Skip to main content

A production-ready Python SDK for running paid AI agents on your marketplace

Project description

orca-agent-sdk

A production-ready Python SDK for running paid AI agents on your marketplace with a single, consistent protocol.

orca-agent-sdk abstracts away:

  • HTTP server and routing
  • Persistent job storage (SQLite)
  • Algorand unsigned transaction generation
  • On-chain payment verification
  • Access token–gated result delivery
  • Background job execution

Agent developers implement one handler function and a small configuration object. Your marketplace controls the payment and verification flow.


1. Primary use case

Enable third-party developers to publish AI agents to your marketplace with:

  • No direct exposure to:
    • Flask or HTTP internals
    • Database schema or SQL
    • Algorand SDK details or on-chain verification logic
  • A standard, enforceable protocol:
    • Jobs are created via /start_job
    • Payments are enforced on Algorand
    • Results are released only after verified payment via /submit_payment
    • Access to outputs is controlled with per-job access tokens

This ensures:

  • Consistent integration across all agents
  • Security and integrity of the payment flow
  • Minimal onboarding friction for developers

2. Installation

For MVP and testing (TestPyPI):

pip install -i https://test.pypi.org/simple/ orca-agent-sdk

(Once promoted to PyPI:)

pip install orca-agent-sdk

Requirements:

  • Python 3.10+
  • Network access to Algorand nodes (defaults to Algonode public endpoints)

3. Quick start for agent developers

Agent developers define:

  • handle_task(job_input: str) -> str
  • AgentConfig with identity and pricing
  • Bootstrapping using AgentServer

Example:

from orca_agent_sdk import AgentConfig, AgentServer

def handle_task(job_input: str) -> str:
    # Implement your agent's core logic here
    # e.g., call an LLM, tools, vector search, etc.
    return f"Echo: {job_input}"

if __name__ == "__main__":
    config = AgentConfig(
        agent_id="my-agent-id",               # Unique identifier used by the marketplace
        receiver_address="MY_ALGO_ADDRESS",  # Algorand address to receive payments
        price_microalgos=1_000_000,          # Price per job in microAlgos
        # Optionally:
        # app_id=YOUR_MARKETPLACE_APP_ID,
        # algod_url="https://testnet-api.algonode.cloud",
        # indexer_url="https://testnet-idx.algonode.cloud",
    )

    AgentServer(config=config, handler=handle_task).run()

When this script runs, the SDK exposes the standardized API surface required by your marketplace.


4. Protocol overview (for marketplace integrators)

The SDK exposes three key endpoints.

4.1 Create job: POST /start_job

Request:

{
  "sender_address": "USER_ALGO_ADDRESS",
  "job_input": "User request or prompt"
}

Response (example):

{
  "job_id": "J1762754481290",
  "unsigned_group_txns": ["...base64_txn_1...", "...base64_txn_2..."],
  "txn_ids": ["EXPECTED_TXID_1", "EXPECTED_TXID_2"],
  "payment_required": 1000000
}

Responsibilities:

  • SDK:

    • Persists the job.
    • Constructs an Algorand transaction group:
      • Payment from sender_address to receiver_address for price_microalgos.
      • Application call to app_id using the configured ABI method.
    • Returns unsigned transactions for the client to sign.
  • Marketplace / client:

    • Requests unsigned_group_txns.
    • Presents them to the user’s wallet for signing and broadcasting.

4.2 Verify payment: POST /submit_payment

After broadcasting the signed transaction group, call:

Request:

{
  "job_id": "J1762754481290",
  "txid": [
    "REAL_TXID_1",
    "REAL_TXID_2"
  ]
}

SDK behavior:

  • Fetches on-chain transaction data via the configured Indexer.
  • Verifies:
    • The provided txids match what was generated for that job.
    • Sender matches the original sender_address.
    • Payment receiver matches receiver_address.
    • Amount equals price_microalgos.
    • Application ID matches app_id.
  • On success:
    • Marks the job as running.
    • Issues an access_token bound to this job.
    • Executes handler(job_input) asynchronously in the background.

Response (example):

{
  "status": "success",
  "message": "Payment verified, job started",
  "access_token": "ACCESS_TOKEN_VALUE"
}

If verification fails, the SDK returns an error and does not execute the handler.


4.3 Retrieve result: GET /job/<job_id>

Without access token (public view):

{
  "job_id": "J1762754481290",
  "status": "running",
  "created_at": 1762754481,
  "output": null
}

With valid access_token:

GET /job/J1762754481290?access_token=ACCESS_TOKEN_VALUE

{
  "job_id": "J1762754481290",
  "status": "succeeded",
  "created_at": 1762754481,
  "completed_at": 1762754504,
  "output": "Echo: Hello from local SDK test"
}

Design guarantees:

  • Output is only available to parties who can present the correct access token.
  • External observers can see job status but not sensitive results.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

orca_agent_sdk-0.1.5.tar.gz (11.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

orca_agent_sdk-0.1.5-py3-none-any.whl (10.2 kB view details)

Uploaded Python 3

File details

Details for the file orca_agent_sdk-0.1.5.tar.gz.

File metadata

  • Download URL: orca_agent_sdk-0.1.5.tar.gz
  • Upload date:
  • Size: 11.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.9

File hashes

Hashes for orca_agent_sdk-0.1.5.tar.gz
Algorithm Hash digest
SHA256 5e566b17968dac22bc88f7faef6ca34356899c9f27c8a3400c034e6055d5c794
MD5 a49909a838a4bf2b3b26b54adf22293c
BLAKE2b-256 853d90b086d876f8d014100cc3de89af03a83f026f458f883b0ac023e50dcae4

See more details on using hashes here.

File details

Details for the file orca_agent_sdk-0.1.5-py3-none-any.whl.

File metadata

  • Download URL: orca_agent_sdk-0.1.5-py3-none-any.whl
  • Upload date:
  • Size: 10.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.9

File hashes

Hashes for orca_agent_sdk-0.1.5-py3-none-any.whl
Algorithm Hash digest
SHA256 d371f1f6eab398aaf544a7cb52a0b45d712433ca81cf5457ae510c6b6c7cec7e
MD5 e85af048794af6bdae7838bb020b33b2
BLAKE2b-256 11022ee64c6152ecfd8766a1ffc9e465eba837f5af4066679fa6662f613731e7

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page