Add initial working implementation using search grounding

[amrit110]

This pull request introduces a new agent evaluation framework for knowledge-grounded question answering (QA) using Gemini's Google Search grounding, along with rich display utilities for evaluation outputs. It adds well-structured modules for agent implementation, configuration, evaluation, and user-friendly display, and updates environment and linting configurations to support the new features.

Agent and Evaluation Framework:

• Introduced a knowledge-grounded QA agent leveraging Gemini's Google Search grounding, including synchronous and asynchronous answering, response parsing, and citation formatting, with clear system instructions and guidelines. (aieng-eval-agents/aieng/agent_evals/knowledge_agent/agent.py)
• Added an asynchronous client manager for efficient resource management and lifecycle control of agent and grounding tool instances. (aieng-eval-agents/aieng/agent_evals/knowledge_agent/agent.py)
• Provided a comprehensive package initialization for the knowledge agent, exposing all key classes for agent creation, configuration, evaluation, and session management. (aieng-eval-agents/aieng/agent_evals/knowledge_agent/__init__.py)
• Added a package-level __init__.py to expose display utilities and document the agent evaluation tools. (aieng-eval-agents/aieng/agent_evals/__init__.py)

Display and Usability Enhancements:

• Implemented rich display utilities for Jupyter and console, including functions for displaying responses, comparisons, evaluation results, metrics tables, and styled messages, all using the rich library for improved readability and user experience. (aieng-eval-agents/aieng/agent_evals/display.py)

Configuration and Tooling:

• Updated the example environment file to include Gemini/OpenAI-compatible LLM settings and optional LangFuse tracing configuration, replacing placeholder variables with relevant API keys and model selection options. (.env.example)
• Expanded pre-commit linting rules to ignore additional style and import errors, improving development workflow for the new codebase. (.pre-commit-config.yaml)

[fcogidi]

This is not really an agent. It doesn't follow the "LLM + tools + loop" format or the ReAct framework introduced in the agent bootcamp. It's tricky in this case, since the tool itself is agent-like.

[amrit110]

Yeah, thanks will update it.

[amrit110]

Oops this got merged by mistake. Need to revert it.