Revert "feat: Implement a parallel voting ensemble for LLM selection based on Jaccard similarity, replacing the sequential fallback mechanism."

feat: Implement a parallel voting ensemble for LLM selection based on Jaccard similarity, replacing the sequential fallback mechanism.

feat: Enhance agent robustness by separating base and fallback LLMs, update Ollama model to 7B, and improve Docker startup reliability.

feat: Enhance LLM resilience with a multi-provider fallback system, upgrade the default Ollama model to 7B, and improve vector store relevance score calculation.

feat: Add stop parameter to chat completion requests and simplify the comment regarding `<think>` tag stripping.

Remove `Qwen/Qwen1.5-110B-Chat` from the list of candidate models.

feat: Augment agent prompts with uploaded file context, clarify data storage consent, and add Qwen/Qwen1.5-110B-Chat to candidate models.

Removed artifacts

enhanced pompt

stripped artifacts

stripped artifacts

stripped of certain artifacts

added swahili language

controlling scope

citations logic

scope guardrails + advice protocol

health economics

introduced scope guardrails and confidence scoring (0-100)

increased token limit + catch <think> tags

uses langchain core

Switched to use Huggingface CHAT

trying new models that use could possibly accept the HuggingFaceEndpoint API inference

added more fallback model options

added list of candidate models

switched to qwen 32B - Instruct due to inference provider support

changed to use huggingface serverless endpoint with local CPU as a fallback

added observation logging

changed default ollama network port to match huggingface default

deployment 1.0

prototype stage