Benchmark

[spyrchat]

This pull request introduces a modular, production-ready Retrieval-Augmented Generation (RAG) system built around a configurable LangGraph agent architecture. The changes add a flexible, YAML-configurable retrieval pipeline, robust agent node implementations, and comprehensive documentation. The new system supports advanced retrieval strategies, reranking, filtering, and seamless agent workflows, making it highly extensible and suitable for production use.

The most important changes are:

Agent System Architecture & Core Nodes:

• Introduced a new LangGraph-based agent in agent/graph.py, wiring together modular nodes for query interpretation, retrieval, answer generation, and memory updating, all configurable via YAML.
• Implemented modular agent nodes:
  • query_interpreter for intent classification and routing
  • retriever supporting configurable pipelines and legacy compatibility
  • generator for answer synthesis using LLMs
  • memory_updater for chat history management
• Defined a comprehensive AgentState schema for agent state management, supporting enhanced retrieval metadata and error handling.

Configurable Retrieval Pipeline:

• Added a highly flexible retriever node (make_configurable_retriever) that loads pipeline configurations from YAML, supports multiple retrieval strategies (dense, sparse, hybrid), reranking, filtering, and rich metadata output.

Productionization & Deployment:

• Added a Dockerfile for containerized, production-ready deployment, including system dependencies and Python requirements installation.

Documentation & Developer Experience:

• Replaced the README.md with a detailed, user- and developer-focused guide covering architecture, configuration, extension, testing, and migration from legacy code.

---

Agent Architecture & Core Logic

• Added agent/graph.py to define the main LangGraph agent, connecting modular nodes for interpretation, retrieval, generation, and memory, all configurable via YAML.
• Implemented agent nodes: query interpreter (LLM-based routing), configurable retriever, generator (LLM answer synthesis), and memory updater for chat history. [1] [2] [3] [4]
• Defined a robust AgentState schema supporting enhanced retrieval, error handling, and extensibility.

Retrieval Pipeline Flexibility

• Introduced a configurable retriever node supporting YAML-based pipeline configuration, advanced reranking, filtering, and rich document metadata, with legacy compatibility for older retrieval strategies.

Deployment & Operations

• Added a Dockerfile for streamlined, production-ready containerization of the system.

Documentation & Usability

• Overhauled README.md with comprehensive instructions, architecture diagrams, extension guides, configuration examples, and migration paths from legacy code.

[spyrchat]

This pull request introduces a modular, production-ready Retrieval-Augmented Generation (RAG) system with a LangGraph agent architecture. It adds a configurable agent graph, modular pipeline components, robust testing and CI/CD setup, and comprehensive documentation for users and contributors. The most important changes are grouped below:

Agent Graph and Modular Pipeline Architecture

• Introduced agent/graph.py, which builds a configurable LangGraph agent using modular nodes for query interpretation, retrieval, generation, and memory updating, all configurable via YAML. This enables flexible and extensible RAG workflows.
• Added new node implementations: query_interpreter, retriever, generator, and memory_updater, each as a separate module under agent/nodes/, supporting clean separation of concerns and easy extension. [1] [2] [3]

Testing and CI/CD

• Added a new GitHub Actions workflow .github/workflows/pipeline-tests.yml to automate minimal, integration, and security/config validation tests, including Qdrant service integration and configuration checks for robustness.

Deployment and Packaging

• Introduced a Dockerfile using a slim Python 3.11 base image, installing dependencies and copying the full source, to support easy containerized deployment of the system.

Documentation and Developer Experience

• Replaced the README.md with a detailed overview of the architecture, features, configuration options, usage instructions, extension guidelines, and project structure, making it easier for users and contributors to understand and work with the system.