Yashvardhan Thanvi
ReportI am the creator and primary maintainer of LLMSlim, an open-source Python project focused on improving context efficiency for LLM and agentic AI systems. I currently lead its architecture, implementation, benchmarking, releases, documentation and integrations. My goal is to build LLMSlim as reliable, vendor-neutral open infrastructure that developers can inspect, benchmark, extend and deploy without being locked into a proprietary model or platform.
LLMSlim
LLMSlim is an open-source Python infrastructure library for optimizing context supplied to large language models and AI agents. It reduces unnecessary context while protecting critical instructions, entities, structured content and execution contracts. LLMSlim supports local deterministic compression, semantic optimization, RAG and document workflows, and tool-heavy agent and MCP use cases. It is designed to remain model- and provider-independent so developers can add context optimization to existing AI stacks without replacing their current infrastructure.
LLMSlim Open Source Development Grant
Seeking INR 3,00,000 to support 12 months of open-source LLMSlim development. Funding will support maintainer time, reproducible benchmarks, multilingual and agentic workload testing, documentation, security and contributor growth. Milestones include stronger RAG/MCP/agent reliability, public evaluation methods and datasets, improved developer onboarding, and regular open development updates.