pareta

Python client for Pareta — deploy open-weights endpoints, run metered inference, browse the benchmark catalog, and eval models on your own data.

pip install pareta        # or: uv add pareta / poetry add pareta

from pareta import Pareta

pa = Pareta.from_env()                       # reads PARETA_API_KEY
# or: Pareta(api_key="pareta_sk_…", base_url="https://api.pareta.ai")

# OpenAI-compatible inference against a deployed endpoint
resp = pa.chat.completions.create(
    model="ep_…",                            # an endpoint id (see pa.models.list())
    messages=[{"role": "user", "content": "Extract the total from this invoice: …"}],
)
print(resp.choices[0].message.content)

# Streaming
for chunk in pa.chat.completions.create(model="ep_…", messages=[...], stream=True):
    print(chunk.choices[0].delta.content or "", end="")

# List the models (endpoints) your org can call
for m in pa.models.list():
    print(m.id)

Async mirrors the sync client:

from pareta import AsyncPareta

async with AsyncPareta.from_env() as pa:
    resp = await pa.chat.completions.create(model="ep_…", messages=[...])

Auth

Mint a pareta_sk_ key in the dashboard (key management is browser-only) and pass it as api_key= or via PARETA_API_KEY. The SDK only ever consumes a key; it never creates, lists, or revokes them.

Inference is OpenAI-compatible

You don't even need this SDK to call a deployed endpoint — point the openai client at base_url + your key:

from openai import OpenAI
client = OpenAI(api_key="pareta_sk_…", base_url="https://api.pareta.ai/v1")

This SDK's unique value is the control plane: deploy, operate, and eval models from code — available both as Python methods and via the two interfaces below.

CLI

pip install "pareta[cli]" adds the pareta command — the same control plane from your shell:

export PARETA_API_KEY=pareta_sk_…

pareta tasks match "extract fields from invoices"     # intent → task
pareta tasks leaderboard invoice-extraction           # ranked open models + savings
pareta endpoints deploy --task invoice-extraction --wait
pareta endpoints list
pareta chat ep_… "Summarize this contract: …"          # prompt arg or piped stdin
pareta endpoints cost ep_…

Add --json to any command for machine-readable output; pareta --help (or pareta <group> --help) documents the full tree — tasks, models, endpoints, evals, chat, audio.

MCP server

pip install "pareta[mcp]" adds pareta-mcp, a Model Context Protocol server that exposes Pareta to an AI agent (Claude Desktop, Cursor, …) as tools — so the agent can find the best open model for a task, benchmark it on your data, and deploy it. Register it (Claude Desktop → Settings → Developer → Edit Config):

{
  "mcpServers": {
    "pareta": {
      "command": "pareta-mcp",
      "env": { "PARETA_API_KEY": "pareta_sk_…" }
    }
  }
}

It exposes the full surface — discovery (match_task, get_leaderboard, …), provisioning (deploy_endpoint, start / stop / delete), eval (run_eval), and metered chat / transcribe / speak. Provisioning and inference tools spend money; your MCP client's per-tool-call approval is the guardrail.

Errors

All errors subclass pareta.ParetaError:

Idempotent GETs and 429/5xx/timeouts are retried with exponential backoff (max_retries, default 2).

Status

Live: the full control plane — chat, models, tasks (browse + match), endpoints (deploy / operate / metrics), evals (bring-your-own-data), and audio — plus two interfaces over it: the pareta CLI (pip install "pareta[cli]") and the pareta-mcp MCP server (pip install "pareta[mcp]"). Sync + async clients.