feat: add new backends to agent

[PeterStaar-IBM]

Overview

This PR introduces a flexible backend abstraction layer and significantly enhances the enricher agent capabilities, making the docling-agent more extensible and powerful.

Major Changes

🔧 Backend Abstraction Layer

• New Backend System: Introduced a pluggable backend architecture supporting multiple LLM providers:
  • Ollama
  • LM Studio
  • LiteLLM
  • Mellea (default)
  • OpenAI-compatible endpoints
• Backend Factory: Added factory pattern for backend instantiation with registry system
• Configuration: Enhanced BackendConfig with support for base URLs, timeouts, API keys, and provider-specific options
• Updated Examples: Refactored all example scripts to use the new backend configuration system

🤖 Enricher Agent Improvements

• Flexible Entity Detection: Redesigned enricher to support more flexible entity finding and extraction
• Metadata Propagation: Enhanced writer agent to propagate summaries and metadata into generated documents
• Improved Orchestration: Better coordination between enricher and other agents

📝 Editor Agent Enhancements

• Configuration Support: Added YAML-based task configurations for editor, enricher, and writer agents
• Validation: Improved editor validation with comprehensive test coverage
• Prompts: Enhanced prompt templates for better agent behavior

🧹 Code Quality & Maintenance

• Type Hints: Added Self return type hints for better type safety
• Refactoring: Cleaned up base functions and removed redundant code
• Dependencies: Upgraded docling-core and other dependencies
• Documentation: Updated README with backend usage examples

✅ Testing

• Added comprehensive tests for:
  • Backend factory and instantiation
  • Direct backend usage
  • Task model backend configuration
  • Editor validation
  • Agent instantiation with new backend system

Breaking Changes

• Agent initialization now requires backend configuration instead of direct model parameters
• Examples updated to reflect new initialization pattern

Migration Guide

Old:

from docling_agent import WriterAgent
agent = WriterAgent(model="gpt-4")

New:

from docling_agent import WriterAgent, BackendConfig
config = BackendConfig(type="mellea")
agent = WriterAgent(backend_config=config)

[github-actions]

✅ DCO Check Passed

Thanks @PeterStaar-IBM, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

🟢 Require two reviewer for test updates

Wonderful, this rule succeeded.

When test data is updated, we require two reviewers

• #approved-reviews-by >= 1

[codecov]

Codecov Report

❌ Patch coverage is 37.02624% with 648 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
docling_agent/agent/enricher.py	10.16%	274 Missing ⚠️
docling_agent/agent/writer.py	11.56%	153 Missing ⚠️
docling_agent/cli/__init__.py	0.00%	44 Missing ⚠️
docling_agent/agent/editor.py	63.41%	30 Missing ⚠️
docling_agent/agent/orchestrator.py	0.00%	29 Missing ⚠️
docling_agent/backends/openai_compatible.py	69.14%	29 Missing ⚠️
docling_agent/backends/mellea_backend.py	50.87%	28 Missing ⚠️
docling_agent/backends/ollama_backend.py	72.97%	20 Missing ⚠️
docling_agent/agent_models.py	41.66%	14 Missing ⚠️
docling_agent/agent/base.py	63.15%	7 Missing ⚠️
... and 6 more

📢 Thoughts on this report? Let us know!

[ceberam]

Great refactoring to make it backend agnostic!
I just found very small details. In addition:

• I would only merge once docling-core is released to avoid local copies of docling-core
• I would remove docs/plans and docs/specs since the implementation plans and steps of this PR will become irrelevant in the long term. But we could put the summary in the PR description for our records.

[ceberam]

🚀