Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/sl-mar/quantcoder-cli

AI-powered CLI tool: Transform trading research papers into QuantConnect algorithms using GPT-4
https://github.com/sl-mar/quantcoder-cli

llm openai pair-programming quantconnect trading trading-algorithms trading-strategies

Last synced: 7 days ago
JSON representation

AI-powered CLI tool: Transform trading research papers into QuantConnect algorithms using GPT-4

Awesome Lists containing this project

README 

          
# QuantCoder 2.0.0



[![Version](https://img.shields.io/badge/version-2.0.0-green)](https://github.com/SL-Mar/quantcoder-cli)

[![Python](https://img.shields.io/badge/python-3.10+-blue)](https://python.org)

[![License](https://img.shields.io/badge/license-Apache%202.0-blue)](LICENSE)



> **AI-powered CLI for generating QuantConnect trading algorithms from research articles**



> **Note**

> This version (v2.0.0) has not been systematically tested yet.

> It represents a complete architectural rewrite from the legacy v1.x codebase.

> Use with caution and report any issues.



Features: Multi-agent system, AlphaEvolve-inspired evolution, autonomous learning, MCP integration.



---



QuantCoder is a command-line tool that allows users to generate QuantConnect trading algorithms from research articles using natural language processing and large language models (LLMs). It was initiated in November 2023 and based on a cognitive architecture inspired by the article ["Dual Agent Chatbots and Expert Systems Design"](https://towardsdev.com/dual-agent-chatbots-and-expert-systems-design-25e2cba434e9)



The initial version successfully coded a blended momentum and mean-reversion strategy as described in ["Outperforming the Market (1000% in 10 years)"](https://medium.com/coinmonks/how-to-outperform-the-market-fe151b944c77?sk=7066045abe12d5cf88c7edc80ec2679c), which received over 10,000 impressions on LinkedIn.



## 🌟 Version 2.0 - Complete Refactoring



**Refactored in December 2025** - Inspired by [Mistral's Vibe CLI](https://github.com/mistralai/mistral-vibe) architecture.



### New Features:

- 🤖 **Interactive Chat Interface** with conversational AI

- 🛠️ **Tool-Based Architecture** for modularity and extensibility

- ⚙️ **Configuration System** with TOML support

- 🎨 **Modern Terminal UI** with Rich library

- 📝 **Programmable Mode** via `--prompt` flag

- 💾 **Persistent Context** and conversation history



📖 **[Architecture](docs/AGENTIC_WORKFLOW.md)** | **[Autonomous Mode](docs/AUTONOMOUS_MODE.md)** | **[Changelog](CHANGELOG.md)**



---



## 🚀 Installation (v2.0)



> ✅ Requires **Python 3.10 or later**



### Quick Start



```bash

# Clone the repository

git clone https://github.com/SL-Mar/quantcoder-cli.git

cd quantcoder-cli



# Create and activate virtual environment

python -m venv .venv

source .venv/bin/activate  # On Windows: .venv\Scripts\activate



# Install the package

pip install -e .



# Download SpaCy model

python -m spacy download en_core_web_sm

```



### First Run



```bash

# Launch interactive mode

quantcoder



# Or use the short alias

qc

```



On first run, you'll be prompted for your OpenAI API key.



---



## 💡 Usage (v2.0)



QuantCoder offers multiple modes of operation to suit different workflows.



### Interactive Mode



Launch the interactive chat interface for a conversational experience:



```bash

# Start interactive mode

quantcoder



# Or use the short alias

qc

```



In interactive mode, you can use natural language or direct commands:



```bash

quantcoder> search "momentum trading strategies"

quantcoder> download 1

quantcoder> summarize 1

quantcoder> generate 1

```



### Programmatic Mode



Run commands non-interactively using the `--prompt` flag:



```bash

quantcoder --prompt "Find articles about mean reversion"

```



---



## 📖 CLI Commands Reference



### Core Commands



#### `search` - Search for Academic Articles



Search for research articles on CrossRef using keywords.



```bash

quantcoder search "algorithmic trading" --num 5

```



**Arguments:**

- `query` - Search query string (required)



**Options:**

- `--num` - Number of results to return (default: 5)



**Example:**

```bash

quantcoder search "momentum trading strategies" --num 10

```



---



#### `download` - Download Article PDF



Download a research article PDF by its ID from the search results.



```bash

quantcoder download 1

```



**Arguments:**

- `article_id` - ID of the article from search results (required)



**Example:**

```bash

quantcoder download 3

```



---



#### `summarize` - Summarize Article



Analyze and summarize a downloaded article using AI.



```bash

quantcoder summarize 1

```



**Arguments:**

- `article_id` - ID of the downloaded article (required)



**Example:**

```bash

quantcoder summarize 1

```



---



#### `generate` - Generate QuantConnect Code



Generate trading algorithm code from a research article.



```bash

quantcoder generate 1

quantcoder generate 1 --open-in-editor

quantcoder generate 1 --open-in-editor --editor code

```



**Arguments:**

- `article_id` - ID of the article to generate code from (required)



**Options:**

- `--max-attempts` - Maximum refinement attempts (default: 6)

- `--open-in-editor` - Open generated code in editor (default: false)

- `--editor` - Editor to use (e.g., zed, code, vim)



**Example:**

```bash

quantcoder generate 2 --open-in-editor --editor zed

```



---



#### `validate` - Validate Algorithm Code



Validate algorithm code locally and optionally on QuantConnect.



```bash

quantcoder validate generated_code/algorithm_1.py

quantcoder validate my_algo.py --local-only

```



**Arguments:**

- `file_path` - Path to the algorithm file (required)



**Options:**

- `--local-only` - Only run local syntax check, skip QuantConnect validation



**Example:**

```bash

quantcoder validate generated_code/algorithm_1.py

```



---



#### `backtest` - Run Backtest on QuantConnect



Execute a backtest on QuantConnect with the specified algorithm.



**Requirements:** Set `QUANTCONNECT_API_KEY` and `QUANTCONNECT_USER_ID` in `~/.quantcoder/.env`



```bash

quantcoder backtest generated_code/algorithm_1.py

quantcoder backtest my_algo.py --start 2022-01-01 --end 2024-01-01

```



**Arguments:**

- `file_path` - Path to the algorithm file (required)



**Options:**

- `--start` - Backtest start date in YYYY-MM-DD format (default: 2020-01-01)

- `--end` - Backtest end date in YYYY-MM-DD format (default: 2024-01-01)

- `--name` - Name for the backtest



**Example:**

```bash

quantcoder backtest my_algo.py --start 2023-01-01 --end 2024-01-01 --name "My Strategy v1"

```



---



### Autonomous Mode Commands



Autonomous mode runs continuously, learning from errors and self-improving strategies over time.



#### `auto start` - Start Autonomous Strategy Generation



```bash

quantcoder auto start --query "momentum trading" --max-iterations 50

```



**Options:**

- `--query` - Strategy query (e.g., "momentum trading") (required)

- `--max-iterations` - Maximum iterations to run (default: 50)

- `--min-sharpe` - Minimum Sharpe ratio threshold (default: 0.5)

- `--output` - Output directory for strategies

- `--demo` - Run in demo mode (no real API calls)



**Example:**

```bash

quantcoder auto start --query "mean reversion" --max-iterations 100 --min-sharpe 1.0

```



---



#### `auto status` - Show Autonomous Mode Status



Display statistics and learning progress from autonomous mode.



```bash

quantcoder auto status

```



Shows:

- Total strategies generated

- Success rate

- Average Sharpe ratio

- Common errors and fix rates



---



#### `auto report` - Generate Learning Report



Generate a comprehensive report from autonomous mode runs.



```bash

quantcoder auto report --format text

quantcoder auto report --format json

```



**Options:**

- `--format` - Report format: text or json (default: text)



---



### Library Builder Commands



Library builder mode systematically generates strategies across all major categories.



#### `library build` - Build Strategy Library



Build a comprehensive library of trading strategies.



```bash

quantcoder library build --comprehensive --max-hours 24

quantcoder library build --categories momentum,mean_reversion

```



**Options:**

- `--comprehensive` - Build all categories

- `--max-hours` - Maximum build time in hours (default: 24)

- `--output` - Output directory for library

- `--min-sharpe` - Minimum Sharpe ratio threshold (default: 0.5)

- `--categories` - Comma-separated list of categories to build

- `--demo` - Run in demo mode (no real API calls)



**Example:**

```bash

quantcoder library build --categories momentum,arbitrage --max-hours 12

```



---



#### `library status` - Show Library Build Progress



Display the current progress of library building.



```bash

quantcoder library status

```



---



#### `library resume` - Resume Library Build



Resume an interrupted library build from checkpoint.



```bash

quantcoder library resume

```



---



#### `library export` - Export Completed Library



Export the completed strategy library.



```bash

quantcoder library export --format zip --output library.zip

quantcoder library export --format json --output library.json

```



**Options:**

- `--format` - Export format: zip or json (default: zip)

- `--output` - Output file path



---



### Evolution Mode Commands (AlphaEvolve-Inspired)



Evolution mode uses LLM-generated variations to optimize trading algorithms through structural changes.



#### `evolve start` - Start Strategy Evolution



Evolve a trading algorithm through multiple generations of variations.



```bash

quantcoder evolve start 1

quantcoder evolve start 1 --gens 5

quantcoder evolve start --code algo.py

quantcoder evolve start --resume abc123

```



**Arguments:**

- `article_id` - Article number to evolve (optional if using --code or --resume)



**Options:**

- `--code` - Path to algorithm file to evolve

- `--resume` - Resume a previous evolution by ID

- `--gens` - Maximum generations to run (default: 10)

- `--variants` - Variants per generation (default: 5)

- `--elite` - Elite pool size (default: 3)

- `--patience` - Stop after N generations without improvement (default: 3)

- `--qc-user` - QuantConnect user ID (or set QC_USER_ID env var)

- `--qc-token` - QuantConnect API token (or set QC_API_TOKEN env var)

- `--qc-project` - QuantConnect project ID (or set QC_PROJECT_ID env var)



**Example:**

```bash

quantcoder evolve start 1 --gens 20 --variants 10 --qc-user 123456 --qc-token abc123 --qc-project 789

```



**Note:** Evolution explores structural variations like:

- Indicator changes (SMA → EMA, adding RSI, etc.)

- Risk management modifications

- Entry/exit logic changes

- Universe selection tweaks



---



#### `evolve list` - List Saved Evolutions



Show all saved evolution runs with their status and performance.



```bash

quantcoder evolve list

```



---



#### `evolve show` - Show Evolution Details



Display detailed information about a specific evolution.



```bash

quantcoder evolve show abc123

```



**Arguments:**

- `evolution_id` - The evolution ID to show (required)



---



#### `evolve export` - Export Best Algorithm



Export the best algorithm from an evolution run.



```bash

quantcoder evolve export abc123

quantcoder evolve export abc123 --output my_best_algo.py

```



**Arguments:**

- `evolution_id` - The evolution ID to export from (required)



**Options:**

- `--output` - Output file path



---



### Utility Commands



#### `config-show` - Show Current Configuration



Display the current QuantCoder configuration.



```bash

quantcoder config-show

```



Shows:

- Model configuration (provider, model, temperature, max tokens)

- UI configuration (theme, auto-approve, token usage display)

- Tools configuration (directories, enabled tools)

- Paths (home directory, config file)



---



#### `version` - Show Version Information



Display the current version of QuantCoder.



```bash

quantcoder version

```



---



### Global Options



These options can be used with any command:



- `--verbose` or `-v` - Enable verbose logging

- `--config` - Path to custom config file

- `--prompt` or `-p` - Run in non-interactive mode with a prompt



**Example:**

```bash

quantcoder --verbose search "algorithmic trading"

quantcoder --config my_config.toml generate 1

```



---



## 📁 Articles and Strategies



The folder 'Strategies and publications' contains articles and trading strategies generated using this CLI tool. These strategies may have been manually refined or enhanced using LLM-based methods. Use them at your own discretion — conduct thorough research and validate before live use.



---



## 📜 License



This project is licensed under the Apache License 2.0. See the [LICENSE](LICENSE) file for details.