Language
HomeLanguages
Sections
OpenDroid is an open-source autonomous AI agent for Android that turns your smartphone into an LLM execution environment with 12 AI provider support and 60+ ready actions.
PR-AF is an open-source tool for deep code auditing that builds dynamic agent graphs, extracts AST trees, and verifies each finding for falsifiability.
Ponytail is a ruleset and hooks for LLM agents that forces models to write minimal code, use native features, and skip overengineering. Benchmarks show 54% less code and 20% lower costs.
Boris Cherni from Anthropic no longer writes prompts manually—automatic loops do it instead. Here's how Loop Engineering changes the game.
txtai is an all-in-one Python library combining vector database, hybrid search with SQL, graph analytics, and LLM orchestration for RAG pipelines.
Kun combines code, text, design, and automation in one local AI workspace with a shared runtime for TUI and GUI, five specialized workspaces, and flexible model support.
Building AI agents with local LLMs? SIE (Superlinked Inference Engine) consolidates multiple inference servers into one, cutting GPU costs and simplifying your container setup.
Vector databases and standard RAG can't answer auditor questions about AI decisions. Semantica adds deterministic graphs, rules, and audit trails.
Developer Vitaliy built code-graph-rag to replace vector search with knowledge graphs for monorepo code analysis. It uses Tree-sitter, Memgraph, and Qdrant to understand code structure.
Aether embeds an Alpine Linux VM into a mobile AI agent, letting language models execute code locally on smartphones without external servers.
Standard inference engines struggle with autonomous AI agents. TokenSpeed offers a specialized solution combining TensorRT-LLM speed with a Python-friendly interface.
LitGPT from Lightning AI rewrites 20+ popular LLM architectures in pure PyTorch, offering fine-tuning, deployment, and benchmarking without heavy framework overhead.
jCodeMunch MCP indexes your codebase with tree-sitter and feeds AI agents only the relevant code snippets, cutting token costs by 95–99%.
Firebase team open-sourced Genkit, a framework for building AI features. Does it help combine Gemini, OpenAI, and Ollama without spaghetti code?
Stop writing adapters for every AI model. LLM Gateway unifies API calls to OpenAI, Anthropic, Vertex AI and more through a single OpenAI-compatible interface with built-in cost tracking.
Turn vague AI prompts into professional documents with 700+ ready-made Markdown instructions for engineers, lawyers, and marketers.
A guide to using avoid-ai-writing, an open-source tool that removes AI-generated clichés from text generated by language models.
A practical guide to building AI agents that actually work. The ai-agent-book repository by Bojie Li covers everything from context engineering to self-evolution without fine-tuning.
Decepticon is an autonomous Red Team agent that runs attack chains like a real pentester. It scored 98% on the XBOW benchmark and uses 16 specialized agents for different attack stages.
nanobot is an open-source AI agent engine that runs locally, supports Telegram/Discord integration, MCP tools, and long-term memory without sending data to the cloud.
Bonsai-27B uses 1-bit and ternary quantization to fit a 27B parameter model into just 3.5 GB of RAM, making it runnable on iPhones and laptops without discrete GPUs.
How to fit AI into smart glasses without going crazy from lag. Meet Needle — a 26M parameter model for instant function calling.
GraphRAG builds knowledge graphs from unstructured data, enabling LLMs to understand connections and answer global questions across massive archives.
I tried Memtrace — a tool that turns your codebase into a knowledge graph for AI agents, indexing 15,000 files in 1.5 seconds without LLM calls.