB – desirable but not mandatory.
✅ Ability to develop Telegram bots.
✅ Experience with GPT API – a plus.
6. What is required from the freelancer:
📌 Create a working system and explain how to run it on my laptop.
📌 Do not use complex server solutions – everything should work locally.
📌 Break the work into stages and report on each completed stage.
📌 Answer questions during the process if something is unclear.
7. Budget and Timeline
⏳ Timeline: 10–14 days (to be discussed).
💰 Budget: Reasonable proposals are needed, without inflated prices.
🚀 If you are ready to perform – please provide timelines, price, and examples of similar work.
