Update Conversation v2

[dillonalaird]

Major updates to improve stability:

• Adds configuration files so you can switch LLM providers and arguments more easily
• Introduces majority voting for tool choice
• Now uses o1 for code summarization from planner, removed coder
• With the above changes we can reduce temperature to 0 on more agents to improve stability

Minor updates and bug fixes:

• Fixes an issue with multi-turn conversations. Previously was not passing the correct context to the planner after 2nd turn
• Updates README with V2 code
• Fixes bug with OpenAI LMM to properly use o1 model
• Fixes prompts for V1 code that used old tool names
• Fixes prompting issue where only 1 tool was getting tested (now test minimum 2 tools)
• Tool tester now tries to mutually exclusive prompts like 'person, car' and not overlapping prompts like 'person, athlete'
• Fixes bug in code parser where it would incorrectly strip code if the function contained a nested function call
• Fixes prompt bug where it accidentally said to use save_file instead of save_image