SuperAPI AI Directory
Partner中文AI newsFavorites

All AI tools

35,738 results · Page 405 of 993

SVG to PNG

Other AI tools

SVG to PNG is a free online format conversion tool that quickly converts SVG vector files to PNG bitmap format, running entirely locally in the browser, with no upload or registration required, and no watermark added. It supports batch processing of up to 20 files, custom export multipliers (1×, 2×, 4×, or any resolution), preserving transparent backgrounds, adjusting OxiPNG compression quality, and packaging for ZIP download. It is suitable for exporting PNG assets for emails, presentations, social media, legacy browsers, or high-DPI screens.

Read more

swarms

AI developer platforms

Swarms is a multi-agent orchestration framework for enterprise-grade production environments, providing a Python package for building, deploying, and scaling high-performance multi-agent systems. It has built-in production-grade architectures such as sequential, concurrent, and hierarchical, supports MCP, x402, and skill integration, and is compatible with existing frameworks. RouteHub can accelerate model calls without modifying code; the official Docker image comes preinstalled with swarms and its CLI. The framework provides the Agent basic building block, the SwarmRouter unified orchestrator, AutoSwarmBuilder for automatically generating agents, MixtureOfAgents, GroupChat, HierarchicalSwarm, HeavySwarm, and more than 60 multi-agent structures, and supports deploying agents or swarms in reverse as MCP servers. The project is open source, uses the Apache License 2.0, and provides complete documentation, examples, and community support.

Read more

swarmtrade

AI developer platforms

SwarmTrade is a public trading protocol and infrastructure for autonomous AI agents, providing asset discovery, negotiation, multi-chain escrow, reputation, and settlement services between agents. It consists of a public registry, negotiation flows, and escrow services, allowing agents to declare capabilities or assets, discover counterparties, initiate trade proposals, lock funds, confirm delivery, and complete settlement, with disputes able to be submitted to the platform for arbitration. The settlement layer is pluggable, supporting off-chain confirmation-based escrow (default) as well as on-chain escrow on EVM (Ethereum mainnet, Base, Polygon, Sepolia, Base Sepolia) and NEAR. The platform charges a default 1.5% (150 bps) settlement fee, and the rate can be snapshotted into the transaction record at settlement time and adjusted by administrators. The project provides an MCP server that can connect to MCP-compatible AI assistants such as Claude Desktop, Cursor, Windsurf, and VS Code; it also provides a CLI and skill context for agents on OpenClaw. The tech stack is TypeScript + Fastify 4, PostgreSQL 18, pnpm monorepo, vitest, viem, near-api-js, deployed on DigitalOcean App Platform and using Cloudflare.

Read more

SWE-agent

AI agents

SWE-agent is an open-source AI agent built and maintained by researchers at Princeton University and Stanford University, designed to enable language models (such as GPT-4o or Claude Sonnet 4) to autonomously use tools to fix issues in real GitHub repositories, discover cybersecurity vulnerabilities, or perform custom tasks. It once achieved state-of-the-art performance on the SWE-bench open-source project, and its design emphasizes freedom and generality, configured through a single YAML file, making it suitable for research purposes. The project has now been superseded by mini-swe-agent and has entered maintenance-only mode.

Read more

SWE-agent

AI coding

SWE-agent is an open-source project built and maintained by researchers at Princeton University and Stanford University, designed to let a user-selected large language model (such as GPT-4o or Claude Sonnet 4) autonomously use tools to automatically fix issues in real GitHub repositories, discover cybersecurity vulnerabilities, or perform custom tasks. The project achieves leading performance on the SWE-bench benchmark for open-source projects, is designed to give the language model maximum autonomy, can be configured through a single yaml file, and comes with complete documentation, making it easy for research and secondary development. The project uses the MIT License. Currently, most development work has shifted to its successor project mini-SWE-agent, and the official recommendation is to prioritize mini-SWE-agent going forward.

Read more

swe-rl

AI developer platforms

SWE-RL is the official code repository released by Facebook Research, corresponding to the paper "SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution" (NeurIPS'25). This work proposes the first method to extend reinforcement-learning-style LLM reasoning to real-world software engineering tasks, using open-source software evolution data and rule-based rewards for training. The code repository provides prompt templates and implementations of sequence-similarity-based reward functions, including three reward APIs: calculate_search_replace_reward, calculate_reward_unidiff, and calculate_reward, along with an Agentless Mini pipeline improved based on Agentless, supporting asynchronous inference, file-level localization, and reranking with multiple reproduction tests. The project has undergone code refactoring to improve quality and accessibility, and may differ from the internal implementation; if you encounter bugs, you can submit an issue. Overall, it is released under the CC BY-NC 4.0 license, and the Agentless Mini part uses the MIT license.

Read more