GigaChat api: README.md

2026-05-27 11:18:34 +00:00
parent 76b3ced312
commit 61edc902c8
1 changed files with 148 additions and 0 deletions
@@ -0,0 +1,148 @@
+# GigaChat API – Function Calling Demo
+
+A lightweight Python demo that shows how to interact with the **Sberbank GigaChat** REST API using the *Function Calling* feature.  
+The script sends a user prompt, receives a response from the model, detects if it requests a function call, executes the local function, and feeds the result back into the conversation.
+
+> **⚠️ Important** – Replace `YOUR_GIGACHAT_API_KEY` in `client.py` with your real GigaChat API key before running.
+
+---
+
+## Table of Contents
+
+- [What It Does](#what-it-does)
+- [Prerequisites](#prerequisites)
+- [Installation](#installation)
+- [Running the Demo](#running-the-demo)
+  - [Single Prompt Example](#single-prompt-example)
+  - [Interactive Session](#interactive-session)
+- [Example Usage](#example-usage)
+- [Extending with More Functions](#extending-with-more-functions)
+
+---
+
+## What It Does
+
+1. **Sends a user prompt** to GigaChat via the `/chat` endpoint.
+2. **Parses the model’s response** for a `function_call`.
+3. If a function call is detected:
+   - Extracts the function name and arguments.
+   - Executes the corresponding local Python function (currently only `calculate_sum`).
+   - Sends the function result back to GigaChat as part of the conversation.
+4. Prints the final assistant message.
+
+---
+
+## Prerequisites
+
+| Component | Minimum Version |
+|-----------|-----------------|
+| Python    | 3.11+           |
+| pip       | –               |
+
+The demo uses only standard libraries (`json`, `time`) and the third‑party package:
+
+```bash
+pip install requests
+```
+
+No external services (Ollama, Qdrant, etc.) are required.
+
+---
+
+## Installation
+
+1. **Clone the repository** (or copy the files into a folder):
+
+   ```bash
+   git clone https://github.com/your-username/gigachat-function-calling.git
+   cd gigachat-function-calling
+   ```
+
+2. **Create a virtual environment** (recommended):
+
+   ```bash
+   python -m venv .venv
+   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
+   ```
+
+3. **Install dependencies**:
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+> `requirements.txt` contains only `requests`.
+
+---
+
+## Running the Demo
+
+### Single Prompt Example
+
+```bash
+python client.py "What is 12 plus 30?"
+```
+
+The script will:
+- Send `"What is 12 plus 30?"` to GigaChat.
+- Detect that the model wants to call `calculate_sum(a=12, b=30)`.
+- Execute the function locally and return the result (`42`) back to the model.
+- Print the final assistant response.
+
+### Interactive Session
+
+If you run the script without arguments, it will start an interactive prompt:
+
+```bash
+python client.py
+```
+
+You can then type any question or request. The session continues until you exit (Ctrl‑C).
+
+---
+
+## Example Usage
+
+**Command**
+
+```bash
+python client.py "Add 7 and 15."
+```
+
+**Output**
+
+```
+User: Add 7 and 15.
+Assistant: Sure! Let me calculate that for you...
+Function call detected: calculate_sum(a=7, b=15)
+Result of calculate_sum(7, 15): 22
+Assistant: The sum of 7 and 15 is 22. Is there anything else I can help with?
+```
+
+---
+
+## Extending with More Functions
+
+1. **Define a new function** in `client.py`:
+
+   ```python
+   def get_current_time() -> str:
+       return time.strftime("%Y-%m-%d %H:%M:%S")
+   ```
+
+2. **Add it to the `FUNCTION_MAP`:**
+
+   ```python
+   FUNCTION_MAP = {
+       "calculate_sum": calculate_sum,
+       "get_current_time": get_current_time,
+   }
+   ```
+
+3. **Update the model’s function schema** (if you control the prompt or system message) so GigaChat knows about the new function.
+
+The script will automatically detect and execute any function present in `FUNCTION_MAP`.
+
+---
+
+Happy coding! 🚀