Metadata-Version: 2.4
Name: co-datascientist
Version: 0.1.9
Summary: A tool for agentic recursive model improvement
Project-URL: Homepage, https://github.com/TropiFloAI/co-datascientist
Project-URL: Issues, https://github.com/TropiFloAI/co-datascientist/issues
Author-email: David Gedalevich <davidgdalevich7@gmail.com>
License: Copyright (c) 2018 The Python Packaging Authority
        
        Permission is hereby granted, free of charge, to any person obtaining a copy
        of this software and associated documentation files (the "Software"), to deal
        in the Software without restriction, including without limitation the rights
        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
        copies of the Software, and to permit persons to whom the Software is
        furnished to do so, subject to the following conditions:
        
        The above copyright notice and this permission notice shall be included in all
        copies or substantial portions of the Software.
        
        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
        SOFTWARE.
License-File: LICENSE
Requires-Python: >=3.12
Requires-Dist: click>=8.1.8
Requires-Dist: fastmcp>=2.2.5
Requires-Dist: httpx>=0.28.1
Requires-Dist: ipdb>=0.13.13
Requires-Dist: keyring>=25.6.0
Requires-Dist: keyrings-alt>=5.0.0
Requires-Dist: pydantic-settings>=2.9.1
Requires-Dist: yaspin>=3.1.0
Description-Content-Type: text/markdown

# CoDatascientist

<div align="center">

![Version](https://img.shields.io/badge/version-1.0.0-blue.svg)
![License](https://img.shields.io/badge/license-MIT-green.svg)

<div align="center">
  <img src="figures/logo2.png" alt="PiñaColada Logo" width="500"/>
</div>

An agentic framework for recursive model improvement.

</div>

## 🚀 Quickstart
Install `co-datascientist`:

```bash
pip install co-datascientist
```

To use from command line, run:
```bash
co-datascientist run --script-path myscript.py
```

For development with a local backend, use the `--dev` flag:
```bash
co-datascientist --dev run --script-path myscript.py
```

To use from cursor or other AI clients, run the MCP server:
```bash
co-datascientist mcp-server
```

For development mode:
```bash
co-datascientist --dev mcp-server
```

And add the MCP configuration to the AI client. For example in cursor go to:
`file -> preferences -> cursor settings -> MCP -> Add new global MCP server`,
and add the co-datascientist mcp server config in the json, should look like this:
```json
{
  "mcpServers": {
    "CoDatascientist": {
        "url": "http://localhost:8001/sse"
    }
  }
}
```

## 🔑 Adding Your OpenAI API Key

CoDatascientist supports using your own OpenAI API key for unlimited usage (instead of the free tier). You can manage your OpenAI key through the CLI:

### Add/Update OpenAI Key
```bash
# This will prompt you to enter your OpenAI API key
co-datascientist --reset-openai-key run --script-path myscript.py
```

Or use the dedicated key management command:
```bash
# Manage your OpenAI key
co-datascientist openai-key
```

### Remove OpenAI Key
```bash
# Switch back to free tier
co-datascientist openai-key --remove
```

### Check Current Status
```bash
# See if you're using your OpenAI key or the free tier
co-datascientist status
```

**Benefits of adding your OpenAI key:**
- 🚀 **Unlimited usage** with your OpenAI account
- 💰 **Direct billing** to your OpenAI account  
- 🔒 **No usage limits** from TropiFlow's free tier

**Note**: Get your OpenAI API key from [https://platform.openai.com/api-keys](https://platform.openai.com/api-keys)

## 💰 Cost Tracking
Track your LLM usage costs automatically:

```bash
# View summary costs
co-datascientist costs

# View detailed breakdown
co-datascientist costs --detailed
```

The system automatically tracks:
- **Total costs** across all workflows
- **Token usage** (input/output tokens)
- **Per-workflow costs** and breakdown
- **Model-specific costs** using LiteLLM's built-in pricing

All costs are calculated in real-time using LiteLLM's `completion_cost()` function with up-to-date pricing for all supported models (OpenAI, Anthropic, DeepSeek, etc.).

## 🧠 How does it work?
AI agents are advanced enough to hypothetize ideas, write the code, and evaluate results. 

While they're not yet advanced enough to do all of this independently, they are good enough to give a *worthy attempt* at each step.

CoDatascientist coordinates AI agents and automates the workflow of
Idea generation → implementation → testing.

The cycle is done iteratively, selecting the best candidates using the agent's own logic - this is why the metric for success has to be simple, which is the case with ML tasks: it's easy to boil the success to a single number or a vector.

## 📝 Example: Before vs After

See how CoDatascientist transforms simple ML code into sophisticated, mathematically-grounded solutions:

<table>
<tr>
<th>📥 Input Code</th>
<th>📤 Enhanced Output</th>
</tr>
<tr>
<td>

```python
from sklearn.ensemble import RandomForestClassifier
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import accuracy_score
import numpy as np

# XOR data
X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
y = np.array([0, 1, 1, 0])

# Define pipeline: preprocessing + model
pipeline = Pipeline([
    ('scaler', StandardScaler()),
    ('clf', RandomForestClassifier(n_estimators=10, random_state=0))
])

# Fit
pipeline.fit(X, y)

# Predict and evaluate
preds = pipeline.predict(X)
accuracy = accuracy_score(y, preds)
print(f"Accuracy: {accuracy:.2f}")
```

</td>
<td>

```python
import numpy as np
from sklearn.base import TransformerMixin, BaseEstimator
from sklearn.ensemble import RandomForestClassifier
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import accuracy_score
from tqdm import tqdm

class ChebyshevPolyExpansion(BaseEstimator, TransformerMixin):
    def __init__(self, degree=3):
        self.degree = degree
    def fit(self, X, y=None):
        return self
    def transform(self, X):
        X = np.asarray(X)
        X_scaled = 2 * X - 1
        n_samples, n_features = X_scaled.shape
        features = []
        for f in tqdm(range(n_features), desc="Chebyshev features"):
            x = X_scaled[:, f]
            T = np.empty((self.degree + 1, n_samples))
            T[0] = 1
            if self.degree >= 1:
                T[1] = x
            for d in range(2, self.degree + 1):
                T[d] = 2 * x * T[d - 1] - T[d - 2]
            features.append(T.T)
        return np.hstack(features)

X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
y = np.array([0, 1, 1, 0])

pipeline = Pipeline([
    ('cheb', ChebyshevPolyExpansion(degree=3)),
    ('scaler', StandardScaler()),
    ('clf', RandomForestClassifier(n_estimators=10, random_state=0))
])

pipeline.fit(X, y)

preds = pipeline.predict(X)
accuracy = accuracy_score(y, preds)
print(f"Accuracy: {accuracy:.2f}")
```

</td>
</tr>
</table>

### 🔧 What CoDatascientist Added:
- **Custom Feature Engineering**: Implemented Chebyshev polynomial expansion for better non-linear feature representation
- **Mathematical Rigor**: Uses orthogonal polynomials for stable numerical computation
- **Enhanced Architecture**: Maintains scikit-learn compatibility while adding sophisticated transformations
- **Better Performance**: Improved feature space for solving non-linearly separable problems like XOR

## 🛠️ Features

### Idea Generation
- 📚 Arxiv paper reader
- 💡 Idea generator
- 🔍 Idea critic
- 📊 Time series tool APIs
- 📖 RAG with ML-trading books

### Implementation
- 🛠️ AIDER integration
- 🔄 Automatic improvement system
- 📈 Metric comparison

### Cost Management
- 💰 Real-time cost tracking per user
- 📊 Token usage monitoring
- 🔍 Workflow-level cost breakdown
- 📈 Model-specific cost analysis

### 📊 Benchmarking
Coming soon!

## Any questions? Contact us!
we're happy to help, discuss and debug.

contact us at `ozsamkilim@gmail.com`

---
<div align="center">
Made with ❤️ by the CoDatascientist team
</div>