Custom agents

TypeScript-фреймворк для разработки голосовых навыков и чат-ботов. Он даёт единую бизнес-логику для всех платформ — но одинаково эффективен, даже если вы работаете только с одной. Поддерживаются: `Яндекс.Алиса`, `Маруся`, `Сбер Салют`, а также `Telegram`, `VK`, `MAX` и `Viber` из коробки.

SQL Server 2025 demo: graph (SHORTEST_PATH) + vector (DiskANN) + Blazor Server. SQL Server 2025: Herkes İçin, Her Rol İçin kitabının referans uygulaması.

Financial Q&A for Indian equities. Deterministic ratio engine + RAG-powered LLM for NSE/BSE companies. CLI, API, dashboard.

Evidence-first product teardown skill for any AI product: journeys, agents, models, tools, context, knowledge, evaluation, safety, and architecture.

Top Knowledge Management AI (Opensource) 🌟 Star if you like it! 🌟

AI-powered LeetCode analytics platform — syncs your solving history and uses RAG (Gemini + ChromaDB) to answer personalized questions about your strengths, weaknesses, and growth areas. Built as a Turborepo monorepo with real-time SSE sync, BullMQ background jobs, and full observability (Prometheus/Loki/Grafana)

Memory infrastructure for AI agents: store, recall by meaning or association, ingest documents, and track beliefs in one Python library.

A framework for Forward Deployed Engineers: from problem statement to a deployable AI project — every decision traced to a fact, every fact to a source.

Lightweight local memory for agents — every dedup, update, and rerank decision inspectable and tunable. 3 dedup detectors × 4 update resolvers × 3 rerank modes · 60+ parameters · 6 prompts replaceable

Build provenance-aware higher-order knowledge graphs and native hypergraphs from documents, with validation, semantic search, and offline visualization.

Stateful workflow checkpoint manager calculating state hashes and rollback points

Sentence-level RAG factual grounding and hallucination detection engine

Long-horizon conversational memory compaction & episodic anchor extractor (Mem0 / LangMem)

Desktop app for Claude Code: a roster of bots, each with its own model, working directory and persistent conversation

AI agents and skills for mobile development: Flutter, Android, iOS and React Native. Plugin for Claude Code and Codex.

Multi-agent debate cross-examination & consensus arbitration (Super Agent / DeepSeek-R1)

Git 提交一键转禅道任务：自动汇总需求、按 T1~T7 难度阶梯定工时、日期铺排与任务闭环，生成经得起审计的工时汇报。AI Agent 技能包，支持 Gitea/GitHub/Gitee，MIT 开源。

Autonomous web browser action planner mapping goals into resilient selector action sequences

Speculative decoding draft-model verification engine calculating acceptance probabilities

Customer review aspect-level sentiment mining engine extracting satisfaction radar breakdown

Cross-border Geo-IP currency converter & dynamic VAT/tax nexus calculator (Stripe Tax)

Multimodal visual spec grounding engine calculating physical dimensions and DoF

Self-hosted AI social media automation: Gemini content generation, multi-platform publishing, analytics & self-optimizing strategies

Self-hosted AI agent harness: web UI + HTTP API, sandboxed tool execution, persistent workspace memory, selection quoting, browser and computer use. TypeScript, zero runtime deps, one command to start.

Autonomous refund and dispute arbitration engine validating carrier proofs and policies

Semantic prompt cache deduplicator calculating Jaccard token overlap to prevent redundant LLM cost

Per-turn model routing for the Pi coding agent: a small judge picks the cheap or the strong tier for each message, with multi-model failover, task-level orchestration, and an optional decision-model judge (Jev) that answers with a calibrated probability instead of prose.

Heterogeneous search synthesis to interactive comparison layout (Genspark Sparkpages)

Autonomous multi-agent orchestration engine implementing the blackboard architectural pattern in Python & Redis.

Local-first Windows desktop workspace for Codex, OpenCode, Cursor Agent, GitHub Copilot, Gemini CLI, and other AI coding agents.

Captain Code: one terminal for every coding agent you already pay for. Local director routing across Claude Code, Codex, Cursor, and remote models via opencode.

A programmable AI agent harness built in Rust and configured with Lua. Create custom coding agents with a terminal UI, multi-provider LLM support, MCP tools, persistent sessions, and an extensible plugin system.

Self-hostable multi-agent AI workbench — LangGraph-orchestrated expert teams, human-in-the-loop plan approval, live run timeline. FastAPI + React + PostgreSQL.

Enforceable, provider-neutral harness for multi-agent AI coding: spec-driven workflow, cross-provider peer review, and safety rails. First-class for Claude Code and Pi; advisory adapters for Cursor, Codex, Copilot, Windsurf, and Aider.

Self-Consistency majority voting engine aggregating multiple stochastic reasoning paths to extract high-confidence consensus solutions.

Production-grade genpark-hnsw-hierarchical-navigable-small-world-skill skill for AI agents

First-Order Logic (FOL) resolution refutation automated theorem prover deriving empty clause contradictions for unsatisfiability proofs.

Cont-Kukanov-Stoikov Order Flow Imbalance (OFI) predictor forecasting short-term price movements via quote level depth dynamics.

Production-grade genpark-neighbor-joining-phylogenetic-tree-skill skill for AI agents

Signal double ratchet state machine advancing symmetric KDF chains and ephemeral Diffie-Hellman keys for forward secrecy.

Robinson's first-order term unification algorithm with occurs-check for automated theorem proving, Prolog engines, and type inference.

Production-grade genpark-lmax-disruptor-ring-buffer-skill skill for AI agents

Production-grade genpark-lmax-disruptor-ring-buffer-skill skill for AI agents

Production-grade genpark-chaitin-briggs-graph-coloring-allocator-skill skill for AI agents

FRI (Fast Reed-Solomon Interactive Oracle Proof of Proximity) folding engine verifying polynomial low-degree bounds for STARKs.

Dynamic Octree 3D hierarchical spatial indexer supporting logarithmic range queries, collision volume culling, and point neighbor queries.

Stoer-Wagner algorithm finding the global minimum cut in an undirected weighted graph without designating source and sink nodes in O(V^3).

Lookahead optimizer wrapper stabilizing optimization trajectories by interpolating fast and slow weights every k steps.

Gale-Shapley deferred acceptance algorithm engine solving stable matching problems with Pareto-optimal proposer guarantees.

Smith-Waterman dynamic programming algorithm finding exact optimal local alignment motifs, homologous domains, and substrings across DNA and protein sequences.

ZooKeeper Atomic Broadcast (Zab) Fast Leader Election protocol resolving epochs, ZXID transaction counters, and majority quorum vote updates.

Exact Variable Elimination inference engine for discrete Bayesian Networks computing conditional posteriors and marginal probability distributions.

Hindley-Milner Algorithm W type inference engine implementing Robinson first-order syntactic unification and principal polymorphic type reconstruction.

Multi-Version Concurrency Control (MVCC) engine guaranteeing non-blocking repeatable reads, snapshot isolation, and version chain garbage collection.

Conflict-Driven Clause Learning (CDCL) 2-Watched Literals (2WL) clause indexer accelerating Boolean constraint propagation.

Mel-scale triangular filterbank spectral analyzer mapping linear frequency spectra to perceptual log-mel energies for speech AI.

Production-grade genpark-geohash-spatial-quadtree-encoder-skill skill for AI agents

Production-grade genpark-cbor-concise-binary-object-representation-skill skill for AI agents

Linear Temporal Logic (LTL) trace validator supporting Globally (G), Finally (F), Next (X), and Until (U) operators.

Poseidon arithmetic hash function optimized for zero-knowledge circuits using MDS mixing matrix and prime field S-boxes.

Shamir (k, n) threshold secret sharing scheme with polynomial interpolation and cryptographic threshold reconstruction.

Reflexion metacognitive actor-evaluator agent maintaining verbal reinforcement reflection memory for iterative self-correction across trials.

Anti-entropy Merkle tree range reconciliation protocol for fast replica divergence detection and state synchronization.

MuSig2 multi-signature and threshold Schnorr aggregation protocol for compact collective authorization proofs.

Production-grade genpark-chase-lev-work-stealing-deque-skill skill for AI agents

Production-grade genpark-raft-state-machine-replication-skill skill for AI agents

Production-grade genpark-tdigest-streaming-quantile-sketch-skill skill for AI agents

Production-grade genpark-michael-scott-lockfree-queue-skill skill for AI agents

Production-grade genpark-hamming-7-4-syndrome-decoder-skill skill for AI agents

Chaitin-Briggs graph coloring register allocator featuring Kempe heuristics, interference graph simplification, and optimistic spill recovery.

Bulletproofs Inner Product Argument (IPA) recursive halving reduction proving inner product satisfiability without trusted setup.

Online Softmax Tiling FlashAttention kernel executing bounded SRAM query/key/value block updates with running max and sum rescaling for O(1) memory footprint.

Strict Two-Phase Locking (SS2PL) concurrency control with Wound-Wait timestamp-based deadlock prevention algorithm.

Flajolet-Martin (FM) distinct elements streaming sketch estimating unique cardinality from trailing zero bit patterns with logarithmic memory.

Production-grade genpark-dpdk-zero-copy-packet-ring-buffer-skill skill for AI agents

Production-grade genpark-michael-scott-lockfree-queue-skill skill for AI agents

Polynomial and Vector Commitment simulation framework proving evaluation correctness f(z)=y without revealing the entire vector payload.

Operating system two-handed clock virtual memory page replacement algorithm with reference bit clearance and proactive page aging.

Ultra-low-latency Price-Time Priority Level-3 (L3) Limit Order Book (LOB) matching engine with continuous price queues and fills.

A bot application for Synology DSM Chat package

Production-ready frontend with Google login, built using Next.js and integrated with a Cloud Run. This project is part of the GCP AI Agent Starter Kit workshop and provides a clean, customizable interface to connect conversational agents to intelligent backend services. Ideal for developers looking for a practical, reusable starting point

The harness that makes consumer-GPU models feel first-class. Local-first: no telemetry, no account, no cloud tier.

Local browser AI agent with indexed action space — Laya local decisions, /opt/homebrew/bin/bash/run. Google Flights demo in ~6.5s. Python.

Claude Code plugin that compacts the conversation at sub-task boundaries instead of at the context limit

Helix — reference runtime for the Agent Execution Protocol (AEP). Source-available under the Fair Source License.

Managed multi-agent system for OpenCode with one coordinator, risk-based routing, specialized roles, model pools and failover, durable memory, controlled context compaction and recovery, independent review, and execution telemetry.

Local code retrieval for codebases too big to grep

Customer briefing and meeting intelligence on AWS. React, Python, Bedrock, AgentCore, RAG, and a durable SQS job pipeline.

End-to-end Claude Code skill for job applications. 8 phases, 2 sub-skills, originally built for UK AI startup hunt, generalised. Bilingual EN/中文.

FEMO插件版 — 可接入dsh的多智能体引擎，差不多能做所有事：我已经用femo和我的主agent玩了好几局狼人杀了，他抽到狼他还刀我！ 我在b站的宣传视频也是用femo剪的。 coding的时候，特别难找的bug、代码重构，我都会用femo来做。 我甚至用femo模式给我的主agent养了只宠物，她很开心。

EdgeLab Trading Skills｜给 AI Agent 用的投研与交易复盘工具集 · 11 Skills · CC BY-NC 4.0

Local message bus for AI coding agents running in Herdr panes — join, DM, presence, offline delivery

A local AI Agent workbench for branching conversations, visual context management, parallel exploration, and tool-driven project workflows.

LT is a local-first AI interactive storytelling platform. Create characters, build worlds, manage story resources, and experience dynamic adventures with AI-powered generation and branching narratives.

A governance layer for AI coding work — not an orchestrator. Seven gated roles with typed submissions, independent audit seats, and deliberately no runner: composition stays with the caller.

A local-first AI learning assistant that turns browser videos and local media into grounded, structured study notes.

Autonomous business operating system for solo founders — an AI C-suite that classifies directives, runs multi-agent debates, tracks every LLM call as a real expense, and executes work 24/7.

Agent 测试转型手册：资深自动化测试工程师的 AI QE 学习路径 + 练手代码

AI assistant for interviews: on-device speech recognition with real-time answers.

RETECO: A SemEval-2027 Shared Task on Reasoning-Oriented Retrieval