feat: solution for '8. Самописный поисковый агент на основе deep agents from scratch'
This commit is contained in:
@@ -1,76 +1,114 @@
|
||||
# Deep Agent from Scratch
|
||||
# Deep Agents from Scratch – LangChain Search Agent
|
||||
|
||||
This repository demonstrates a **Deep Agent** implementation using the **LangChain** library.
|
||||
The agent follows the “Deep Agents from Scratch” template and can answer arbitrary questions by leveraging an LLM (OpenAI GPT‑3.5‑Turbo by default). It also showcases how to integrate a simple tool (`Echo`) and use a Planner/Executor pattern for a more realistic agent workflow.
|
||||
This project demonstrates a **Deep Agent** built from scratch using **LangChain**.
|
||||
The agent can answer user questions by searching the web with DuckDuckGo and
|
||||
providing concise, up‑to‑date responses.
|
||||
|
||||
> **Author**: Artur Kuzakhmetov
|
||||
> **Course**: Deep Agents from Scratch (Lecture: Perplexity, 09.04.2026)
|
||||
> **Deadline**: 31.08.2026
|
||||
|
||||
---
|
||||
|
||||
## Features
|
||||
|
||||
- Implements the **Planner** and **Executor** pattern from the Deep Agents from Scratch template.
|
||||
- Uses LangChain’s `OpenAI`, `Tool`, `PromptTemplate`, and `ConversationBufferMemory`.
|
||||
- Configurable LLM model, temperature, and token limits.
|
||||
- Simple command‑line interface for quick testing.
|
||||
- Environment‑variable based configuration for API keys and model selection.
|
||||
- Demonstrates tool integration (Echo tool) and the full agent template.
|
||||
- **Custom Search Tool** – queries DuckDuckGo’s instant answer API.
|
||||
- **Conversation Memory** – keeps context across turns.
|
||||
- **REACT Agent** – follows the “Reason → Act → Think” pattern.
|
||||
- **CLI** – simple command‑line interface for interactive use.
|
||||
- **Unit Tests** – basic tests for the search tool.
|
||||
|
||||
## Prerequisites
|
||||
|
||||
- Node.js 18+ (or any LTS version)
|
||||
- An OpenAI API key
|
||||
---
|
||||
|
||||
## Setup
|
||||
|
||||
```bash
|
||||
# Clone the repository
|
||||
git clone https://git.brojs.ru/kuzakhmetovartur/8.-samopisnyy-poiskovyy-agent-na-osnove-
|
||||
cd 8.-samopisnyy-poiskovyy-agent-na-osnove-
|
||||
1. **Clone the repository**
|
||||
|
||||
# Install dependencies
|
||||
npm install
|
||||
```
|
||||
```bash
|
||||
git clone https://git.brojs.ru/kuzakhmetovartur/8.-samopisnyy-poiskovyy-agent-na-osnove-.git
|
||||
cd 8.-samopisnyy-poiskovyy-agent-na-osnove-
|
||||
```
|
||||
|
||||
Create a `.env` file in the project root:
|
||||
2. **Create a virtual environment**
|
||||
|
||||
```dotenv
|
||||
OPENAI_API_KEY=your_openai_api_key_here
|
||||
OPENAI_MODEL=gpt-3.5-turbo # optional, defaults to gpt-3.5-turbo
|
||||
```
|
||||
```bash
|
||||
python -m venv .venv
|
||||
source .venv/bin/activate # On Windows: .venv\Scripts\activate
|
||||
```
|
||||
|
||||
> **Tip:** Keep your `.env` file out of version control. Add it to `.gitignore` if you plan to push the repo.
|
||||
3. **Install dependencies**
|
||||
|
||||
```bash
|
||||
pip install -r requirements.txt
|
||||
```
|
||||
|
||||
4. **Set up OpenAI API key**
|
||||
|
||||
Create a `.env` file in the project root:
|
||||
|
||||
```dotenv
|
||||
OPENAI_API_KEY=sk-...
|
||||
```
|
||||
|
||||
Replace `sk-...` with your actual key.
|
||||
|
||||
---
|
||||
|
||||
## Usage
|
||||
|
||||
Run the agent with a question:
|
||||
Run the agent:
|
||||
|
||||
```bash
|
||||
npm start -- "What is the tallest mountain in the world?"
|
||||
python -m src.index
|
||||
```
|
||||
|
||||
Or simply:
|
||||
You will see:
|
||||
|
||||
```
|
||||
Deep Agents from Scratch - LangChain Search Agent
|
||||
Type 'exit' or 'quit' to stop.
|
||||
|
||||
Enter your question:
|
||||
```
|
||||
|
||||
Type a question, e.g.:
|
||||
|
||||
```
|
||||
What is the capital of France?
|
||||
```
|
||||
|
||||
The agent will search the web and return an answer.
|
||||
|
||||
---
|
||||
|
||||
## Running Tests
|
||||
|
||||
```bash
|
||||
node src/index.js "Your question here"
|
||||
python -m unittest discover -s tests
|
||||
```
|
||||
|
||||
The agent will output the answer to the console.
|
||||
---
|
||||
|
||||
## Project Structure
|
||||
|
||||
```
|
||||
├── package.json # Project metadata and dependencies
|
||||
├── src/
|
||||
│ ├── deepAgent.js # Core DeepAgent implementation (Planner/Executor)
|
||||
│ └── index.js # CLI entry point
|
||||
├── src
|
||||
│ └── index.py # Main agent implementation
|
||||
├── tests
|
||||
│ └── test_search_tool.py # Unit tests for the search tool
|
||||
├── requirements.txt # Project dependencies
|
||||
└── README.md # Documentation
|
||||
```
|
||||
|
||||
## Extending the Agent
|
||||
---
|
||||
|
||||
- **Add more sophisticated prompts**: Edit the `Planner` prompt in `deepAgent.js`.
|
||||
- **Integrate additional tools**: Use LangChain’s `Tool` and add them to the `tools` array.
|
||||
- **Switch LLM providers**: Replace `OpenAI` with another LangChain LLM implementation (e.g., `AzureOpenAI`, `Anthropic`).
|
||||
## Contributing
|
||||
|
||||
Feel free to fork the repository, create a feature branch, and submit a pull request.
|
||||
Please ensure tests pass before merging.
|
||||
|
||||
---
|
||||
|
||||
## License
|
||||
|
||||
MIT © 2026
|
||||
|
||||
---
|
||||
MIT License.
|
||||
+4
-5
@@ -1,5 +1,4 @@
|
||||
transformers
|
||||
torch
|
||||
requests
|
||||
click
|
||||
pytest
|
||||
langchain==0.1.0
|
||||
openai==1.3.0
|
||||
python-dotenv==1.0.0
|
||||
requests==2.31.0
|
||||
+139
@@ -0,0 +1,139 @@
|
||||
import os
|
||||
import asyncio
|
||||
from typing import Any
|
||||
|
||||
import requests
|
||||
from dotenv import load_dotenv
|
||||
from langchain.chat_models import ChatOpenAI
|
||||
from langchain.agents import initialize_agent, AgentType
|
||||
from langchain.memory import ConversationBufferMemory
|
||||
from langchain.tools import BaseTool
|
||||
|
||||
|
||||
class DuckDuckGoSearchTool(BaseTool):
|
||||
"""
|
||||
A simple web search tool that queries DuckDuckGo's instant answer API.
|
||||
"""
|
||||
|
||||
name: str = "duckduckgo_search"
|
||||
description: str = (
|
||||
"Use this tool to search the web for up-to-date information. "
|
||||
"Input should be a search query."
|
||||
)
|
||||
|
||||
def _run(self, query: str) -> str:
|
||||
"""
|
||||
Execute the search query and return a concise answer.
|
||||
|
||||
Parameters
|
||||
----------
|
||||
query : str
|
||||
The search query string.
|
||||
|
||||
Returns
|
||||
-------
|
||||
str
|
||||
A short answer extracted from the search results.
|
||||
"""
|
||||
if not query:
|
||||
return "No query provided."
|
||||
|
||||
url = "https://api.duckduckgo.com/"
|
||||
params = {
|
||||
"q": query,
|
||||
"format": "json",
|
||||
"no_html": 1,
|
||||
"skip_disambig": 1,
|
||||
}
|
||||
try:
|
||||
response = requests.get(url, params=params, timeout=10)
|
||||
response.raise_for_status()
|
||||
data = response.json()
|
||||
except Exception as exc:
|
||||
return f"Error during search: {exc}"
|
||||
|
||||
# Prefer abstract text if available
|
||||
abstract = data.get("AbstractText")
|
||||
if abstract:
|
||||
return abstract
|
||||
|
||||
# Fallback to the first related topic
|
||||
topics = data.get("RelatedTopics", [])
|
||||
if topics:
|
||||
first = topics[0]
|
||||
if isinstance(first, dict):
|
||||
return first.get("Text", "No relevant information found.")
|
||||
return "No relevant information found."
|
||||
|
||||
async def _arun(self, query: str) -> str:
|
||||
"""
|
||||
Asynchronous run implementation that delegates to the synchronous _run method.
|
||||
"""
|
||||
loop = asyncio.get_running_loop()
|
||||
return await loop.run_in_executor(None, self._run, query)
|
||||
|
||||
|
||||
def create_agent() -> Any:
|
||||
"""
|
||||
Create and configure the Deep Agent using LangChain.
|
||||
|
||||
Returns
|
||||
-------
|
||||
Any
|
||||
The initialized agent executor.
|
||||
"""
|
||||
# Load environment variables (e.g., OPENAI_API_KEY)
|
||||
load_dotenv()
|
||||
|
||||
# Initialize the LLM
|
||||
llm = ChatOpenAI(temperature=0)
|
||||
|
||||
# Memory to keep conversation context
|
||||
memory = ConversationBufferMemory(memory_key="chat_history")
|
||||
|
||||
# Instantiate the custom search tool
|
||||
search_tool = DuckDuckGoSearchTool()
|
||||
|
||||
# Initialize the agent with the REACT description template
|
||||
agent = initialize_agent(
|
||||
tools=[search_tool],
|
||||
llm=llm,
|
||||
agent=AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
|
||||
memory=memory,
|
||||
verbose=True,
|
||||
)
|
||||
return agent
|
||||
|
||||
|
||||
def main() -> None:
|
||||
"""
|
||||
Simple CLI to interact with the Deep Agent.
|
||||
"""
|
||||
agent = create_agent()
|
||||
print("Deep Agents from Scratch - LangChain Search Agent")
|
||||
print("Type 'exit' or 'quit' to stop.\n")
|
||||
|
||||
while True:
|
||||
try:
|
||||
query = input("Enter your question: ").strip()
|
||||
except (EOFError, KeyboardInterrupt):
|
||||
print("\nExiting.")
|
||||
break
|
||||
|
||||
if query.lower() in {"exit", "quit"}:
|
||||
print("Goodbye!")
|
||||
break
|
||||
|
||||
if not query:
|
||||
print("Please enter a non-empty query.")
|
||||
continue
|
||||
|
||||
try:
|
||||
result = agent.run(query)
|
||||
print("\nAnswer:\n", result)
|
||||
except Exception as exc:
|
||||
print(f"Error: {exc}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,21 @@
|
||||
import unittest
|
||||
from src.index import DuckDuckGoSearchTool
|
||||
|
||||
|
||||
class TestSearchTool(unittest.TestCase):
|
||||
def setUp(self):
|
||||
self.tool = DuckDuckGoSearchTool()
|
||||
|
||||
def test_run_returns_string(self):
|
||||
result = self.tool.run("Python programming language")
|
||||
self.assertIsInstance(result, str)
|
||||
self.assertTrue(len(result) > 0)
|
||||
|
||||
def test_run_handles_empty_query(self):
|
||||
result = self.tool.run("")
|
||||
self.assertIsInstance(result, str)
|
||||
self.assertTrue(len(result) > 0)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
Reference in New Issue
Block a user