Dateno LLM

Framework and toolkit for interacting with Dateno LLM services.

Installation

To install the package, run the following command:

pip install .

Or from GitHub:

pip install git+https://github.com/datenoio/datenollm

Python >=3.10 required. Dependencies: gradio_client, logfire>=3.24.2

Framework

The main modules are located in src/datenollm/:

• client.py — API client for Dateno LLM services
• dateno.py — Dateno search API core logic
• file_utils.py — file operations
• jupiter_nb.py — Jupyter/Colab notebook helpers
• server.py — server logic
• cli/ — command-line tools:
  • ask.py, logs.py, like.py, flagged_log.py, collab2gist.py
• test/ — test data (context files)

Examples

The examples/ directory contains Jupyter notebooks (Google Colab compatible).

• examples/dateno_llm_context_search.ipynb — An example research study using Dateno's context-switching LLM search agent.
• dateno-deep-research-workflow.ipynb — An example of Deep Research Workflow

Utilities

This package provides the following command-line tools:

dateno-ask-llm

Generate structured queries for Dateno from user input using LLM.

Usage:

dateno-ask-llm <addr> <query> [--history <history_file>] [--prompt <prompt_file>] [--model <model_id>] [--max-tokens <max_tokens>] [--temperature <temperature>] [--top-p <top_p>]

dateno-get-logs

Download Gradio's flagged logs from the Gradio's app.

Usage:

dateno-get-logs <addr>

dateno-like

Flag logs in the app.

Usage:

dateno-like <addr> <index> <like|dislike> <csv_path>

dateno-flagged-log

Read and print Gradio's flagged logs from a CSV file.

Usage:

dateno-flagged-log <flagged_log_csv_file>

dateno-collab2gist

Fix Jupyter notebooks from Google Colab to a format that displays properly on GitHub Gist. It does this by reading a JSON notebook file from stdin, removing the metadata.widgets section, and writing the modified JSON to stdout.

Usage:

dateno-collab2gist < input.ipynb > output.ipynb

License

This project is licensed under the Apache-2.0 License. See LICENSE for details.