[UdayKumar9381]

Summary

• add OpenAIGenerator using the OpenAI Chat Completions API
• support model, API key/env fallback, system prompt, and max tokens
• add mocked unit tests and OpenAI usage example
• export generator and update changelog

Testing

• pytest tests/ -v — 93 passed
• ruff check ragframework/generator/openai.py tests/test_generator/test_openai.py — passed
• mypy ragframework/ — passed
• python -m py_compile examples/openai_rag.py — passed

Closes adaumsilva#8

[UdayKumar9381]

Hi @adaumsilva,

I've completed the OpenAI LLM Generator implementation for Issue adaumsilva#8 and opened this draft PR.

The implementation includes:

• OpenAIGenerator with configurable model, API key/env fallback, system prompt, and max tokens
• Context formatting with numbered chunks
• GeneratorError handling
• Mocked unit tests
• Package export
• OpenAI RAG example
• CHANGELOG update

Validation:

• 93 tests passed
• Ruff passed
• Mypy passed
• Example compilation passed

The PR is currently conflict-free and ready for your review.

Thanks!

[adaumsilva]

Hi @UdayKumar9381, thanks for picking this up! The implementation looks solid overall. A few things before I can review it properly:

1. The PR is opened against your fork, not this repo.
The base is UdayKumar9381/RAG-framework:main, so it doesn't show up here. Please close it and open a new one from the same branch with base repository: adaumsilva/RAG-framework, base: main. On the compare page, use the "base repository" dropdown.

2. Please sync with upstream first. Your fork is about 40 commits behind, and the branch currently conflicts with main in CHANGELOG.md and ragframework/generator/__init__.py (AnthropicGenerator has landed since you forked).

git remote add upstream https://github.com/adaumsilva/RAG-framework.git
git fetch upstream
git rebase upstream/main
git push --force-with-lease

When resolving: keep both exports in __init__.py, and move your changelog line under ## [Unreleased] → ### Added (0.2.0 is already released).

3. The import guard needs to be lazy (this breaks CI). openai.py raises ImportError at module import time, and generator/__init__.py imports it unconditionally. So without the openai extra installed, from ragframework.generator import EchoGenerator fails. CI doesn't install openai, so the whole generator test package errors out. Please move the from openai import OpenAI inside __init__, the same way ragframework/generator/anthropic.py does it.

4. Tests shouldn't need the real openai package. Right now they construct a real client and patch its methods. Please follow the pattern in tests/test_generator/test_anthropic.py (a fake module injected via sys.modules), so they pass in CI without the extra. While you're there, please add tests for (a) the missing-package ImportError and (b) a missing API key.

Smaller things:

• Validate the API key up front and raise a clear GeneratorError ("Pass api_key or set OPENAI_API_KEY"), like AnthropicGenerator does.
• Add a trailing newline to examples/openai_rag.py and generator/__init__.py, and run ruff format on the new files.
• Update the roadmap row in README.md for the OpenAI generator.

Once that's up against this repo with CI green, I'll review it right away. Thanks again!