mirror of
https://github.com/katanemo/plano.git
synced 2026-06-08 14:55:14 +02:00
Update README with clearer project explanation and quick start guide
This commit is contained in:
parent
b71a555f19
commit
50a0bc9b1e
1 changed files with 82 additions and 0 deletions
82
README.md
82
README.md
|
|
@ -19,6 +19,88 @@
|
|||
Star ⭐️ the repo if you found Plano useful — new releases and updates land here first.
|
||||
</div>
|
||||
|
||||
# What is Plano?
|
||||
|
||||
Plano is an **AI-native proxy server and data plane** designed specifically for agentic applications. Instead of embedding routing, orchestration, safety, and observability code directly into your application, Plano externalizes these concerns into a centralized, high-performance data plane that sits between your agents and your users.
|
||||
|
||||
## Key Capabilities
|
||||
|
||||
| Capability | Description |
|
||||
|-----------|-------------|
|
||||
| **🚦 Agent Orchestration** | Intelligently route requests between multiple agents based on intent, with low-latency decision making powered by purpose-built LLMs |
|
||||
| **🔗 Smart LLM Routing** | Route to any model by name, alias, or automatic preference-based selection across 200+ LLMs from 50+ providers |
|
||||
| **🕵 Agentic Signals™** | Automatically capture rich signals and OpenTelemetry traces across every agent interaction — zero instrumentation required |
|
||||
| **🛡️ Guardrails & Safety** | Apply consistent moderation, jailbreak protection, and policy enforcement via configurable filter chains |
|
||||
| **🧠 Memory & Context** | Add memory hooks and context management consistently across all your agents |
|
||||
| **📊 Observability** | Get end-to-end tracing, metrics, and logs out of the box for continuous improvement |
|
||||
|
||||
## Why Plano?
|
||||
|
||||
Traditional agentic applications require you to build custom routing logic, handle multiple LLM provider APIs, implement tracing and observability, and add safety filters — all within your application code. This creates tight coupling and makes it hard to iterate.
|
||||
|
||||
**Plano decouples these concerns:**
|
||||
- ✅ **Language agnostic** — Use any programming language or framework (Python, Node.js, Go, Rust, etc.)
|
||||
- ✅ **Framework agnostic** — Works with LangChain, LlamaIndex, AutoGen, or raw HTTP servers
|
||||
- ✅ **Model agnostic** — Switch between OpenAI, Anthropic, Google, or self-hosted models without code changes
|
||||
- ✅ **Production-ready** — Built on Envoy proxy, battle-tested at scale
|
||||
|
||||
---
|
||||
|
||||
# Quick Start
|
||||
|
||||
Get Plano running in under 5 minutes:
|
||||
|
||||
## 1. Install Plano
|
||||
|
||||
```bash
|
||||
# macOS (Homebrew)
|
||||
brew tap katanemo/plano
|
||||
brew install plano
|
||||
|
||||
# Linux (from releases)
|
||||
curl -fsSL https://planoai.dev/install.sh | sh
|
||||
|
||||
# Docker
|
||||
docker run -p 12001:12001 katanemo/plano:latest
|
||||
```
|
||||
|
||||
## 2. Create a Configuration File
|
||||
|
||||
Create `config.yaml`:
|
||||
```yaml
|
||||
version: v0.3.0
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
default: true
|
||||
|
||||
listeners:
|
||||
- type: llm
|
||||
port: 12001
|
||||
```
|
||||
|
||||
## 3. Start Plano
|
||||
|
||||
```bash
|
||||
planoai up config.yaml
|
||||
```
|
||||
|
||||
## 4. Make Your First Request
|
||||
|
||||
```bash
|
||||
curl http://localhost:12001/v1/chat/completions \
|
||||
-H "Content-Type: application/json" \
|
||||
-d '{
|
||||
"model": "openai/gpt-4o",
|
||||
"messages": [{"role": "user", "content": "Hello, Plano!"}]
|
||||
}'
|
||||
```
|
||||
|
||||
**Next steps:** Check out the [Build Agentic Apps with Plano](#build-agentic-apps-with-plano) section below for a complete multi-agent example, or visit our [documentation](https://docs.planoai.dev).
|
||||
|
||||
---
|
||||
|
||||
# Overview
|
||||
Building agentic demos is easy. Shipping agentic applications safely, reliably, and repeatably to production is hard. After the thrill of a quick hack, you end up building the “hidden middleware” to reach production: routing logic to reach the right agent, guardrail hooks for safety and moderation, evaluation and observability glue for continuous learning, and model/provider quirks scattered across frameworks and application code.
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue