archestack/best-of-ai：155 个开源 AI 项目排名榜

Open-source AI, ranked. 🏆

Apps, agents, models and the tools around them: what each one does, what it needs, where it breaks. No fluff, and paid never moves a rank.

Best of Open-Source AI is a ranked list of 155 open-source AI projects: chat apps and assistants, desktop apps, agent platforms and frameworks, workflow automation, RAG, memory, model serving, open models, training, image, video and voice, vector databases, gateways, observability, sandboxes, data tools, MCP servers and learning material. Each entry has strengths, weaknesses, specs and a score from public GitHub signals. _{Updated 2026-10-10 (UTC).}

Tip

🔎 Every project fact-checked on GitHub before it is listed  ·  ✍️ Every entry has strengths and weaknesses  ·  ▶️ Demo links so you can try before you deploy

Note

🤝 Independent. We sell none of the tools listed here, and nobody can pay to move up. Ranks come from public GitHub data and are re-checked twice a day; hover a score to see what it is made of.

Looking for coding agents, starters and the stack you ship on? 👉 Best of Vibe Coding

🔥 Top 10 right now

_{155 projects ranked in 16 categories · data from 2026-10-10 · re-ranked twice a day}

_{How to read it: rank 🥇🥈🥉 then #⁠4, #⁠5… (equal scores: more stars first) · score out of 100 sets the order: hover it to see what it is made of, or read how we rank · click a project name for its review · ↗ opens another site (Cmd/Ctrl-click for a new tab)}

_{Ordered by score: adoption, freshness, how easy it is to run, and a few more signals as we measure them (how we rank).}

⚡ Pick by vibe

• 💬 ChatGPT, but at home → Chat UIs
• 🧠 Run models on my own GPU → Model serving
• 🤖 An assistant in my Telegram or WhatsApp → Assistants
• 📚 Chat with my own documents → RAG and knowledge
• 🔁 Wire AI into the rest of my tools → Workflow automation with AI
• 🔀 One API key for every provider → Gateways
• 🕸️ Turn websites and PDFs into clean text → Data and scraping for AI
• 📈 See what my LLM app is actually doing → Observability
• 🛡️ Let agents run code without wrecking my server → Sandboxes

🔓 Open-source alternatives to …

_{The top open-source picks by score in the matching categories. ✋ means the project's own README calls it an alternative to that product.}

💻 Runs on what you have

_{From each project README: whether it needs a GPU, and whether it mentions Apple Silicon or arm64. Apps and tools you run, not libraries.}

🗂️ All categories

🧑‍💻 Apps you run

• 🤖 Assistants · 12
• 💬 Chat UIs · 13
• 🧩 Agent platforms · 10
• 🔁 Workflow automation with AI · 2
• 📚 RAG and knowledge · 15
• 🔎 Search · 5
• 🎙️ Voice · 11
• 🎨 Image and video · 8
• 🔐 Security · 1

🧱 Build with it

• 🗂️ Memory · 12
• 🧠 Model serving · 20
• 🔀 Gateways · 14
• 📈 Observability · 13
• 🧮 Vector databases · 8
• 🛡️ Sandboxes · 8
• 🕸️ Data and scraping for AI · 3

🧑‍💻 Apps you run

🤖 Assistants

Personal AI assistants you run yourself and talk to through chat apps, with memory and the ability to act. _{12 ranked · 📝 all reviews}

nanobot leads with 84, ahead of OpenClaw (79) and Hermes Agent (76).

_{Top 10 of 12. See all 12, ranked, with reviews →}

💡 How to choose

• Check which channels it supports today (WhatsApp, Telegram, Slack, Discord, email) and whether each needs a paid API.
• Look at how memory is stored and whether you can inspect or wipe it.
• Actions need credentials; prefer assistants that scope them per tool.

↑ categories

💬 Chat UIs

Web front-ends for local or API models, usually with user accounts, chat history and file upload. _{13 ranked · 📝 all reviews}

LobeHub leads with 80, ahead of AnythingLLM (77) and Open WebUI (76).

_{Top 10 of 13. See all 13, ranked, with reviews →}

💡 How to choose

• Confirm it talks to your backend (Ollama, OpenAI-compatible, Anthropic) without a plugin.
• Multi-user auth, RBAC and SSO are where free and paid editions differ most.
• Check what the default compose pulls in (database, vector store) before sizing the host.

↑ categories

🧩 Agent platforms

Visual or code-first builders for agents and workflows, with orchestration, tools and deployment. _{10 ranked · 📝 all reviews}

Langflow leads with 73, ahead of Dify (72) and AutoGPT (70).

💡 How to choose

• Decide between a visual builder (faster to start) and code-first (easier to test and version).
• Check how workflows are exported; vendor-specific JSON makes migration costly.
• Look at the license for the server part; several are AGPL or source-available.

↑ categories

🔁 Workflow automation with AI

Workflow and integration platforms with AI steps or agent nodes, for wiring models into the rest of your tools. _{2 ranked · 📝 all reviews}

n8n leads with 76, ahead of Activepieces (65).

💡 How to choose

• Count the integrations you actually need; the long tail is where platforms differ.
• Check whether AI steps can use your own model endpoint or only hosted providers.
• Read the license: several are fair-code or source-available, not OSI open source.

↑ categories

📚 RAG and knowledge

Document Q&A, knowledge bases and enterprise search over your own files and data. _{15 ranked · 📝 all reviews}

LightRAG leads with 76, ahead of Open Notebook (75) and RAGFlow (72).

_{Top 10 of 15. See all 15, ranked, with reviews →}

💡 How to choose

• Match the ingestion formats you need (PDF with tables, Office, web, Confluence) before anything else.
• Check which embedding models and vector stores are supported and whether they can run offline.
• Look for citations in answers; without them, RAG output is hard to trust.

↑ categories

🔎 Search

Private search engines and AI answer engines that keep queries on your host. _{5 ranked · 📝 all reviews}

Morphic leads with 70, ahead of Local Deep Research (65) and GPT Researcher (64).

💡 How to choose

• Metasearch engines need upstream providers; check rate limits and whether results are cached.
• Answer engines call an LLM per query; budget tokens before exposing them to a team.
• Check how results are attributed; answer engines without citations are hard to verify.

↑ categories

🎙️ Voice

Speech-to-text, text-to-speech, voice agents and meeting tools that run locally. _{11 ranked · 📝 all reviews}

Voicebox leads with 72, ahead of Speech-to-Speech (67) and Pocket TTS (64).

_{Top 10 of 11. See all 11, ranked, with reviews →}

💡 How to choose

• Latency decides usability for voice agents; check the end-to-end numbers the project publishes.
• Most quality models need a GPU; CPU-only setups are slower and limited to smaller models.
• Check language coverage and whether models are downloaded at build or at first run.

↑ categories

🎨 Image and video

Generation UIs and pipelines for images and video, usually around diffusion models. _{8 ranked · 📝 all reviews}

ComfyUI leads with 75, ahead of MoneyPrinterTurbo (69) and InvokeAI (62).

💡 How to choose

• A GPU with enough VRAM is the hard requirement; check the minimum the project states.
• Node-based UIs are flexible but have a learning curve; form-based UIs are quicker to use.
• Check the model licensing separately from the app licensing.

↑ categories

🔐 Security

AI agents and tools for penetration testing, red-teaming and finding vulnerabilities in your own apps and models. _{1 ranked · 📝 all reviews}

Strix leads with 69.

💡 How to choose

• Run them only against targets you own or are authorized to test; check the tool's scope controls.
• Check which model backends are supported and whether findings stay on your host.
• Look at how exploitation is sandboxed (containers, VMs) before pointing it at production.

↑ categories

🧱 Build with it

🗂️ Memory

Long-term memory engines that store and retrieve facts for agents across sessions. _{12 ranked · 📝 all reviews}

hindsight leads with 78, ahead of MemPalace (77) and agentmemory (75).

_{Top 10 of 12. See all 12, ranked, with reviews →}

💡 How to choose

• Check what gets stored (raw messages, extracted facts, graphs) and how it is retrieved.
• Multi-tenant isolation matters if several users or agents share one store.
• Look at the backing store (Postgres, a vector DB, a graph DB) and whether you already run it.

↑ categories

🧠 Model serving

Inference engines and model servers that expose local models over an API. _{20 ranked · 📝 all reviews}

LocalAI leads with 82, ahead of llama.cpp (79) and vLLM (75).

_{Top 10 of 20. See all 20, ranked, with reviews →}

💡 How to choose

• Pick by hardware first (CPU, NVIDIA, AMD, Apple Silicon) and by model format (GGUF, safetensors).
• Throughput engines (continuous batching, paged attention) need GPUs; single-user servers do not.
• An OpenAI-compatible API keeps the rest of your stack portable.

↑ categories

🔀 Gateways

LLM gateways and proxies for routing, caching, rate limits and cost control across providers. _{14 ranked · 📝 all reviews}

OmniRoute leads with 86, ahead of LiteLLM (84) and freellmapi (76).

_{Top 10 of 14. See all 14, ranked, with reviews →}

💡 How to choose

• List the providers you call today; check the gateway's native support rather than generic passthrough.
• Key management, per-team budgets and audit logs separate gateways from simple proxies.
• Check latency overhead and whether streaming is passed through unchanged.

↑ categories

📈 Observability

Tracing, evaluation and prompt management for LLM applications. _{13 ranked · 📝 all reviews}

Langfuse leads with 74, ahead of Phoenix (74) and promptfoo (71).

_{Top 10 of 13. See all 13, ranked, with reviews →}

💡 How to choose

• Confirm SDK support for your framework (OpenTelemetry, LangChain, OpenAI SDK) and your language.
• Evaluation features vary widely; check whether evals run on your data without a cloud account.
• Retention and storage backend decide the host size; traces grow fast.

↑ categories

🧮 Vector databases

Vector stores and hybrid search engines for embeddings. _{8 ranked · 📝 all reviews}

Milvus leads with 81, ahead of Meilisearch (69) and Chroma (64).

💡 How to choose

• If you already run Postgres, check pgvector-based options before adding a new database.
• Hybrid search (keyword + vector) and filtering are where engines differ most.
• Check memory use per million vectors; it decides the host size.

↑ categories

🛡️ Sandboxes

Isolated runtimes where agents execute code, browse or use tools safely. _{8 ranked · 📝 all reviews}

Lightpanda leads with 78, ahead of Obscura (71) and NemoClaw (71).

💡 How to choose

• Isolation level (container, microVM, gVisor) sets the risk you accept when agents run code.
• Check startup time per sandbox; slow starts limit agent loops.
• Look at what is persisted between runs and how secrets are injected.

↑ categories

🕸️ Data and scraping for AI

Tools that turn web pages, PDFs and documents into clean text or structured data that models can use. _{3 ranked · 📝 all reviews}

Crawl4AI leads with 72, ahead of Firecrawl (68) and Jina Reader (44).

💡 How to choose

• Test on your hardest pages (JavaScript-heavy, tables, scans) before choosing.
• Check whether it runs fully self-hosted or needs the vendor's API for parts.
• Look at output formats: markdown, JSON with a schema, or both.

↑ categories

🏅 How we rank

Every project gets a score out of 100 from the signals below. A signal we cannot measure yet is left out and its weight goes to the others, so nobody loses points for things we have not checked. The score sets the order; 🥇🥈🥉 mark the top three places in each table.

_{Penalties: no license −10 · data must leave your host −10 · star-growth anomaly −15 · unpatched security advisory −20. Archived projects are removed. Paid placements never touch the score.}

_{The weights are our judgment, not science. Think one is wrong? Tell us. Method changes: 2026-10-09: Docker alone earns no points, medals mark places 1–3 instead of score bands.}

🧠 How this works

• 🧠 On this list: open-source projects whose main job is AI, from apps you run to libraries you build with. Tools for coding with AI, starters and the stack you ship on live on the sister list, Best of Vibe Coding.
• 🔎 Found in curated lists, app stores and template galleries, then fact-checked on GitHub: stars, last commit, license, Docker files.
• ✍️ Written from the README, never copied from other lists: what it does, what it needs, strengths and weaknesses as claims you can check. Specs say unknown rather than guess.
• 🔄 Kept fresh: entries are rewritten when the README or the latest release changes; projects quiet for 12 months get marked stale, archived ones drop off.
• 🚫 No pay-to-rank. Sponsors and affiliate links, if any, are labeled and never touch the order.

📬 Your turn

➕ Add a project · 🛠️ Fix wrong data · 🚪 Opt out (done within 24 hours). Everything here is generated, so use the forms instead of editing files.

📚 Where candidates come from

These lists, app stores and galleries (facts and links only, no text copied), plus community submissions: awesome-selfhosted · awesome-local-llm · awesome-homelab · self-hosting-guide · awesome-ai-agent-platforms · awesome-openclaw · umbrel-apps · runtipi-appstore · casaos-appstore · dokploy-templates · gh-search-self-hosted-ai · gh-search-ai-agents · gh-search-llm-apps · gh-search-ai-security · awesome-opensource-ai · awesome-llm-services · awesome-llmops · awesome-private-ai · awesome-local-llms · awesome-llm-webapps · 0xsojalsec-llms-local · inftyai-awesome-llmops · bradagi-awesome-cli-coding-agents · best-of-ml-python · awesome-generative-ai · awesome-ai-agents-e2b · awesome-mcp-servers · gh-search-llm-top · gh-search-ml-top · gh-search-agents-top · gh-search-mcp-top.

📄 License

Data (data/, this README, reviews/) is CC BY 4.0; see LICENSE-DATA. Code is MIT; see LICENSE. Project names and descriptions belong to their owners.

_{Curated with care, powered by public GitHub data, funded by freelance gigs. Found a gem? Tips keep the servers happy, but rankings stay honest. 🚀
For freelance collabs or coffee-fueled conversations, say hello on LinkedIn.}

_{Made with ☕ by Archestack · see also Best of Vibe Coding}