OpenAI / Anthropic LLM Generator

[github-actions]

Description

Add OpenAIGenerator and/or AnthropicGenerator that format retrieved chunks as context and call the respective chat completion API to produce a grounded answer.

Motivation

EchoGenerator is a placeholder that just returns the retrieved text. Real applications need an LLM to synthesise a coherent answer from the retrieved context. This is the final piece needed for a fully functional RAG pipeline.

Acceptance criteria

• OpenAIGenerator in ragframework/generator/openai.py and/or AnthropicGenerator in ragframework/generator/anthropic.py
• Subclasses Generator from ragframework/base.py
• Configurable model, api_key (env var fallback), system_prompt, max_tokens
• Formats context chunks into a clear prompt (chunk number + content)
• Uses openai / anthropic SDK (already in optional extras in pyproject.toml)
• Import guarded; raises GeneratorError on failure
• Unit tests using unittest.mock — no real API calls in CI
• Exported from ragframework/generator/__init__.py
• Usage example added to examples/
• CHANGELOG.md updated under [Unreleased]

Files to touch

• ragframework/generator/openai.py and/or ragframework/generator/anthropic.py
• ragframework/generator/__init__.py — export
• tests/test_generator/ — new test files
• examples/ — add example script

Resources

• OpenAI Chat Completions
• Anthropic Messages API
• EchoGenerator in ragframework/generator/echo_generator.py as reference