peaq-os-sdk

Python SDK for the peaqOS protocol. Provides a typed wrapper around the peaq on-chain capabilities so integrators can onboard machines, submit events, mint NFTs, and query credit ratings without writing raw web3.py boilerplate.

Tokenomics 2.0 machine-ID compatibility

Tokenomics machine IDs are non-boolean Python int values in the full range 0 <= machine_id < 2**256. At JSON, URL, log, or other text boundaries they must be canonical base-10 strings. ID size never selects a mode: a non-None client.tokenomics20 configuration explicitly selects Tokenomics mode.

from peaq_os_sdk import (
    parse_tokenomics_machine_id,
    serialize_tokenomics_machine_id,
    validate_tokenomics_machine_id,
)

machine_id = parse_tokenomics_machine_id("9223372036854775808")
validate_tokenomics_machine_id(machine_id)
assert serialize_tokenomics_machine_id(machine_id) == "9223372036854775808"

Still-disabled surfaces raise TokenomicsIntegrationUnavailableError with code TOKENOMICS_INTEGRATION_UNAVAILABLE, plus .integration and .owning_ticket, at their first local boundary. Select the mode with Tokenomics20Config(deployment_id=...), or TOKENOMICS_DEPLOYMENT_ID for PeaqosClient.from_env(). See MCR queries, events, and the machine-ID inventory for every audited field and owner.

Tokenomics 2.0 machine activation

On peaq, one transaction mints the machine NFT, stores its DID document, activates its subscription, and registers its home chain — replacing the deprecated register_machine() + mint_nft() pair.

Activation is opt-in. Supplying tokenomics20 selects a deployment by ID from the SDK's approved snapshot; contract addresses are never passed by callers. A client without it keeps its existing behaviour unchanged.

from peaq_os_sdk import PeaqosClient, Tokenomics20Config

client = PeaqosClient(
    rpc_url="https://peaq-agung.api.onfinality.io/public",
    private_key=os.environ["PRIVATE_KEY"],
    # ... existing contract addresses ...
    tokenomics20=Tokenomics20Config(deployment_id="agung-2026-08-28"),
)

preview = client.preview_machine_activation(params)   # no signing, no writes
result = client.activate_machine(params)              # one atomic transaction
print(result.machine_id, result.net_peaq_amount)

Selecting a deployment also disables the deprecated registration, mint, and bridge entry points, which then raise TokenomicsUnsupportedError locally — before any RPC, approval, or signature — rather than quietly doing the old thing. Outside Tokenomics mode they keep working and emit a DeprecationWarning.

See docs/18_TOKENOMICS_ACTIVATION.md for configuration, the preview/activate flow, the bond and voucher arithmetic, the full error taxonomy, reconciling an unconfirmed transaction, the machine-signed/operator-controlled migration path, and the release-day procedure for adding a network.

Tokenomics 2.0 machine management

After onboarding, the same deployment-selected client exposes aggregate state and availability reads, suspend/resume, PEAQ and USDT subscriptions, ERC-721 ownership, DID updates, side-effect-free action previews, and read-only reconciliation for a known transaction hash. Owner, controller, and payer authority remain distinct, and every write correlates contract-scoped events before reconciling post-state.

See docs/19_TOKENOMICS_MACHINE_MANAGEMENT.md for the authority matrix, bounded payment flows, migration notes, relocation-read semantics, and remaining release gates.

Solana onboarding and reads

The existing activation APIs support reservation-first Solana onboarding, keyless previews/status, staged submission and recovery. Native machine creation executes on Solana after its peaq reservation and subscription prerequisites. Existing EVM activation behavior is unchanged.

See the Solana onboarding guide for configuration, signer selection, examples, status and recovery, and migration guidance. See Solana reads for home resolution and native account observations.

peaq RPC: retries, rate limits and call counts

PeaqosClient talks to peaq through PeaqHTTPProvider (TICKET-TKNA-007):

• Chain id. It is asked once per client.
• Retries, reads only. Rate limits (HTTP 429, -32029) back off from 0.5 s, at most five attempts, honouring Retry-After. A transient failure is retried once after 1 s. Submissions are sent exactly once.
• Time budget. Solana onboarding operations bound every retry pause by their timeout_seconds.
• One block per read. A read operation pins one finalized peaq block.
• Log windows. History reads split eth_getLogs into windows of at most 5,000 blocks, so an endpoint that caps the range (QuickNode: 10,000) keeps working as machines age.
• Errors. A failed read names the request: TokenomicsActivationError.rpc_method and .rpc_host (host only, never the path or query where an API key lives). An exhausted rate limit is RPC_RATE_LIMITED.

To see how many calls an operation makes, enable DEBUG on the peaq_os_sdk.rpc logger (one rpc calls for <operation>: total=N … line per operation), or run the counting proxy:

python scripts/count_rpc_calls.py --upstream-env PEAQOS_RPC_URL --port 18545 --log rpc-calls.log

Point the client at http://127.0.0.1:18545, run the operation, then press Ctrl-C for per-method totals. The proxy never prints the upstream URL.

Solana machine management

Writes for Solana-homed machines go through the same entry points as their peaq equivalents — transfer_machine, set_machine_controller, the three DID setters, suspend_machine/resume_machine, preview_machine_action and reconcile_machine_management_transaction all take an optional solana=SolanaMachineWriteParams(...). A caller who passes none sees exactly the shipped EVM behaviour; no machine-keyed method gained a chain argument. set_machine_evm_operator is the one addition with no peaq sibling, because on peaq the owner already is the EVM address.

The signer is an argument rather than client configuration: Solana's key space and the EVM account's are unrelated, and the action is authorized as that signer's public key. Three ERC-721 operations — approve_machine, set_machine_approval_for_all, safe_transfer_machine — are refused by name with SOLANA_OPERATION_UNSUPPORTED before any RPC or signature, rather than rejected as unknown.

A transaction that landed and then disagreed comes back as a result carrying its conflicts, not as an exception, because a caller told only "mismatch" cannot tell that their own action succeeded. Confirmation requires the transaction, the action's own event from the expected program for this machine, and the complete expected post-state including the fields the action was not supposed to change. pending means "not yet known", never "failed".

Confirming locally never implies peaq has applied anything: set_machine_evm_operator returns the two in two fields, and synchronization is established only by peaq's own lastAppliedMirrorPushSeq watermark keyed by Solana's protocol chain ID (5) — not its LayerZero endpoint ID and not peaq's EVM chain ID. Matching operators, NothingToPush, a skipped send and a source sequence are each true of a push that was dropped in transit, so none of them counts. The SDK sends no cross-chain message and changes no bridge configuration.

See docs/21_SOLANA_MACHINE_MANAGEMENT.md for the authority table, the explicit clearing rules, DID rent direction, and how expired is the only reconciliation state that licenses a fresh attempt.

Two subscription accounts, selected by deployment record

Two generations of the Solana programs are live. Before CORE-778 they read a machine's subscription from the adapter-owned SubscriptionMirror; since CORE-778 they read the terminal-owned SubscriptionTerminal, which carries a status as well as a tier. The deployment record says which, through subscription_reader ("mirror" or "terminal") and machine_subscription_terminal. Only that field decides — not the packaged IDL's account name, not metadata.version (0.1.0 in both), and not whether either account exists. A "terminal" record with no terminal program ID fails at resolution with DEPLOYMENT_UNAVAILABLE, before any RPC, rather than falling back to an account the deployed programs no longer read.

The eligibility rules differ with it, and neither reads the tier: onboard_machine needs last_seq != 0 and a status of Active or Grace; submit_event needs the status alone. Tier 0 is Entry, a real purchasable tier — never read it as "no subscription". An absent terminal account is ineligible for both, not tier zero.

Status reads report both accounts and name the authoritative one in subscription_source; nothing merges them into a single tier. subscription_terminal is None means the account was not read, never that it is absent. The two last_seq values count different streams — the adapter's tier push and peaq's status signal — and are never comparable.

The shipped peaq-mainnet record reads "terminal" and names the deployed terminal program 4EFBHr9AyazPA6cxZRYYU5eK6uguogG5YhPQX6cs6A9d: the official Solana set has run the CORE-778 readers since the production upgrade of 2026-09-16. The mirror generation stays implemented for a record whose programs still run pre-CORE-778 builds; none is shipped. See docs/20_SOLANA_READS.md for the full field list, the readiness table, what the shipped record uses, and the migration notes for callers that construct deployment records.

Solana event submission

submit_event and batch_submit_events resolve each machine and write to the chain it is homed on. Callers never name a chain, and an EVM-only caller sees no change. The native path needs pip install peaq-os-sdk[solana].

from peaq_os_sdk import SubmitEventParams

# A Solana-homed machine: the same call, routed natively.
signature, data_hash = client.submit_event(
    machine_id=machine_id,          # full uint256, never narrowed
    event_type=0,                   # 0 revenue, 1 activity
    value=1_23,                     # subunit integer: US$1.23
    currency="USD",                 # omit for the same default
    timestamp=observed_at,          # checked against the cluster clock
    raw_data=payload,               # keccak256(payload); None means 32 zero bytes
    trust_level=0,
    source_chain_id=3338,           # provenance, not routing
    source_tx_hash=None,
    metadata=b"",                   # at most 256 bytes, refused not truncated
)

A batch is one transaction, which is what makes it atomic:

signatures = client.batch_submit_events([
    {**event_for(machine_a), "currency": "USD"},
    {**event_for(machine_a), "currency": "USD"},   # same log, consecutive indices
    {**event_for(machine_b), "currency": ""},      # activity event
])

Mixing peaq-homed and Solana-homed machines in one batch is refused before either chain is written. An oversized batch is refused rather than split, because a split batch would no longer be atomic.

The first event for a machine creates its log and the submitter pays the rent shortfall, once per machine however many events target it. Each intentional submission carries a random 128-bit ID in one SPL Memo instruction, which is what keeps two identical calls against one blockhash from collapsing into a single transaction. Retries reuse that ID and rebroadcast the same signed bytes; a timeout means unresolved, not failed.

Local confirmation proves the event was recorded on Solana. It does not prove that any downstream rating service has ingested it, that a score changed, or that peaq has synchronized. On chain the log keeps only its count and the most recent event, so there is no event enumeration and this SDK adds no history API.

See docs/03_EVENTS.md for the full gate list, the field differences from EVM, and what completion does and does not establish.

Install

pip install peaq-os-sdk

Dependencies installed automatically: web3, eth-account, requests, posthog.

For OWS wallet management (optional):

pip install peaq-os-sdk[ows]

For Solana transaction signing via OWS (optional):

pip install peaq-os-sdk[ows,solana]

For Stream buyer SPL token payments (optional):

pip install peaq-os-sdk[solana]

The solana extra installs solders (native SOL transfers and signing) and solana (provides spl.token for SPL transfers such as USDC).

For P2P stream delivery (optional):

pip install peaq-os-sdk[p2p]

The p2p extra enables machine-to-machine P2P delivery via the peaqos-p2p transport (P2PDeliveryChannel seller, P2PDeliveryReceiver buyer).

Solana signing from an OWS wallet

When open-wallet-standard and solders are installed, construct a standalone Solana signer from a vault wallet that has a Solana account:

from solders.hash import Hash
from solders.keypair import Keypair
from solders.message import Message
from solders.pubkey import Pubkey
from solders.system_program import TransferParams, transfer
from solders.transaction import Transaction

from peaq_os_sdk import PeaqosClient, SOLANA_MAINNET_CHAIN_ID

signer = PeaqosClient.solana_signer_from_wallet("my-wallet", "s3cret")
print(signer.address, signer.chain_id)  # chain_id == SOLANA_MAINNET_CHAIN_ID

from_pubkey = Pubkey.from_string(signer.address)
instruction = transfer(
    TransferParams(
        from_pubkey=from_pubkey,
        to_pubkey=Keypair().pubkey(),
        lamports=1_000,
    ),
)
message = Message.new_with_blockhash([instruction], from_pubkey, Hash.default())
tx = Transaction.new_unsigned(message)

signed = signer.sign_transaction(tx)
# broadcast with your Solana RPC client using signed.serialize()

By default (ows_signing=True), the ed25519 key is decrypted only per sign via OWS. Pass ows_signing=False to decrypt at construction and sign locally with solders. See docs/06_WALLET.md for details.

Requirements

• Python >= 3.10
• A peaq RPC endpoint
• A funded wallet for the proxy operator (or for the machine itself, in self-managed mode)

Quick start

from peaq_os_sdk import PeaqosClient

client = PeaqosClient(
    rpc_url="https://peaq.api.onfinality.io/public",
    private_key="0xYOUR_PRIVATE_KEY",
    identity_registry="0x...",
    identity_staking="0x...",
    event_registry="0x...",
    machine_nft="0x...",
    did_registry="0x...",
    batch_precompile="0x...",
)

print("signer address:", client.address)

PeaqosClient is the only class consumers instantiate. All feature methods hang off it. The constructor performs synchronous validation and wires up the underlying Web3 provider + signing account — no network I/O is issued at construction time.

Read More here >>

MCR queries

Use PeaqosQueryClient(deployment_id="peaq-mainnet") for HTTP-only MCR reads without RPC, contracts, keys or wallet configuration. The SDK selects the paired endpoint. Legacy callers explicitly select PeaqosQueryClient(legacy_api_url=...). All three methods reuse the existing typed validation, response correlation and pooled HTTP transport; full PeaqosClient query calls remain supported. See the query client guide for examples and mode-specific DID formats.

MCRResponse:

{
    "did": str,
    "machine_id": int,
    "mcr_score": int | None,    # 0–100, None if Provisioned
    "mcr": str,                 # "AAA" | "AA" | "A" | "BBB" | "BB" | "B" | "NR" | "Provisioned"
    "bond_status": str,         # "bonded" | "unbonded"
    "negative_flag": bool,      # active negative event flag
    "event_count": int,
    "revenue_event_count": int,
    "activity_event_count": int,
    "revenue_trend": str,       # "up" | "stable" | "down" | "insufficient"
    "total_revenue": float,
    "average_revenue_per_event": float,
    "last_updated": int | None,
}

OperatorMachinesResponse carries operator_did, a machines list of {did, machine_id, mcr_score, mcr, negative_flag} entries, and a pagination object. MachineProfileResponse returns structured NFT metadata with a validated peaqos sub-object. Every failure path is an ApiError with a stable .code (NOT_FOUND, SERVICE_UNAVAILABLE, SERVER_ERROR, HTTP_ERROR, BAD_RESPONSE, TIMEOUT, NETWORK_ERROR). See docs/03_QUERIES.md for the full endpoint and error reference.

Smart account deployment

ERC-4337 smart accounts are provisioned via the MachineAccountFactory contract using CREATE2 — the deployed address is deterministic from (owner, machine, daily_limit, salt), so the same tuple always resolves to the same address.

Because CREATE2 is deterministic, predicted == deployed for the same inputs — callers can preview the address, pre-fund it, or display it in a UI before paying gas. Configure the factory address via the optional machine_account_factory kwarg on PeaqosClient or the MACHINE_ACCOUNT_FACTORY_ADDRESS env var. See docs/04_SMART_ACCOUNTS.md for parameter rules and receipt-decoding error codes.

Cross-chain NFT bridging

Machine NFTs move between peaq and Base over LayerZero v2. Two chains are recognised today; direction is inferred from the source / destination arguments.

client.bridge_nft(...) estimates the LayerZero messaging fee via quoteSend before broadcasting and attaches that fee as msg.value on the actual send transaction so the message is correctly paid for. PeaqosClient.wait_for_bridge_arrival(...) is a static method (no client instance needed) that polls MachineNFT.ownerOf(token_id) on the destination every 10 seconds and returns True on arrival or False at timeout (default 300 s). See docs/05_BRIDGE.md for the full walkthrough including options handling, Base-source setup, and the complete error-code table.

Deprecated in Tokenomics mode. bridge_nft raises TokenomicsUnsupportedError (MACHINE_RELOCATION_UNAVAILABLE) when the client selected a Tokenomics deployment — before the adapter approval it would otherwise submit. Tokenomics 2.0 relocates the whole machine record rather than sending an ONFT, and relocation is not exposed in this release. wait_for_bridge_arrival takes no client and so cannot detect the mode; it is deprecated but not gated. See docs/18_TOKENOMICS_ACTIVATION.md.

Provider-node provisioning

The peaq_os_sdk.provisioning namespace is a generic, provider-agnostic runner that turns a schema-driven ProviderProvisioningManifest (node-provider.peaq.network/v1alpha1) into an ordered, auditable install. All provider-specific complexity lives in the manifest — the runner understands only the schema. It runs entirely on the machine and calls no external API.

from peaq_os_sdk import provisioning

manifest = provisioning.fetch_manifest("akash", "latest", repo_base_url="https://manifests.example")
inputs = provisioning.resolve_inputs(manifest, {"providerDomain": "node.example.com", ...})
executor = provisioning.LocalShellExecutor()

preflight = provisioning.run_preflight(manifest, executor)
preflight.raise_if_blocked()                       # gate: no phase starts while blocked

result = provisioning.provision(
    manifest, inputs, mode="manual", executor=executor,
    context={"machine_wallet_address": "0x..."},   # never the operator's
    on_owner_action=lambda e: input(e.instructions) or "confirmed",
)

verification = provisioning.verify_provisioning(
    manifest, executor, inputs=inputs.values, captures=result.captures
)
verification.raise_if_failed()                     # live only when all probes pass

• Manual vs auto modes are declared per step (manual / auto / both). Eligibility is asymmetric: a manual-only step still runs (confirmed) in an auto run, while an auto-only step is skipped in a manual run and left un-recorded for a later auto pass — so completed means this selected-mode pass finished. Auto mode requires a SudoGrant scoped to the manifest's allowedCommands; a sudo: required command outside the grant fails closed, and unmarked commands are never elevated.
• Owner-action handoffs (funding, DNS, SSH, signing) are real pause points — on_owner_action must return the step's expectedConfirmation; they are never auto-confirmed.
• Secrets never leak — and are never persisted. Secret inputs and secret captures are redacted before any output is emitted; a secret capture with no pattern withholds its step's raw stream entirely. Errors and resume state carry no secret material, and secret captures are never written to a resume state (not even encrypted).
• Idempotency & durable resume. A step's idempotency check short-circuits already-provisioned work. Every run emits a typed, JSON-serializable ResumeState (with a schema_version, completed phase/step ids, and resolved non-secret inputs) at each step boundary via on_state_change — and as ProvisionResult.resume_state / StepExecutionError.resume_state — which you pass back as resume_state= to skip completed steps and restore non-secret captures. A resumed step needing a non-persisted secret re-runs its producer when rerun: safe, else the caller can re-provide it via resume_secrets (manifest-declared secret keys only), else ResumeNotPossibleError.
• Deterministic manifest pinning. fetch_manifest(..., include_source=True) returns a FetchedManifest exposing the raw source bytes, so a caller can pin a stable sha256:<hex> digest.
• Verification, not exit codes, defines success — verify_provisioning succeeds only when every success.all_of probe passes (the caller passes inputs=inputs.values and a redaction registry seeded from inputs.secret_values()).

The runner is standalone; a CLI (or any other consumer) calls these functions. CLI-surface decisions (headless vs prompts) and provider manifest content (commission/API-key fields) are the consumer's and the manifest's concern, not the runner's. Remote execution is out of scope — only the default local-shell executor and a swappable local agent/session executor are supported.

Telemetry

The SDK collects anonymous, aggregate usage telemetry via PostHog to measure install counts, onboarding success rates, and MCR query volume. No PII is collected — GeoIP is disabled, IP addresses are nulled, and all UUIDs are randomly generated.

Telemetry is enabled by default and can be disabled via environment variables:

DO_NOT_TRACK=1 always takes precedence. When telemetry is disabled, no PostHog client is created and no events are sent.

Analytics events are captured automatically from existing SDK methods — no additional API calls are needed:

client = PeaqosClient(...)                         # peaqos_sdk_initialize (first install only)
addr, key = PeaqosClient.generate_keypair()        # peaqos_sdk_generate_keypair (via active instance)
machine_id = client.register_machine()             # peaqos_sdk_generate_wallet
tx = client.write_machine_did_attributes(...)      # peaqos_sdk_generate_did
tx = client.mint_nft(machine_id, addr)             # peaqos_sdk_generate_nft
mcr = client.query_mcr("did:peaq:0x...")           # peaqos_sdk_mcr_request
client.close()                                     # flush queued events

All analytics calls are fire-and-forget — they never block or raise. See docs/10_ANALYTICS.md for the full event reference and privacy details.

Documentation

Per-feature deep-dives live under docs/:

• Quick Start — client initialization, configuration, environment variables, types, exception classes, validation, utilities, and constants.
• Machine identity & registration — faucet 2FA enrollment, gas-station funding, self-managed and proxy-managed registration. Includes the full faucet error reference and the on-chain revert mapping.
• MCR queries — query_mcr, query_machine, query_operator_machines, response shapes, rating tiers, and the full HTTP error-code table.
• Smart account deployment — deploy_smart_account and get_smart_account_address, CREATE2 determinism, parameter rules, and receipt-decoding error codes.
• Cross-chain NFT bridging — bridge_nft direction handling (peaq ↔ Base via LayerZero v2), supported chain IDs and LayerZero EIDs, LayerZero fee estimation, and wait_for_bridge_arrival polling semantics.
• Provider-node provisioning — the generic, manifest-driven peaq_os_sdk.provisioning runner: fetch/validate a ProviderProvisioningManifest, resolve inputs, gate on pre-flight, execute phases/steps in manual/auto modes with scoped sudo and owner-action handoffs, redact secrets before emission, thread captures, verify success by probes, and resume interrupted runs. Covers the CommandExecutor protocol, machine-wallet context, idempotency, the error hierarchy, and the consumer/runner boundary.
• Monetisation opt-in — the Tokenomics 2.0 signed on/off toggle and public state read: deployment-selected MCR configuration, full uint256 IDs and decimal DIDs, owner/controller authorization, live compatibility checks, bounded retries, one end-to-end deadline, strict response correlation, typed failures, and migration from independent legacy state.
• Tokenomics 2.0 machine activation — atomic PEAQ activation: deployment selection, preview_machine_activation / activate_machine, the bond/voucher arithmetic and why only the net amount is approved, receipt event correlation and post-state reconciliation, reconcile_activation_transaction for a submitted-but-unconfirmed hash, the typed error taxonomy, migrating off register-then-mint, and the release-day diff for adding a network.
• Tokenomics 2.0 machine management — post-onboarding reads and previews, lifecycle, bounded PEAQ/USDT subscriptions, ERC-721 ownership, DID updates, relocation evidence, authority rules, and known-hash reconciliation.
• Solana machine management — native writes through the existing entry points: the authority table, why the signer is an argument, results that report conflicts instead of raising, keeping local confirmation apart from peaq application, explicit clearing, DID rent direction, and reconciling after a crash.
• Solana onboarding and recovery — reservation-first phases, keyless previews/status, signer selection, durable recovery and migration.
• Solana reads — home resolution, native identity and subscription observations, which of the two subscription accounts a deployment reads and the two eligibility rules, cluster verification and packaged bindings.
• MCR API compatibility — Tokenomics 2.0 and Solana-homed machines: the three breaking type changes, the additive home_chain / rating_unavailable / owner / controller fields, the distinct 400 / 401 / 409 / 503 codes on ApiError, and the planned owner-signed Solana monetization path (implemented, gated off until a compatible server release is recorded).
• Event submission — single and batch event submission, pipeline details, limits, hashing, metadata mode, and rate-limiting behavior.
• NFT minting & DID attributes — minting machine NFTs, querying token IDs, writing machine and proxy DID attributes atomically via the Batch precompile. Covers attribute key reference, data visibility options, and the atomic batch guarantee.
• OWS wallet lifecycle — create, import, list, get, export, and delete encrypted wallets via the Open Wallet Standard. Multi-chain accounts, passphrase management, vault storage, OWS-native EVM signing via from_wallet(), and Solana signing via solana_signer_from_wallet().
• Orchestration service (experimental) — client.orchestration, policies, observability, market lifecycle, pagination, challenge-sign workflows, planned type stubs, and live integration tests.
• Stream data signing & encryption (experimental) — local field-level privacy rules, EIP-191 data package signing, and verification (Phase 2 Step 2; no network/distribution).
• Stream data chunking (experimental) — per-chunk XChaCha20-Poly1305 encryption, inline Ed25519 signatures, owner/operator/machine key wrapping, and separate buyer access docs; build_chunk_chain / verify_chunk_chain / decrypt_chunk / create_buyer_access_entry / build_buyer_access_files.
• Analytics & telemetry — anonymous usage telemetry, event inventory, opt-out, privacy guarantees, and onboarding flow integration.
• Stream distribution (experimental) — seller-side payment confirmation and delivery (distribute_data, PollingConfirmationProvider, S3DeliveryChannel) plus buyer-side token payment (transfer_token, submit_payment_proof, pay_and_submit_proof) on peaq, Base, and Solana.
• P2P delivery (experimental) — machine-to-machine delivery via the peaqos-p2p transport: P2PDeliveryChannel (seller) and P2PDeliveryReceiver (buyer) with verify-before-decrypt chunk reception, session routing, and replay protection.
• Stream delivery setup (in development) — pre-purchase payment rails discovery, delivery transports discovery, delivery capability registration, and stream listing purchase field updates.
• Stream purchases (in development) — full purchase lifecycle: creation with buyer identity, payment intent, payment proof, delivery retrieval, and purchase events.

Development

git clone https://github.com/peaqnetwork/peaq-os-sdk-py.git
cd peaq-os-sdk-py
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Activate .env for Testing

set -a
source .env
set +a

Quality gates

ruff check src tests       # lint — zero warnings
black --check src tests    # formatting
mypy src                   # strict type check, zero errors
pytest -q                  # all green

The unit tests are hermetic — they mock Web3 and requests.Session with handwritten stubs and never touch the network.

The Tokenomics activation surface additionally carries a branch-coverage floor:

pytest -q -m "not integration" \
  --cov=src/peaq_os_sdk/tokenomics --cov-branch --cov-fail-under=90

Coverage is measured from the hermetic suite alone, so the floor is meaningful without spending PEAQ.

Optional integration tests

The repo ships an opt-in suite that talks to a real peaq devnet. It is skipped silently in normal pytest -q and only runs when the required environment variables are supplied.

Registration integration suite (tests/integration/registration/test_registration_integration.py) — three end-to-end tests for the self-managed flow, proxy-managed flow, and double-registration revert. Gated behind the integration marker plus seven environment variables:

PEAQOS_RPC_URL=https://peaq.api.onfinality.io/public
PEAQOS_PRIVATE_KEY=0xYOUR_FUNDED_TREASURY_KEY
IDENTITY_REGISTRY_ADDRESS=0xYOUR_DEPLOYED_REGISTRY_ADDRESS
IDENTITY_STAKING_ADDRESS=0xYOUR_DEPLOYED_STAKING_ADDRESS
PEAQOS_OWNER_ADDRESS=5GrwvaEF...
PEAQOS_FAUCET_URL=https://depinstation.peaq.xyz
PEAQOS_2FA_CODE=123456

pytest -m integration tests/integration/

See docs/02_REGISTRATION.md → Integration tests for the full breakdown of what each test verifies and operational warnings.

Tokenomics 2.0 query & event suites (tests/integration/tokenomics/) — MCR-20 coverage for the three enabled queries (lossless int machine IDs, identifier correlation, per-entry DID/ID agreement) and for Tokenomics event submission. Requires the eight base contract env vars plus:

PEAQOS_TOKENOMICS_TEST_MACHINE_ID=9007199254740992   # canonical decimal, registered on MachineRegistry
PEAQOS_TOKENOMICS_TEST_OPERATOR_DID=did:peaq:0x...   # queries only
PEAQOS_TOKENOMICS_EVENT_WRITES=1                     # events only — submits real transactions, spends gas

pytest -m integration tests/integration/tokenomics/

The query suite pins routing by building its client with a deliberately invalid api_url, so any fallback to the legacy endpoint fails loudly. agung is explicitly skipped with a reason — it has no approved paired MCR to test against.

NFT & DID integration suite (tests/integration/nft_did/test_nft_and_did_integration.py) — end-to-end tests for the full NFT lifecycle, atomic DID writes, and atomicity guarantee verification (all-or-nothing batch semantics). Includes 7 tests across minting, token queries, machine DID attributes, proxy DID attributes, and atomic revert scenarios. Requires the same env vars as the registration suite plus:

EVENT_REGISTRY_ADDRESS=0x...
MACHINE_NFT_ADDRESS=0x...
DID_REGISTRY_ADDRESS=0x0000000000000000000000000000000000000800
BATCH_PRECOMPILE_ADDRESS=0x0000000000000000000000000000000000000805

See docs/04_NFT_AND_DID.md for the full breakdown.

OWS wallet integration suite (tests/integration/wallet/test_wallet_integration.py) — six end-to-end tests for wallet lifecycle (create, list, get, delete), create-export-reimport round-trip, private key import, empty vault, and error paths. Gated behind PEAQOS_INTEGRATION=1 and requires the open-wallet-standard package:

pip install peaq-os-sdk[ows]
PEAQOS_INTEGRATION=1 pytest -m integration tests/integration/wallet/

See docs/06_WALLET.md for the full wallet API reference.

Orchestration live API suite (tests/integration/orchestration/test_orchestration_integration.py) — health, readiness, machine lifecycle, market lifecycle, policy CRUD, and audit-event listing against a real orchestration deployment. Gated behind PEAQOS_ORCHESTRATION_INTEGRATION=1:

PEAQOS_ORCHESTRATION_INTEGRATION=1
PEAQOS_ORCHESTRATION_URL=https://markets.peaq.xyz
PEAQOS_API_KEY=your-platform-api-key
pytest -m integration tests/integration/orchestration/test_orchestration_integration.py

Optional overrides: PEAQOS_ORCHESTRATION_PROVIDER_KEY, PEAQOS_ORCHESTRATION_ENDPOINT_URL, PEAQOS_ORCHESTRATION_MARKET_SERVICE_TYPE, PEAQOS_ORCHESTRATION_MARKET_OPERATION, PEAQOS_ORCHESTRATION_MARKET_ORDER_INPUT (JSON object), PEAQOS_ORCHESTRATION_MARKET_PAYMENT_RAIL. See docs/07_ORCHESTRATION.md → Integration tests.

A separate mocked market lifecycle test lives in tests/integration/orchestration/test_market_lifecycle_integration.py and runs without orchestration env vars.

Tokenomics integration suite (tests/integration/tokenomics/test_activation_integration.py) — live Agung coverage for snapshot and InfoDesk.peer drift, pricing, deterministic IDs, genuine revert decoding, write-free previews, activation guards, atomic activations, management reads, reversible suspend/resume, submission reporting, and known-hash reconciliation.

This suite spends real PEAQ and cannot be undone. Activation mints a permanent machine identity and bonds tokens against it — there is no deactivate and no refund. A full run costs roughly 0.8 PEAQ of bond plus activation and management gas at the tier-0 price stamped on Agung. It is gated behind a dedicated variable on top of the integration marker.

PEAQOS_TOKENOMICS_INTEGRATION=1 \
PEAQOS_RPC_URL=https://peaq-agung.api.onfinality.io/public \
PEAQOS_PRIVATE_KEY=0xYOUR_FUNDED_KEY \
pytest -m integration tests/integration/tokenomics/

Contract addresses are deliberately not environment variables — resolving them from the SDK's approved snapshot is the behaviour under test. See the activation and management guides for the API reference and live-suite limitations.

PEAQOS_PRIVATE_KEY must match ^0x[0-9a-fA-F]{64}$. Never commit a real key. Never run the suite against mainnet.

Build

python -m build

Produces dist/peaq_os_sdk-<version>-py3-none-any.whl and dist/peaq_os_sdk-<version>.tar.gz for downstream testing.

License

See LICENSE for details.