Skip to content

DeepNext is an AI-powered junior software engineering assistant designed to automate the process of converting issue descriptions into actionable pull requests.

License

Notifications You must be signed in to change notification settings

stxnext/deep-next

Repository files navigation

DeepNext

Your AI-powered junior software engineering assistant that takes an issue description and converts it into a pull request.

DeepNext is an advanced AI system that automatically analyzes a code repository, prepares a solution plan, and implements the necessary code for software engineering tasks using Large Language Models, saving developers hours of repetitive work.

See documentation for more information.

Quick start

Getting Started

GitHub/GitLab integration

Running as a service to automatically process issues:

See integration details

Configuration

DeepNext supports multiple LLM providers:

  • OpenAI
  • AWS Bedrock (Claude, Mistral, and others)
  • Ollama (for local LLM usage)

Create an llm-config.yaml file based on the example provided to configure model preferences for each pipeline stage. See Ollama integration for details on using local models.

For tracking and metrics, DeepNext integrates with LangSmith. Set up your credentials in the .env file.

See configuration details

Roadmap

  • May 2025: First public release (for registered STX Next employees)
  • June 2025: Open-source

Why choose DeepNext?

  • End-to-End Automation: Complete pipeline from issue to merge request
  • Multiple LLM Support: Works with various models to fit your needs
  • Flexible Integration: Compatible with GitHub, GitLab and AWS
  • Customizable: Configure different models for different pipeline stages
  • Modular: Easy to modify, test, and improve by swapping small pieces of code

License

This project is licensed under the Apache License 2.0. See the LICENSE file in this repository for full terms.


Third-Party Licenses

This project depends on open-source libraries released under the following licenses:


Deprecated, PyJWT, PyYAML, SQLAlchemy, annotated-types, anyio, appdirs, attrs, backoff, cffi, cfgv, charset-normalizer, dataclasses-json, h11, httpx-sse, identify, iniconfig, jiter, jmespath, langchain, langchain-aws, langchain-community, langchain-core, langchain-ollama, langchain-openai, langchain-text-splitters, langfuse, langgraph, langgraph-checkpoint, langgraph-prebuilt, langgraph-sdk, langsmith, libcst, loguru, markdown-it-py, marshmallow, mdurl, mypy_extensions, orjson, ormsgpack, platformdirs, pluggy, pre_commit, pydantic, pydantic-settings, pydantic_core, pyee, pyppeteer, pytest, RapidFuzz, rich, six, slack_sdk, tiktoken, typing-inspect, typing-inspection, unidiff, urllib3, virtualenv


PyNaCl, aiohttp, aiosignal, boto3, botocore, cryptography, distro, frozenlist, importlib_metadata, openai, orjson, ormsgpack, packaging, propcache, requests, requests-toolbelt, s3transfer, sniffio, sortedcontainers, tenacity, yarl, regex, python-dateutil


Pygments, click, cryptography, httpcore, httpx, idna, jsonpatch, jsonpointer, nodeenv, numpy, packaging, pycparser, python-dateutil, python-dotenv, scipy, websockets, wrapt, xxhash


aiohappyeyeballs, distlib, greenlet, typing_extensions


certifi, pytest-rerunfailures, tqdm


PyGithub


python-gitlab


About

DeepNext is an AI-powered junior software engineering assistant designed to automate the process of converting issue descriptions into actionable pull requests.

Topics

Resources

License

Contributing

Stars

Watchers

Forks

Packages

No packages published

Contributors 3

  •  
  •  
  •  

Languages