Papayya

Durable background jobs for AI agents. Bring your own LLM — Papayya handles execution, checkpointing, and deployment. Budget enforcement and cost tracking live on the cloud runtime.

Install

pip install papayya

Try it in 30 seconds (no LLM key needed)

papayya init                                    # writes papayya.yaml
papayya example                                 # scaffolds local_demo_agent.py
python local_demo_agent.py                      # one keyless durable run
papayya dev                                     # open the local dashboard

The demo agent runs a two-step durable workflow against canned data — no provider key, no network. Open papayya dev to see the run, the per-step input/output, and the lineage. That's the iteration loop.

Quick Start with your own LLM

Define an agent

from papayya import agent, tool

@tool
def search_web(query: str) -> str:
    """Search the web for information."""
    # Your implementation here
    return "..."

@agent(name="research-bot", model="gpt-4o-mini", budget_usd=1.0)
def research_bot(input_data):
    from openai import OpenAI
    client = OpenAI()
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": input_data}],
    )
    return response.choices[0].message.content

budget_usd on the @agent decorator is metadata the cloud runtime uses to cap per-run spend. It is not enforced when you call the function directly on your laptop.

Durable execution

Wrap long-running work in checkpoint-able steps. If a run crashes, it resumes from the last checkpoint instead of re-executing completed steps.

from papayya import papayya

run = papayya().run("my-agent")

fetch    = run.step("fetch",     fetch_data)
analyse  = run.step("analyse",   analyse_results)

data    = fetch(query)        # cached on replay
summary = analyse(data)       # cached on replay

run.complete(summary)

step() and task() are aliases. Use whichever reads better in context.

When PAPAYYA_API_KEY is set, checkpoints round-trip through the cloud control plane. Without a key, the SDK writes to a local SQLite at .papayya/local.db — the same database papayya dev reads from.

Budget enforcement (cloud only)

Budget caps and cost tracking are enforced by the runtime shim when your agent runs in a Papayya container. Set a cap on the @agent decorator (shown above) or pass budget_cents when triggering a run via the API. There is no local budget API — local PapayyaRun is durability-only.

Deploy

papayya login
papayya deploy

Key Concepts

• BYOF (Bring Your Own Function) — Papayya doesn't wrap your LLM calls. You use any SDK (OpenAI, Anthropic, Bedrock, etc.) directly inside your agent function.
• @agent decorator — Registers your function for deployment. The function stays callable locally.
• @tool decorator — Defines tools your agent can call, with automatic JSON Schema generation from type hints.
• Durable runs — Checkpoint-and-replay execution. Steps are cached so replayed runs skip completed work.
• Local dashboard — papayya dev reads from the same SQLite the SDK writes to. No control plane required.
• Budgets — Cloud-only. The runtime shim reserves cost before each LLM call and pauses the run when the cap is hit.

Requirements

• Python 3.10+

License

MIT