Skip to content

Continuum v0.2.3

Latest
Latest

Choose a tag to compare

View all tags
@bhavik-shyftlabs bhavik-shyftlabs released this 10 Jun 21:35
v0.2.3
0561a0e

🛠️ Fixed in 0.2.3

The published wheel now ships the continuum CLI. 0.2.2 on PyPI was built from a commit that predated the CLI, so it installed without the continuum command. After pip install shyftlabs-continuum you now get continuum up / status / logs / down exactly as the README describes.

The agent runtime for builders who ship. Build, run, and deploy reliable AI agents at enterprise scale — multi-LLM routing, persistent memory, MCP-native tools, durable workflows, and full observability, behind one small, type-safe API.

PyPI
Python 3.13+
License

📦 Install

pip install shyftlabs-continuum

Optional extras: pip install "shyftlabs-continuum[temporal]" (durable workflows), "[eval]" (evaluation), "[embeddings]" (local embeddings).

🚀 Quick start

Requirements: Python 3.13+ and Docker (for Redis · Qdrant/Milvus · Langfuse).

python3.13 -m venv .venv && source .venv/bin/activate
pip install shyftlabs-continuum

continuum up                          # start local infra (Redis + Qdrant); writes ./.env
echo "OPENAI_API_KEY=sk-…" >> .env     # add your provider key(s)

continuum up ships with the package and starts the bundled Docker stack for you. Pick a bigger profile with continuum up standard (adds Langfuse tracing) or continuum up full (adds Temporal + Milvus).

import asyncio
from continuum.agent import BaseAgent, AgentRunner

async def main():
    agent = BaseAgent(
        name="hello-agent",
        instructions="You are a friendly assistant.",
        model="gpt-4o-mini",
    )
    runner = AgentRunner()
    response = await runner.run(agent, "Hi!")
    print(response.content)

asyncio.run(main())

✨ What's inside

  • 🤖 Agentic core & orchestration — a strongly-typed agent primitive with lifecycle hooks, schema-validated structured outputs, and nine composable multi-agent patterns (sequential, parallel, loop, routing, planning, reflection, debate, scatter, supervised).
  • 🔀 Smart Inference — cost-aware routing that classifies each request and dispatches it to the cheapest capable model, with cross-provider failover and zero lock-in.
  • 🧠 Stateful memory — persistent long-term recall plus low-latency working memory, with multi-tenant isolation scopes and built-in PII redaction.
  • 🔌 Open tool calling — Model Context Protocol (MCP) across multiple transports, with capability scoping, context capture/injection, and generative-UI artifacts.
  • 🔁 Durable execution — crash- and restart-safe workflows with human-in-the-loop approval gates.
  • 🔭 Full observability — distributed tracing, token/latency/error telemetry, and one-line instrumentation.
  • 🌐 Model-agnostic — target frontier or open-weight models through a single model string; swap providers without touching agent code.
  • 🤝 Multi-agent handoffs — context-preserving delegation with history summarization, cycle detection, and depth control.
  • 📡 Real-time streaming — token-, tool-, handoff-, and memory-level events as they happen.
  • ✅ Built-in evaluation — turn production traces into golden datasets and regression-test agent quality.

🔗 Links

Built by ShyftLabs. Licensed under Apache-2.0.

Assets 2
Loading