Best for
- ALWAYS use this skill when the user mentions almanak, DeFi strategy, trading strategy, yield farming, liquidity provision, token swap, borrowing, lending, perpetuals, staking, vault deposit, bridging tokens, backtesting…
almanak-co/sdk/almanak/skills/almanak-strategy-builder/SKILL.md
Build, test, and deploy DeFi trading strategies using the Almanak SDK. ALWAYS use this skill when the user mentions almanak, DeFi strategy, trading strategy, yield farming, liquidity provision, token swap, borrowing, lending, perpetuals, staking, vault deposit, bridging tokens, backtesting, paper trading, or on-chain execution. Use for writing strategy.py files, composing intents (Swap, LP, Borrow, Supply, Perp, Bridge, Stake, Vault, Prediction), working with config.json strategy parameters, run
Decision brief
You are helping a quant build DeFi strategies using the Almanak SDK. Strategies are Python classes that return Intent objects. The framework handles compilation to transactions, execution, and state management.
Compatibility matrix
| Platform | Status | Evidence | What to check |
|---|---|---|---|
| Codex | Not declared | No explicit evidence | Portability before use |
| Claude Code | Not declared | No explicit evidence | Portability before use |
| Cursor | Not declared | No explicit evidence | Portability before use |
| Gemini CLI | Not declared | No explicit evidence | Portability before use |
Installation
The source command is displayed only when detected. A safe inspection prompt is always available so your agent can explain every action before execution.
npx skills add https://github.com/almanak-co/sdk --skill "almanak/skills/almanak-strategy-builder"Inspect the Agent Skill "almanak-strategy-builder" from https://github.com/almanak-co/sdk/blob/f06a1b441d5f76fd0bbe63eaf68b7ad40dc4da8e/almanak/skills/almanak-strategy-builder/SKILL.md at commit f06a1b441d5f76fd0bbe63eaf68b7ad40dc4da8e. List every install step, command, network request, credential, file read/write, external action, and rollback step. Explain whether it fits my task. Do not install or execute anything until I approve.
Workflow
Review the “Quick Start” section in the pinned source before continuing.
For permissionless, market-keyed protocols (morphoblue) a marketid must be verified on-chain before it is pinned into a config — a same-pair market can carry a hostile oracle or IRM.
Review the “Usage” section in the pinned source before continuing.
Review the “Multi-Step with IntentSequence” section in the pinned source before continuing.
This file is documentation for AI coding assistants - it contains no executable code or live credentials.
Permission review
The documentation asks the agent to run terminal commands or scripts.
pipx install almanakThe documentation includes network, browsing, or remote request actions.
# RPC_URL=https://...The documentation asks the agent to create, modify, or delete local files.
almanak strat permissions -o permissions.json # Write to fileThe documentation asks the agent to create, modify, or delete local files.
# Generate Zodiac permissions and write to file (recommended)Evidence record
| Signal | Value | Evidence type | Meaning |
|---|---|---|---|
| Quality score | 96/100 | Computed | Documentation, specificity, maintenance, and trust rules |
| Repository stars | 60 | Source | Repository attention, not individual Skill quality |
| Compatibility | 0 platforms | Source | Declared in the catalog source record |
| Usage guide | automated source guide | Editorial | Generated or reviewed according to the visible evidence level |
Pinned source
You are helping a quant build DeFi strategies using the Almanak SDK. Strategies are Python classes that return Intent objects. The framework handles compilation to transactions, execution, and state management.
This file is documentation for AI coding assistants - it contains no executable code or live credentials.
The Almanak SDK uses a gateway-isolated architecture:
.env file referenced below is for local development only and is never deployed to production environments..env are used solely for local Anvil fork testing and have no access to mainnet funds in production.See the Gateway section for details.
IMPORTANT — Zodiac permissions auto-generation: Whenever you create a new
strategy or modify strategy.py in a way that changes protocols, intent types,
tokens, or chains, you MUST regenerate the Zodiac Roles permission manifest
by running almanak strat permissions -o permissions.json
from the strategy directory. The file MUST be named exactly permissions.json
— the platform deploy pipeline hardcodes this filename. This ensures the Safe
wallet permissions stay in sync with the strategy's actual contract interactions.
Always do this as a final step after writing or editing strategy code.
# Install the CLI globally
pipx install almanak
# Scaffold a new strategy (creates a self-contained Python project)
almanak strat new --template ta_swap --name my_rsi --chain arbitrum
# Run on local Anvil fork (auto-starts gateway + Anvil)
cd my_rsi
almanak strat run --network anvil --once
# Run a single iteration on mainnet
almanak strat run --once
# Browse and copy a working demo strategy
almanak strat demo
Each scaffolded strategy is a self-contained Python project with its own
pyproject.toml, .venv/, and uv.lock. The same files drive both local
development and the platform's cloud Docker build.
Strategy project structure:
my_strategy/
strategy.py # IntentStrategy subclass with decide() method
config.json # Runtime parameters (tokens, thresholds, funding)
pyproject.toml # Dependencies + [tool.almanak] metadata
uv.lock # Locked dependencies (created by uv sync)
.venv/ # Per-strategy virtual environment
.env # Local dev credentials (not deployed; see Security Model)
.gitignore # Git ignore rules
.python-version # Python version pin (3.12)
__init__.py # Package exports
tests/ # Test scaffold
AGENTS.md # AI agent guide
pyproject.toml example:
[project]
name = "my-strategy"
version = "0.1.0"
requires-python = ">=3.12"
dependencies = [
"almanak>=2.15.0",
]
[tool.almanak.run]
interval = 60
The [tool.almanak.run] section is required — it sets the execution interval (in seconds)
for the strategy loop in production. Always include it when writing pyproject.toml manually.
Adding dependencies:
uv add pandas-ta # Updates pyproject.toml + uv.lock + .venv/
uv run pytest tests/ -v # Run tests in the strategy's venv
For Anvil testing, add anvil_funding to config.json so your wallet is auto-funded on fork start
(see Configuration below).
# strategy.py
from decimal import Decimal
from almanak import MarketSnapshot
from almanak.framework.strategies import IntentStrategy, almanak_strategy
from almanak.framework.intents import Intent
@almanak_strategy(
name="my_strategy",
version="1.0.0",
supported_chains=["arbitrum"],
supported_protocols=["uniswap_v3"],
intent_types=["SWAP", "HOLD"],
default_chain="arbitrum",
)
class MyStrategy(IntentStrategy):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.trade_size = Decimal(str(self.config.get("trade_size_usd", "100")))
def decide(self, market: MarketSnapshot) -> Intent | None:
rsi = market.rsi("WETH", period=14)
if rsi.value < 30:
return Intent.swap(
from_token="USDC", to_token="WETH",
amount_usd=self.trade_size, max_slippage=Decimal("0.005"),
)
return Intent.hold(reason=f"RSI={rsi.value:.1f}, waiting")
Note:
amount_usd=requires a live price oracle from the gateway. If swaps revert with "Too little received", switch toamount=(token units) which bypasses USD-to-token conversion. Always verify pricing on first live run with--dry-run --once.
All strategies inherit from IntentStrategy and implement one method:
def decide(self, market: MarketSnapshot) -> Intent | None
The framework calls decide() on each iteration with a fresh MarketSnapshot.
Return an Intent object (swap, LP, borrow, etc.) or Intent.hold().
__init__: Extract config parameters, set up statedecide(market): Called each iteration - return an Intenton_intent_executed(intent, success, result): Optional callback after executionget_status(): Optional - return dict for monitoring dashboardssupports_teardown() / generate_teardown_intents(): Optional safe shutdownAttaches metadata used by the framework and CLI:
@almanak_strategy(
name="my_strategy", # Unique identifier
description="What it does", # Human-readable description
version="1.0.0", # Strategy version
author="Your Name", # Optional
tags=["trading", "rsi"], # Optional tags for discovery
supported_chains=["arbitrum"], # Which chains this runs on
supported_protocols=["uniswap_v3"], # Which protocols it uses
intent_types=["SWAP", "HOLD"], # Intent types it may return
default_chain="arbitrum", # Default chain for execution
quote_asset="USD", # Asset performance is measured in (USD default, or a token)
)
IMPORTANT — Intent Type Teardown Complements: intent_types must include
both the "open" and "close" side of every operation. These are used to generate
Zodiac Roles permissions for Safe wallet deployments. If you declare the open
side without its complement, the strategy will deploy but teardown will fail
on-chain because the wallet lacks permission for the close operation.
| If you declare... | You MUST also declare... |
|---|---|
SUPPLY | WITHDRAW |
BORROW | REPAY |
LP_OPEN | LP_CLOSE |
VAULT_DEPOSIT | VAULT_REDEEM |
PERP_OPEN | PERP_CLOSE |
The decorator emits a UserWarning at import time if complements are missing.
The permission generator also auto-expands missing complements as a safety net,
but always declare them explicitly.
quote_asset declares the asset your strategy's performance is measured in. It defaults
to USD and sets the numeraire for performance reporting: backtests and paper runs
compute their canonical performance metrics in it (performance_denomination in the
result summary names the unit; *_usd counterparts are kept alongside), and the hosted
platform reports performance in it. It does
not change execution behaviour — only how results are measured — so a wrong value reports
performance in the wrong unit: a BTC-growth strategy declared "USD" shows USD PnL and no
BTC-denominated metrics. Choose by asking what quantity the strategy is trying to grow —
if the goal is stated ("increase BTC"), the denomination must match it.
quote_asset="USD". Declare it explicitly rather than omitting it —
the scaffold and packaged demos do, and an explicit value makes the choice reviewable.quote_asset={"type": "token", "chain_id": <int>, "address": "0x..."} (or
QuoteAsset.token(chain_id, address) from almanak.core.models.quote_asset),
identifying the token by its canonical (chain_id, address). Use a numeric chain_id
only, never a chain name. Represent native gas tokens by their wrapped ERC-20
(ETH->WETH, MNT->WMNT, 0G->W0G).Set a token quote asset only when the strategy's goal is to grow a quantity of that
token — pure accumulators, ETH-denominated LST leverage loops (collateral and borrow are
ETH-family), native-asset staking, and same-asset-family LP pools built to grow that asset
(e.g. a WBTC/tBTC pool as a BTC accumulator quotes in WBTC). Mixed-family LP (e.g.
WETH/USDC), USD-yield lending, stablecoin, delta-neutral, and USD-collateral perp
strategies stay on the USD default. quote_asset is distinct from quote_token (a
trading-pair leg) and starting_asset (an LP round-trip asset).
You can also set it per-deployment in config.json ("quote_asset": "USD" or the token
object), which overrides the decorator default on live runs at boot (backtests read the
decorator value). It is frozen at boot — not hot-reloadable. When denominating a strategy
for backtesting, set it on the decorator.
In __init__, read parameters from self.config (dict loaded from config.json):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.trade_size = Decimal(str(self.config.get("trade_size_usd", "100")))
self.rsi_period = int(self.config.get("rsi_period", 14))
self.base_token = self.config.get("base_token", "WETH")
Also available: self.chain (str), self.wallet_address (str), self.chains (list[str]),
self.get_wallet_for_chain(chain) (str).
All intents are created via Intent factory methods. Import:
from almanak.framework.intents import Intent
BaseIntent.registry_handle (added by VIB-4192 / T06b; factory ergonomics
lifted by VIB-4285) is an optional opaque field that disambiguates multiple positions on
the same (primitive, semantic_group).
ix_registry_auto_mode
partial unique index) does not reject the second open. Use stable per-position
handles (leg_narrow, leg_wide) — NOT action-scoped suffixes
(leg_narrow:open / leg_narrow:close), so the same handle survives the full
open → close → rebalance lifecycle. Example:
Intent.lp_open(..., registry_handle="hedge_leg_long").Synthesising a handle that does not match the prior open will fail at
save_ledger_and_registry with RegistryAutoCollisionError. See
../../../blueprints/28-position-registry.md §3.5 and §6 anti-pattern #13 for the
contract.
The reserved-field rule above gets registry_handle right; the rule below is about
the dispatch cadence. They are complementary — both must be right for a
multi-position strategy to work.
Emit one opening intent per decide() iteration, not as a list. Drive iterations
with a _phase field that advances only when on_intent_executed observes a real
position_id on the receipt. The list-return shape (return [open_a, open_b]) and
Intent.sequence([open_a, open_b]) both commit two legs to a single market snapshot,
give leg 2 no opportunity to re-size against leg 1's actual on-chain output, and
provide no clean partial-success state. Reference implementation:
strategies/accounting/lp_dual/strategy.py (two LPs, one pool, phase machine,
self-sized amounts, position-id-keyed close).
Self-size each leg from live market.balance(...) at the moment the open is built —
leg #1 takes commit_pct of the available balance, leg #2 takes 0.99 of what
remains (the 1% safety margin absorbs gas / dust / slippage drift between balance
read and tx submission). Hardcoded per-leg amounts in config.json work in steady
state but desync against any real-world mint slippage; the live-balance pattern is
what lp_dual / lp_triple use because mint slippage is observable on every
real-Anvil run.
Skeleton:
PHASE_INIT = "init"
PHASE_LP1_OPEN = "lp1_open"
PHASE_BOTH_OPEN = "both_open"
def decide(self, market):
if self._phase == PHASE_INIT:
return self._build_lp_open(market, position_index=1)
if self._phase == PHASE_LP1_OPEN:
return self._build_lp_open(market, position_index=2)
if self._phase == PHASE_BOTH_OPEN:
return Intent.hold(reason="Both LPs open — awaiting teardown")
return Intent.hold(reason=f"Unknown phase {self._phase!r}")
def _build_lp_open(self, market, *, position_index):
token0_balance = Decimal(str(market.balance(self.token0_symbol).balance))
token1_balance = Decimal(str(market.balance(self.token1_symbol).balance))
if position_index == 1:
commit_pct = self.lp_capital_split_pct # e.g. 0.50
handle = "leg_narrow"
else:
commit_pct = Decimal("0.99") # leg 2 takes what's left
handle = "leg_wide"
return Intent.lp_open(
pool=self.pool,
amount0=token0_balance * commit_pct,
amount1=token1_balance * commit_pct,
range_lower=...,
range_upper=...,
registry_handle=handle,
)
def on_intent_executed(self, intent, success, result):
if not success or intent.intent_type.value != "LP_OPEN":
return # phase stays put → next iteration retries
position_id = getattr(result, "position_id", None)
if not position_id:
return # mint without id → don't advance, retry next tick (prevents stranding)
if self._phase == PHASE_INIT:
self._position_id_1 = str(position_id)
self._phase = PHASE_LP1_OPEN
elif self._phase == PHASE_LP1_OPEN:
self._position_id_2 = str(position_id)
self._phase = PHASE_BOTH_OPEN
State-transition shape:
INIT ──LP_OPEN(narrow)──▶ LP1_OPEN ──LP_OPEN(wide)──▶ BOTH_OPEN ──teardown──▶ DONE
For richer patterns (out-of-order middle close on three positions), see
strategies/accounting/lp_triple/strategy.py. The full design contract lives in
../../../blueprints/04-strategy-layer.md §Multi-position dispatch.
The following IntentType strings are placeholders in the canonical taxonomy. They
exist to reserve the name and primitive classification but have no compiler / executor
behind them. Emitting one of these from strategy code raises PlaceholderIntentError
at intent-compile time (before any on-chain action) and is also refused at the
PolicyEngine boundary for LLM-mediated surfaces:
| IntentType | Future primitive |
|---|---|
LIQUIDATE | LIQUIDATION |
OPEN_CDP | CDP |
MINT_STABLE | CDP |
REPAY_STABLE | CDP |
CLOSE_CDP | CDP |
Do not emit these. When the corresponding primitive ships, the placeholder rows will be swapped for real handlers atomically (taxonomy + compiler + handler in one PR) — your strategy code does not change.
Intent.swap - Exchange tokens on a DEX
Intent.swap(
from_token="USDC", # Token to sell
to_token="WETH", # Token to buy
amount_usd=Decimal("1000"), # Amount in USD (use amount_usd OR amount)
amount=Decimal("500"), # Amount in token units (alternative to amount_usd)
max_slippage=Decimal("0.005"), # Max slippage (0.5%)
max_price_impact=Decimal("0.10"), # Optional: max quoter-vs-oracle deviation (default: 10%; override for thin venues)
protocol="uniswap_v3", # Optional: specific DEX
chain="arbitrum", # Optional: override chain
destination_chain="base", # Optional: cross-chain swap
swap_params=None, # Optional connector-owned exact-route constraints
)
Use amount="all" to swap the entire balance.
amount= vs amount_usd=: Use amount_usd= to specify trade size in USD (requires a live
price oracle from the gateway). Use amount= to specify exact token units (more reliable for live
trading since it bypasses USD-to-token conversion). When in doubt, prefer amount= for mainnet.
For uniswap_v3, sushiswap_v3, pancakeswap_v3, and agni_finance, pin a same-chain
swap to one immutable pool with swap_params={"pool": <address>}. The compiler reads
token0(), token1(), and fee() from that pool and verifies factory.getPool(...) before
building the transaction. A wrong pair, foreign factory, unsupported protocol, or unreadable
pool fails compilation; it never falls back to another pool or route.
APPROVED_POOL = "0xc655e1a100a084d9ac91c269b0a7cb0e62263fcf"
Intent.swap(
from_token=self.quote_token_address,
to_token=self.base_token_address,
amount=Decimal("100"),
max_slippage=Decimal("0.0075"),
protocol="pancakeswap_v3",
chain="bsc",
swap_params={"pool": APPROVED_POOL},
)
Use swap_params={"fee_tier": 500} only when any factory pool at that fee tier is acceptable;
use pool when the address itself is an invariant. Pool pinning is not supported for cross-chain
aggregator swaps. Use the same swap_params in normal execution and teardown.
Intent.lp_open - Open a concentrated LP position
Intent.lp_open(
pool="WETH/USDC", # Pool identifier
amount0=Decimal("1.0"), # Amount of token0
amount1=Decimal("2000"), # Amount of token1
range_lower=Decimal("1800"), # Lower price bound
range_upper=Decimal("2200"), # Upper price bound
range_spec=None, # Typed range: PriceBand | TickBand (alternative to range_lower/range_upper)
protocol="uniswap_v3", # Default: uniswap_v3
chain=None, # Optional override
coin_amounts=None, # Multi-coin pools (e.g. Curve 3pool): per-coin amounts by pool index
max_slippage=None, # Optional slippage bound on the deposit floor
)
Typed ranges (VIB-5555):
range_specacceptsPriceBand(lower=..., upper=...)(human prices, token1-per-token0 — the portable default, converted to ticks by each connector) orTickBand(lower=..., upper=...)(raw protocol ticks, escape hatch). Import both fromalmanak.framework.intents. The legacyrange_lower/range_upperpair is still accepted and equivalent to aPriceBand; pass one form, not both.
Intent.lp_close - Close an LP position
Intent.lp_close(
position_id="12345", # NFT token ID returned by lp_open; ALSO the registry handle
pool="WETH/USDC", # Optional pool identifier
collect_fees=True, # Collect accumulated fees
protocol="uniswap_v3",
amount=None, # "all" = chain off prior LP_OPEN's minted LP (fungible-LP allowlist, e.g. Pendle)
max_slippage=None, # Optional slippage bound on withdrawal min-amounts (Curve; default 50 bps)
coin_index=None, # Single-sided exit: withdraw all as one pool coin (Curve only, VIB-5437)
imbalanced_amounts=None, # Exact per-coin exit amounts, fail-closed max-burn (Curve StableSwap only, VIB-5438)
)
Curve exit selectors:
coin_indexroutes viaremove_liquidity_one_coin;imbalanced_amountsroutes viaremove_liquidity_imbalance. They are mutually exclusive; leave bothNonefor the proportional all-coin close. Only connectors declaring thelp_close_exit_selectorscapability (currently Curve) compile them.
position_idfromlp_open's result is the registry handle (VIB-4192 / T06b). Persist it in strategy state (self.state["lp_position_id"] = result.position_id) and pass it back tolp_closeat teardown. The framework uses it to look up the open row inposition_registry; a mismatch (handle present but no live row) raisesRegistryAutoCollisionErrorbefore any on-chain call.
Intent.collect_fees - Harvest LP fees without closing
Intent.collect_fees(
pool="WETH/USDC",
protocol="traderjoe_v2",
)
Intent.supply - Deposit collateral into a lending protocol
Intent.supply(
protocol="aave_v3",
token="WETH",
amount=Decimal("10"),
use_as_collateral=True, # Enable as collateral (default: True)
market_id=None, # Required for Morpho Blue
)
Intent.borrow - Borrow tokens against collateral
Intent.borrow(
protocol="aave_v3",
collateral_token="WETH",
collateral_amount=Decimal("10"),
borrow_token="USDC",
borrow_amount=Decimal("5000"),
interest_rate_mode="variable", # Aave: "variable" only (stable deprecated)
market_id=None, # Required for Morpho Blue
)
Intent.repay - Repay borrowed tokens
Intent.repay(
protocol="aave_v3",
token="USDC",
amount=Decimal("5000"),
repay_full=False, # Set True to repay entire debt
market_id=None,
)
Intent.deleverage - Emergency repay triggered by risk management (e.g. HF below threshold)
Intent.deleverage(
protocol="aave_v3",
token="USDC",
amount=Decimal("5000"),
trigger_reason="health_factor_below_threshold", # Human-readable reason for the deleverage
observed_hf=Decimal("1.05"), # Health factor at trigger time (persisted as health_factor_before)
target_hf=Decimal("1.5"), # Target HF after deleverage
repay_full=False, # Set True to repay entire debt
market_id=None,
)
Compiles to the same on-chain execution as Intent.repay. The trigger_reason, observed_hf,
and target_hf are stored in the accounting layer so dashboards can surface why the deleverage
was forced. The observed_hf is persisted as health_factor_before in the accounting event.
DELEVERAGE is a mandatory live event type (fail-closed) — the runner will log a WARNING when it
detects a deleverage.
Intent.withdraw - Withdraw from lending protocol
Intent.withdraw(
protocol="aave_v3",
token="WETH",
amount=Decimal("10"),
withdraw_all=False, # Set True to withdraw everything
market_id=None,
is_collateral=True, # Morpho Blue only: True = collateral, False = loan token
)
Intent.perp_open - Open a perpetual futures position
Intent.perp_open(
market="ETH/USD",
collateral_token="USDC",
collateral_amount=Decimal("1000"),
size_usd=Decimal("5000"),
is_long=True,
leverage=Decimal("5"),
max_slippage=Decimal("0.01"),
protocol="gmx_v2",
)
Intent.perp_close - Close a perpetual futures position
Intent.perp_close(
market="ETH/USD",
collateral_token="USDC",
is_long=True,
size_usd=None, # None = close full position
max_slippage=Decimal("0.01"),
protocol="gmx_v2",
position_id=None, # Required for venues keyed on bytes32 (e.g. pancakeswap_perps)
)
Intent.perp_withdraw - Withdraw free margin off a perp venue's off-chain account back to L1 (a cash movement, not a trade — no position, no PnL). On Hyperliquid this compiles to a CoreWriter perp→spot usdClassTransfer followed by a spot→L1 spotSend HyperCore→HyperEVM bridge of USDC (VIB-5617).
Intent.perp_withdraw(
amount=Decimal("6.99"), # human token amount, or "all" ONLY as a chained amount (a prior step's output)
asset="USDC", # the only HyperCore bridge-linked token today
protocol="hyperliquid",
chain="hyperevm",
destination=None, # defaults to the deployment wallet; the bridge always credits the sender
)
Intent.perp_cancel_order - Cancel a pending (unfilled) perp order and recover its committed collateral and unspent execution fee (VIB-5568). Not a position open/close — a refund of committed-but-unspent collateral, e.g. to recover a stranded pending order discovered during teardown.
Intent.perp_cancel_order(
order_key="0x...", # bytes32 order key (0x-prefixed, 66 chars) from the open receipt or residual discovery
protocol="gmx_v2", # Default: gmx_v2
chain=None, # Optional override
)
Intent.bridge - Cross-chain token transfer
Intent.bridge(
token="USDC",
amount=Decimal("1000"),
from_chain="arbitrum",
to_chain="base",
max_slippage=Decimal("0.005"),
preferred_bridge=None, # Optional: specific bridge protocol
)
Intent.stake - Liquid staking deposit
Intent.stake(
protocol="lido",
token_in="ETH",
amount=Decimal("10"),
receive_wrapped=True, # Receive wrapped token (e.g., wstETH)
)
Intent.unstake - Withdraw from liquid staking
Intent.unstake(
protocol="lido",
token_in="wstETH",
amount=Decimal("10"),
protocol_params=None, # Optional: e.g. {"phase": "cooldown"} for Ethena
)
Intent.flash_loan - Borrow and repay in a single transaction
Intent.flash_loan(
provider="aave", # "aave", "balancer", "morpho", or "auto"
token="USDC",
amount=Decimal("100000"),
callback_intents=[...], # Intents to execute with the borrowed funds
)
Intent.vault_deposit - Deposit into an ERC-4626 vault
Intent.vault_deposit(
protocol="metamorpho", # Vault protocol
vault_address="0x...", # Vault contract address
amount=Decimal("1000"), # Amount of underlying to deposit (or "all")
deposit_token="USDC", # Underlying token symbol (for backtesting)
chain="ethereum", # Optional: override chain
)
Intent.vault_redeem - Redeem shares from an ERC-4626 vault
Intent.vault_redeem(
protocol="metamorpho", # Vault protocol
vault_address="0x...", # Vault contract address
shares=Decimal("1000"), # Shares to redeem (or "all")
deposit_token="USDC", # Underlying token symbol (for backtesting)
chain="ethereum", # Optional: override chain
)
Intent.prediction_buy(
market_id="will-bitcoin-exceed-100000", # Polymarket market ID or slug
outcome="YES", # "YES" or "NO"
amount_usd=Decimal("100"), # USDC to spend (or use shares=)
protocol="polymarket",
)
Intent.prediction_sell(
market_id="will-bitcoin-exceed-100000",
outcome="YES",
shares=Decimal("50"), # Shares to sell (or "all")
protocol="polymarket",
)
Intent.prediction_redeem(
market_id="will-bitcoin-exceed-100000", # Redeem after market resolves
protocol="polymarket",
)
Intent.ensure_balance - Meta-intent that resolves to a BridgeIntent (if balance is insufficient) or HoldIntent (if already met). Call .resolve(market) before returning from decide().
intent = Intent.ensure_balance(
token="USDC",
min_amount=Decimal("1000"),
target_chain="arbitrum",
max_slippage=Decimal("0.005"),
preferred_bridge=None,
)
# Must resolve before returning - returns BridgeIntent or HoldIntent
resolved = intent.resolve(market)
return resolved
Intent.wrap (WrapNative) - Wrap native tokens to ERC-20 (ETH -> WETH, MATIC -> WMATIC, etc.)
Intent.wrap(
token="WETH", # Wrapped token symbol to receive
amount=Decimal("0.5"), # Amount of native token to wrap (or "all")
chain="arbitrum", # Target chain
)
Intent.unwrap (UnwrapNative) - Unwrap wrapped native tokens (WETH -> ETH, WMATIC -> MATIC, etc.)
Intent.unwrap(
token="WETH", # Wrapped token symbol
amount=Decimal("0.5"), # Amount to unwrap (or "all")
chain="arbitrum", # Target chain
)
Intent.hold - Do nothing this iteration
Intent.hold(reason="RSI in neutral zone")
Intent.sequence - Execute multiple intents in order
Intent.sequence(
intents=[
Intent.swap(from_token="USDC", to_token="WETH", amount_usd=Decimal("1000")),
Intent.supply(protocol="aave_v3", token="WETH", amount=Decimal("0.5")),
],
description="Buy WETH then supply to Aave",
)
Use "all" to reference the full output of a prior intent:
Intent.sequence(intents=[
Intent.swap(from_token="USDC", to_token="WETH", amount_usd=Decimal("1000")),
Intent.supply(protocol="aave_v3", token="WETH", amount="all"), # Uses swap output
])
The MarketSnapshot passed to decide() provides these methods:
price = market.price("WETH") # Decimal, USD price
price = market.price("WETH", quote="USDC") # Price in USDC terms
pd = market.price_data("WETH") # PriceData object
pd.price # Decimal - current price
pd.price_24h_ago # Decimal
pd.change_24h_pct # Decimal
pd.high_24h # Decimal
pd.low_24h # Decimal
pd.timestamp # datetime
For a non-crypto reference such as XAU/USD, use the dedicated exact-feed API. Never substitute
market.price("XAU"): generic token pricing does not carry the required feed identity or market
session state.
reference = market.reference_price("XAU", chain="bsc", quote="USD")
reason = reference.trade_block_reason(max_age_seconds=300, min_confidence=0.90)
if reason is not None:
return Intent.hold(reason=reason)
reference.price # Decimal | None
reference.source # exact provider/feed identity
reference.observed_at # provider observation time, not gateway receipt time
reference.market_status # OPEN | CLOSED | UNKNOWN
reference.market_status_as_of # time at which the gateway evaluated the session
reference.stale # provider heartbeat result
The API returns a typed unavailable result rather than an inferred value. is_tradeable(...) and
trade_block_reason(...) fail closed for unavailable, malformed, stale, closed/unknown-session,
future-dated, over-age, or low-confidence observations.
bal = market.balance("USDC")
bal.balance # Decimal - token amount
bal.balance_usd # Decimal - USD value
bal.symbol # str
bal.address # str - token contract address
TokenBalance supports numeric comparisons: bal > Decimal("100").
All indicators accept token, period (int), and timeframe (str, default "4h").
rsi = market.rsi("WETH", period=14, timeframe="4h")
rsi.value # Decimal (0-100)
rsi.is_oversold # bool (value < 30)
rsi.is_overbought # bool (value > 70)
rsi.signal # "BUY" | "SELL" | "HOLD"
macd = market.macd("WETH", fast_period=12, slow_period=26, signal_period=9)
macd.macd_line # Decimal
macd.signal_line # Decimal
macd.histogram # Decimal
macd.is_bullish_crossover # bool
macd.is_bearish_crossover # bool
bb = market.bollinger_bands("WETH", period=20, std_dev=2.0)
bb.upper_band # Decimal
bb.middle_band # Decimal
bb.lower_band # Decimal
bb.bandwidth # Decimal
bb.percent_b # Decimal (0.0 = at lower band, 1.0 = at upper band)
bb.is_squeeze # bool
stoch = market.stochastic("WETH", k_period=14, d_period=3)
stoch.k_value # Decimal
stoch.d_value # Decimal
stoch.is_oversold # bool
stoch.is_overbought # bool
atr_val = market.atr("WETH", period=14)
atr_val.value # Decimal (absolute)
atr_val.value_percent # Decimal, percentage points (2.62 means 2.62%, not 0.0262)
atr_val.is_high_volatility # bool
sma = market.sma("WETH", period=20)
ema = market.ema("WETH", period=12)
# Both return MAData with: .value, .is_price_above, .is_price_below, .signal
adx = market.adx("WETH", period=14)
adx.value # Decimal
adx.plus_di # Decimal
adx.minus_di # Decimal
adx.is_trending # bool
adx.is_uptrend # bool
obv = market.obv("WETH", signal_period=21)
obv.value # Decimal
obv.signal # Decimal
obv.is_bullish # bool
cci = market.cci("WETH", period=20)
cci.value # Decimal
cci.is_overbought # bool
cci.is_oversold # bool
ich = market.ichimoku("WETH", tenkan_period=9, kijun_period=26, senkou_b_period=52)
ich.tenkan_sen # Decimal (conversion line)
ich.kijun_sen # Decimal (base line)
ich.senkou_span_a # Decimal (leading span A)
ich.senkou_span_b # Decimal (leading span B)
ich.is_bullish_crossover # bool
ich.is_above_cloud # bool
# Per-token reads work today on every deployment surface.
weth_price = market.price("WETH") # Decimal
wbtc_price = market.price("WBTC") # Decimal
usdc_bal = market.balance("USDC") # TokenBalance
weth_bal = market.balance("WETH") # TokenBalance
usd_val = market.balance_usd("WETH") # Decimal - USD value of holdings
total = market.total_portfolio_usd() # Decimal
Batch helpers are Phase 2. The deprecated data-layer class exposes
market.prices([...])/market.balances([...])batch fetchers. The canonical strategy-facing class deliberately does NOT lift these names in the ALM-2696 fix because legacy callers (runner_state.py, trust tests) historically usedhasattr(market, "prices")/market.prices.get(...)patterns whose absence was load-bearing. Phase 2 (VIB-4065 / GH#2126) migrates those callers in lockstep before lifting these batch names. Use the per-token form above until then.
# USD value of an arbitrary collateral amount (for perp position sizing)
col_usd = market.collateral_value_usd("WETH", Decimal("2")) # Decimal - amount * price
df = market.ohlcv("WETH", timeframe="1h", limit=100) # pd.DataFrame
# Columns: open, high, low, close, volume
Use these (not
market.price()) for anything execution-facing — LP range bounds, range-exit tests.market.price()is a USD valuation oracle (hardcoded1.0for stablecoins) and can silently diverge from the pool's actual price. See "LP Rebalancing" above.
pool = market.pool_price("0x...") # DataEnvelope[PoolPrice]
pool = market.pool_price_by_pair("WETH", "USDC") # DataEnvelope[PoolPrice]
reserves = market.pool_reserves("0x...") # PoolReserves
history = market.pool_history("0x...", resolution="1h", protocol="uniswap_v3") # DataEnvelope[list[PoolSnapshot]]
# `protocol` is REQUIRED keyword-only (VIB-4755 D-2 — closes silent cross-protocol surface).
# Must match the pool's actual protocol slug: "uniswap_v3", "aerodrome", "pancakeswap_v3", etc.
# A defaulted protocol on a non-uniswap_v3 pool address would have routed through
# CoinGecko Onchain (which does not filter on protocol slug) and silently labelled the
# served data with the wrong protocol — see docs/internal/uat-cards/VIB-4755.md §D-2.
analytics = market.pool_analytics("0x...") # DataEnvelope[PoolAnalytics]
best = market.best_pool("WETH", "USDC", metric="fee_apr") # DataEnvelope[PoolAnalyticsResult]
Provider availability:
pool_*,twap/lwap,liquidity_depth,estimate_slippage,pool_analytics/best_pool,il_exposure/projected_il,realized_vol/vol_cone,portfolio_risk/rolling_sharpe,yield_opportunities,lst_*, prediction-market methods, and the rate-history methods are all provider-driven. The runner wires the corresponding provider (pool reader registry, price aggregator, IL calculator, …); when a provider is not wired the method raisesValueError("No <X> configured for MarketSnapshot")rather than returning silently. Do not guard these calls withhasattr(market, ...)— the methods always exist; catchValueError(or one of the typed*UnavailableErrorsubclasses defined inalmanak.framework.data.market_snapshot) if you need to degrade gracefully.Carve-out —
prediction_price(): unlike the other prediction-market methods (prediction(),prediction_positions(),prediction_orders(), all of which raiseValueErrorwhen no provider is wired),prediction_price()returnsNoneas a soft-signal fallback. Strategies that use it as a side-channel signal can therefore branch onif (p := market.prediction_price(...)) is not None:instead of wrapping the call intry / except ValueError. This matches the existing convention preserved by ALM-2696 and is pinned by the regression suite.
twap = market.twap("WETH/USDC", window_seconds=300) # DataEnvelope[AggregatedPrice]
# Explicit-pool form: when you pass `pool_address` directly, you must also
# pass token decimals (or have a `pool_reader_registry` wired so the snapshot
# can resolve them automatically). There is no "WETH/USDC default" — the
# decimals are required for the tick-to-price conversion.
twap = market.twap(
"WBTC/WETH",
pool_address="0x...",
token0_decimals=8, token1_decimals=18,
)
lwap = market.lwap("WETH/USDC") # DataEnvelope[AggregatedPrice]
depth = market.liquidity_depth("0x...") # DataEnvelope[LiquidityDepth]
slip = market.estimate_slippage("WETH", "USDC", Decimal("10000")) # DataEnvelope[SlippageEstimate]
# Canonical fields live on slip.value and are integer basis points:
if not slip.value.within_limits(max_slippage_bps=75, max_price_impact_bps=100):
return Intent.hold(reason="pre-trade slippage or price impact exceeds limit")
prices = market.price_across_dexs("WETH", "USDC", Decimal("1")) # list[DexQuote]
best_dex = market.best_dex_price("WETH", "USDC", Decimal("1")) # BestDexResult
For an approved-pool strategy, pin every execution-facing read to the same address. A useful USD depth contract is to simulate a trade equal to the configured minimum depth and require its effective slippage to remain inside the limit. Check both directions when both entry and exit are possible; convert the base-side USD amount with the already-validated reference price.
pool = market.pool_price(APPROVED_POOL, chain="bsc")
depth = market.liquidity_depth(APPROVED_POOL, chain="bsc")
trade_amount = Decimal("27000") # Quote-token units; the quote token is a USD stablecoin.
quote_side = market.estimate_slippage(
self.quote_token_address,
self.base_token_address,
trade_amount,
chain="bsc",
protocol="pancakeswap_v3",
pool_address=APPROVED_POOL,
)
base_side = market.estimate_slippage(
self.base_token_address,
self.quote_token_address,
trade_amount / reference.price,
chain="bsc",
protocol="pancakeswap_v3",
pool_address=APPROVED_POOL,
)
if not all(
estimate.value.within_limits(max_slippage_bps=75, max_price_impact_bps=100)
for estimate in (quote_side, base_side)
):
return Intent.hold(reason="approved pool fails the $27k / 75 bps depth contract")
return Intent.swap(
self.quote_token_address,
self.base_token_address,
amount=trade_amount,
protocol="pancakeswap_v3",
swap_params={"pool": APPROVED_POOL},
max_slippage_bps=75,
max_price_impact_bps=100,
)
pool_price includes the exact pool address, fee, block number, and timestamp. liquidity_depth
contains current tick/liquidity plus initialized ticks. estimate_slippage(pool_address=...)
simulates those ticks and is the execution-grade pre-trade quote/impact check. If any read is
unavailable, propagate the typed data error or return HOLD; do not retry against another pool.
The result fields are effective_slippage_bps and price_impact_bps; there are no generic
slippage, slippage_pct, price_impact, or price_impact_pct compatibility aliases. Do not
probe guessed field names with getattr fallbacks—use within_limits() when comparing configured
bps limits, and pass those same limits to Intent.swap with the bps-native keywords.
Explicit-pool decimals contract (
twap): any call that suppliespool_addressdirectly must either passtoken0_decimals/token1_decimalsexplicitly OR have apool_reader_registrywired on the snapshot so the decimals can be resolved from pool metadata. There is no "WETH/USDC fallback" — the tick-to-price conversion needs the real decimals. Without either path, the call raisesValueErrorrather than returning a price that can be off by powers of ten for pools like WBTC/WETH (8/18) or USDC/USDT (6/6).lwapdoes not acceptpool_addressand is unaffected (it scans pools internally via the registry).
rate = market.lending_rate("aave_v3", "USDC", side="supply") # LendingRate
best = market.best_lending_rate("USDC", side="supply") # BestRateResult
fr = market.funding_rate("binance", "ETH-PERP") # FundingRate
spread = market.funding_rate_spread("ETH-PERP", "binance", "hyperliquid") # FundingRateSpread
For permissionless, market-keyed protocols (morpho_blue) a market_id must
be verified on-chain before it is pinned into a config — a same-pair market
can carry a hostile oracle or IRM.
# List candidates from the curated catalog (verified=False, source="curated_catalog").
# Returns EVERY match — the gateway never ranks or auto-picks.
markets = market.lending_markets("morpho_blue", collateral="sUSDe", loan="USDC") # list[LendingMarketInfo]
# Verify one exact market_id on-chain (verified=True, source="onchain_verify").
# The gateway reads idToMarketParams(id), recomputes the id and compares;
# a mismatch / missing market raises LendingMarketResolutionError.
info = market.lending_market("morpho_blue", "0x...") # LendingMarketInfo
Use these at build or boot time to verify, not per-iteration to select. An empty
lending_markets(...)result means zero candidates matched — treat that as an error to act on, never a silent continue. Config still PINS the market identity; these accessors re-check it.The same universe is browsable from the CLI:
almanak ax --chain ethereum lending-reserves --protocol morpho_blue --collateral sUSDe --loan USDC
il = market.il_exposure("position_id", fees_earned=Decimal("50")) # ILExposure
proj = market.projected_il("WETH", "USDC", price_change_pct=Decimal("0.1")) # ProjectedILResult
mkt = market.prediction("market_id") # PredictionMarket
price = market.prediction_price("market_id", "YES") # Decimal
positions = market.prediction_positions("market_id") # list[PredictionPosition]
orders = market.prediction_orders("market_id") # list[PredictionOrder]
hist = market.lending_rate_history("aave_v3", "USDC", days=90) # DataEnvelope[list[LendingRateSnapshot]]
for snap in hist.value:
print(f"Supply: {snap.supply_apy}%, Borrow: {snap.borrow_apy}%")
fh = market.funding_rate_history("binance", "ETH-PERP", hours=168) # DataEnvelope[list[FundingRateSnapshot]]
ph = market.position_health("morpho_blue", market_id="0x...") # PositionHealth
ph.health_factor # Decimal
ph.ltv # Decimal
pt = market.pt_position_health("0x...", principal_token_market_address="0x...") # PTPositionHealth
rate = market.lst_exchange_rate("jitoSOL") # LSTExchangeRate
rate.rate # Decimal - rate vs SOL
rate.apy # Decimal
all_rates = market.lst_all_rates() # dict[str, LSTExchangeRate]
vol = market.realized_vol("WETH", window_days=30) # RealizedVol
cone = market.vol_cone("WETH") # VolCone
# portfolio_risk / rolling_sharpe consume a periodic PnL return series
# (each element is a fractional period return — e.g. 0.01 = 1% gain).
risk = market.portfolio_risk(pnl_series) # PortfolioRisk
sharpe = market.rolling_sharpe(pnl_series, window_days=30) # RollingSharpe
yields = market.yield_opportunities("USDC", min_tvl=100_000, sort_by="apy") # DataEnvelope[list[YieldOpportunity]]
gas = market.gas_price() # GasPrice
health = market.health() # HealthReport
signals = market.wallet_activity(action_types=["SWAP", "LP_OPEN"]) # list
market.chain # str - current chain name
market.wallet_address # str - wallet address
market.timestamp # datetime - snapshot timestamp
market.fork_rpc_url # str | None - Local Anvil fork RPC URL (paper trading only; bypasses the gateway and is None in production)
market.fork_block # int | None - current fork block number (paper trading only)
The runner uses these methods to detect when a strategy returns HOLD while market-data lookups
were failing (e.g. price oracle timeouts, unknown tokens), and escalates those cycles into
IterationStatus.DATA_ERROR so the consecutive-error circuit breaker fires correctly.
market.has_critical_data_failures() # bool - True if any data lookup failed this cycle
market.critical_data_failure_count() # int - number of tracked failures
market.classify_critical_data_failures() # str - "transient", "permanent", "mixed", or "none"
market.summarize_critical_data_failures() # str - human-readable summary (for logs)
market.clear_critical_data_failures() # None - reset all failures (called by runner after pre-warm)
The framework automatically persists runner-level metadata (iteration counts, error counters,
multi-step execution progress) after each iteration. However, strategy-specific state --
position IDs, trade counts, phase tracking, cooldown timers -- is only persisted if you implement
two hooks: get_persistent_state() and load_persistent_state().
Without these hooks, all instance variables are lost on restart. This is especially dangerous for LP and lending strategies where losing a position ID means the strategy cannot close its own positions.
Required for any stateful strategy:
def __init__(self, **kwargs):
super().__init__(**kwargs)
self._position_id: int | None = None
self._phase: str = "idle"
self._entry_price: Decimal = Decimal("0")
def get_persistent_state(self) -> dict:
"""Called by framework after each iteration to serialize state for persistence."""
return {
"position_id": self._position_id,
"phase": self._phase,
"entry_price": str(self._entry_price), # Decimal -> str for JSON
}
def load_persistent_state(self, saved: dict) -> None:
"""Called by framework on startup to restore state from previous run."""
self._position_id = saved.get("position_id")
self._phase = saved.get("phase", "idle")
self._entry_price = Decimal(saved.get("entry_price", "0"))
Guidelines:
.get() with defaults in load_persistent_state() so older saved state doesn't
crash when you add new fields.Decimal values as strings (str(amount)) and parse back (Decimal(state["amount"]))
for safe JSON round-tripping. All values must be JSON-serializable.on_intent_executed() callback is the natural place to update state after a trade (e.g.,
storing a new position ID), and get_persistent_state() then picks it up for saving.Use --fresh to clear saved state when starting over: almanak strat run --fresh --once.
After execution, access results (position IDs, swap amounts) via the callback. The framework
automatically enriches result with protocol-specific data - no manual receipt parsing needed.
# In your strategy file, import logging at the top:
# import logging
# logger = logging.getLogger(__name__)
def on_intent_executed(self, intent, success: bool, result):
if not success:
logger.warning(f"Intent failed: {intent.intent_type}")
return
# Capture LP position ID (enriched automatically by ResultEnricher)
# Store in instance variables -- persisted via get_persistent_state()
if result.position_id is not None:
self._lp_position_id = result.position_id
logger.info(f"Opened LP position {result.position_id}")
# Store range bounds for rebalancing strategies (keep as Decimal)
if (
hasattr(intent, "range_lower") and intent.range_lower is not None
and hasattr(intent, "range_upper") and intent.range_upper is not None
):
self._range_lower = intent.range_lower
self._range_upper = intent.range_upper
# Capture swap amounts
if result.swap_amounts:
self._last_swap = {
"amount_in": str(result.swap_amounts.amount_in),
"amount_out": str(result.swap_amounts.amount_out),
}
logger.info(
f"Swapped {result.swap_amounts.amount_in} -> {result.swap_amounts.amount_out}"
)
Contains the target chain and tunable runtime parameters. name, description, and
supported_chains still live in the @almanak_strategy decorator on your strategy class; the
config.json chain field acts as an explicit override of the decorator's default_chain and
lets tooling (sdk-planner, operators, deployment UIs) read the target chain without importing
the strategy module.
Single-chain:
{
"chain": "arbitrum",
"base_token": "WETH",
"quote_token": "USDC",
"rsi_period": 14,
"rsi_oversold": 30,
"rsi_overbought": 70,
"trade_size_usd": 1000,
"max_slippage_bps": 50,
"anvil_funding": {
"0xaf88d065e77c8cc2239327c5edb3a432268e5831": "10000",
"0x82af49447d8a07e3bd95bd0d56f35241523fbab1": "5"
}
}
Multi-chain:
{
"chains": ["base", "arbitrum"],
"swap_amount_usdc": "100",
"max_slippage_bps": 100,
"anvil_funding": {
"base": {
"0x833589fcd6edb6e08f4c7c32d4f71b54bda02913": 500
},
"arbitrum": {
"0xaf88d065e77c8cc2239327c5edb3a432268e5831": 500
}
}
}
The chains field lists the chains the strategy operates on and is read by the platform at
deployment time. It should match supported_chains from the @almanak_strategy decorator.
For single-chain strategies, chain (singular) is also accepted. A flat
anvil_funding object applies to all managed chains; use per-chain sections for
multi-chain strategies because ERC-20 contract addresses are chain-specific.
An optional "network" field ("mainnet" or "anvil") sets the default network for
almanak strat run when no --network flag is passed (local runs only; the flag always wins).
All other fields are strategy-specific and accessed via self.config.get(key, default).
Security note: The
.envfile is for local development and Anvil fork testing only. In production, secrets are managed by the platform and injected into the gateway sidecar - they never reach the strategy container. See Security Model.
# Required for local development
ALMANAK_PRIVATE_KEY=<your-private-key>
# RPC access (set at least one)
ALCHEMY_API_KEY=<your-alchemy-key>
# RPC_URL=https://...
# Optional
# ENSO_API_KEY=<key>
# COINGECKO_API_KEY=<key>
# ALMANAK_API_KEY=<key>
Structured list declaring exactly which tokens the strategy needs to be funded before the first tick.
Each entry specifies the token symbol, on-chain address, amount, and how to interpret the amount.
strat new generates this automatically when the template includes token fields.
"token_funding": [
{"symbol": "WETH", "address": "0x82aF49447D8a07e3bd95BD0d56f35241523fBab1", "chain": "arbitrum", "amount": "1", "amount_type": "token"},
{"symbol": "USDC", "address": "0xaf88d065e77c8cC2239327C5EDb3A432268e5831", "amount": "5000", "amount_type": "usd"}
]
| Field | Required | Description |
|---|---|---|
symbol | yes | Token symbol (e.g. "WETH") |
address | yes | ERC-20 contract address |
chain | no | Defaults to strategy chain |
amount | yes | Quantity (string to preserve precision) |
amount_type | yes | "token" (native units), "usd" (dollar value), or "percentage" (of held balance) |
When running on Anvil (--network anvil), the framework auto-funds the wallet
with tokens specified in anvil_funding. ERC-20 keys must be exact chain-specific
contract addresses; symbols are display metadata and are rejected. Use the shared
EVM native sentinel 0xEeeeeEeeeEeEeeEeEeEeeEEEeeeeEeeeeeeeEEeE for the active
chain's native gas asset.
Values are in token units (not USD). Funding failure aborts managed gateway startup.
Use get_token_resolver() for all token lookups. Never hardcode addresses.
Symbol references are deprecated. Symbols are metadata, not stable asset identity — the same ticker maps to a different contract on every chain and is spoofable.
TokenResolver,MarketSnapshot, andIntentconstruction emit aSymbolTokenResolutionWarning(aFutureWarning) when handed a bare symbol, once per external callsite. Symbols keep working for the rest of the 2.x line and are rejected in Almanak SDK 3.0.0 withSymbolTokenResolutionError. When writing new strategies, prefer a chain-specific contract address or a CAIP-19 asset identifier; resolve the symbol once at config time rather than per-iteration.
from almanak.framework.data.tokens import get_token_resolver
resolver = get_token_resolver()
# Preferred: resolve by address (stable identity)
token = resolver.resolve("0xaf88d065e77c8cC2239327C5EDb3A432268e5831", "arbitrum")
# -> ResolvedToken(symbol="USDC", address="0xaf88...", decimals=6, chain="arbitrum")
# Preferred: resolve by CAIP-19 asset identifier
token = resolver.resolve_caip19("eip155:42161/erc20:0xaf88d065e77c8cC2239327C5EDb3A432268e5831")
# Deprecated: resolve by symbol (warns now, raises in 3.0.0)
token = resolver.resolve("USDC", "arbitrum")
# Convenience
decimals = resolver.get_decimals("arbitrum", "USDC") # -> 6
address = resolver.get_address("arbitrum", "USDC") # -> "0xaf88..."
# For DEX swaps (auto-wraps native tokens: ETH->WETH, MATIC->WMATIC)
token = resolver.resolve_for_swap("ETH", "arbitrum") # -> WETH
# Resolve trading pair
usdc, weth = resolver.resolve_pair("USDC", "WETH", "arbitrum")
# Check whether a value already carries address-based identity
from almanak.framework.data.tokens.deprecation import is_address_based_token_reference
is_address_based_token_reference("USDC", "arbitrum") # -> False
Resolution order: memory cache -> disk cache -> static registry -> gateway on-chain lookup.
Never default to 18 decimals. If the token is unknown, TokenNotFoundError is raised.
almanak strat backtest pnl -s my_strategy \
--start 2024-01-01 --end 2024-06-01
Starting capital comes from the token_funding list in the strategy's
config.json (see Configuration) — PnL backtests fail with a
usage error when it is missing.
almanak strat backtest paper start -s my_strategy \
--chain arbitrum --duration 1h --tick-interval 60
Paper trading runs the full strategy loop on an Anvil fork with real transaction execution, equity curve tracking, and JSON result logs.
almanak strat backtest sweep -s my_strategy \
--start 2024-01-01 --end 2024-06-01 \
--param "rsi_oversold:20,25,30" \
--param "rsi_overbought:70,75,80"
Runs the PnL backtest across all parameter combinations and ranks by Sharpe ratio.
from datetime import UTC, datetime
from almanak.framework.backtesting import PnLBacktestConfig, PnLBacktester
config = PnLBacktestConfig(
start_time=datetime(2024, 1, 1, tzinfo=UTC),
end_time=datetime(2024, 6, 1, tzinfo=UTC),
token_funding=[
{
"symbol": "USDC",
"address": "0xaf88d065e77c8cC2239327C5EDb3A432268e5831",
"chain": "arbitrum",
"amount": "10000",
"amount_type": "usd",
}
],
)
backtester = PnLBacktester(data_provider, fee_models, slippage_models)
result = await backtester.backtest(strategy, config)
result.metrics.sharpe_ratio
result.metrics.max_drawdown_pct
result.metrics.total_return_pct
The CLI (almanak strat backtest pnl) wires the data provider and fee/slippage
models for you — prefer it unless you need custom providers.
backtest pnl, backtest paper, backtest sweep) require an explicit -s strategy_name flag. They do not auto-discover strategies from the current directory like strat run does.total_return_pct, annualized_return_pct, and max_drawdown_pct are actual percentages (33 = 33%). Ratio fields such as win_rate remain decimal fractions (0.33 = 33%).almanak strat new # Interactive scaffolding (creates pyproject.toml, .venv/, uv.lock)
almanak strat new -t ta_swap -n my_rsi -c arbitrum # Non-interactive
almanak strat demo # Browse and copy a working demo strategy
Templates:
| Template | Description |
|---|---|
blank | Minimal scaffold for custom implementations |
ta_swap | Technical-analysis swap strategy (RSI, Bollinger Bands, or combined signals) |
dynamic_lp | Price-based LP range management with rebalancing |
lending_loop | Supply/borrow leverage loop with state machine and health monitoring |
basis_trade | Spot+perp delta-neutral funding-rate arbitrage |
vault_yield | ERC-4626 vault deposit/redeem yield strategy |
copy_trader | Monitor leader wallets and replicate trades |
perps | Perpetual futures with take-profit / stop-loss |
multi_step | Atomic multi-step operations using IntentSequence (e.g., LP rebalancing) |
staking | Liquid staking with optional pre-swap |
Each scaffolded strategy is a self-contained Python project. After scaffolding, uv sync runs
automatically to create .venv/ and uv.lock. Add dependencies with uv add <package>.
almanak strat run --once # Single iteration (from strategy dir)
almanak strat run -d path/to/strat --once # Explicit directory
almanak strat run --network anvil --once # Local Anvil fork
almanak strat run --interval 30 # Continuous (30s between iterations)
almanak strat run --dry-run --once # No transactions submitted
almanak strat run --fresh --once # Clear state before running
almanak strat run --dashboard # Launch live monitoring dashboard
There is no --id flag: the deployment ID is derived deterministically from the
wallet address and chain, so re-running the same strategy resumes the same
deployment automatically (use --fresh to start over instead).
almanak strat backtest pnl -s my_strategy # Historical PnL simulation
almanak strat backtest paper start -s my_strategy # Paper trading on Anvil fork
almanak strat backtest sweep -s my_strategy # Parameter sweep optimization
almanak strat teardown request -s <deployment_id> # Request teardown (runner unwinds on its loop)
almanak strat teardown status -s <deployment_id> # Check teardown progress
almanak strat teardown execute # Execute directly from the strategy working directory
almanak strat permissions # Zodiac Roles Target[] format (default)
almanak strat permissions -o permissions.json # Write to file
almanak strat permissions -d path/to/strat # Explicit directory
almanak strat permissions --chain base # Override chain
Generates a JSON manifest of minimum-privilege contract permissions needed for Safe wallet deployments with Zodiac Roles. Reads supported_protocols and intent_types from @almanak_strategy metadata and compiles synthetic intents to discover required contract addresses and function selectors. Non-EVM chains are automatically skipped. The default output format is Zodiac Roles Target[].
almanak gateway # Start standalone gateway
almanak gateway --network anvil # Gateway for local Anvil testing
almanak gateway --port 50052 # Custom port
almanak agent install # Auto-detect platforms and install
almanak agent install -p claude # Install for specific platform
almanak agent install -p all # Install for all 10 platforms
almanak agent update # Update installed skill files
almanak agent status # Check installation status
almanak strat list # List deployed strategies
almanak strat status # Show strategy status
almanak strat logs # View strategy logs
almanak strat pause # Pause a running strategy
almanak strat resume # Resume a paused strategy
almanak copy validate # Validate a copy trading config
almanak copy replay # Replay copy trades
almanak copy report # Generate copy trading report
almanak ax # Almanak agentic execution
almanak backtest-service # Start backtest service
almanak dashboard # Launch strategy dashboard
almanak mcp serve # Start MCP server
almanak info matrix # Show chain/protocol support matrix
almanak docs path # Path to bundled LLM docs
almanak docs dump # Print full LLM docs
almanak docs agent-skill # Path to bundled agent skill
almanak docs agent-skill --dump # Print agent skill content
Every strategy deployed on a Safe wallet uses Zodiac Roles to enforce minimum-privilege access. The permissions system automatically discovers which contracts and function selectors the strategy needs by compiling synthetic intents.
Regenerate permissions whenever you:
@almanak_strategy(supported_protocols=[...])@almanak_strategy(intent_types=[...])config.json (base_token, quote_token, collateral_token, etc.)@almanak_strategy(supported_chains=[...])supported_protocols and intent_types from the @almanak_strategy() decoratorapprove permissions for tokens found in config.json# Generate Zodiac permissions and write to file (recommended)
almanak strat permissions -o permissions.json
# Preview on stdout
almanak strat permissions
# Single chain override
almanak strat permissions --chain arbitrum -o permissions.json
The Zodiac Roles Target[] format is a JSON array ready for Safe wallet configuration:
[
{
"address": "0x68b3465833fb72A70ecDF485E0e4C7bD8665Fc45",
"clearance": 2,
"executionOptions": 0,
"functions": [
{ "selector": "0x04e45aaf", "wildcarded": true }
]
}
]
clearance: 2 = function-level (specific selectors), 1 = target-level (all functions)executionOptions: 0 = None, 1 = Send, 2 = DelegateCall, 3 = Bothwildcarded: true means the selector applies regardless of input argumentsFor permissions to generate correctly, ensure your @almanak_strategy decorator declares all protocols and intent types:
@almanak_strategy(
name="my_strategy",
default_chain="arbitrum",
supported_chains=["arbitrum", "base"],
supported_protocols=["uniswap_v3", "aave_v3"],
intent_types=["SWAP", "SUPPLY", "WITHDRAW", "BORROW", "REPAY"],
)
class MyStrategy(IntentStrategy):
...
| Chain | Config Name |
|---|---|
| Ethereum | ethereum |
| Arbitrum | arbitrum |
| Optimism | optimism |
| Base | base |
| Avalanche | avalanche |
| Polygon | polygon |
| BSC | bsc |
| Sonic | sonic |
| Plasma | plasma |
| Blast | blast |
| Linea | linea |
| Mantle | mantle |
| Berachain | berachain |
| Monad | monad |
| X-Layer | xlayer |
| 0G Chain | zerog |
| HyperEVM | hyperevm |
| Robinhood Chain | robinhood |
| Solana | solana |
| Protocol | Enum Value | Type | Config Name |
|---|---|---|---|
| Uniswap V3 | UNISWAP_V3 | DEX / LP | uniswap_v3 |
| Uniswap V4 | UNISWAP_V4 | DEX / LP | uniswap_v4 |
| PancakeSwap V3 | PANCAKESWAP_V3 | DEX / LP | pancakeswap_v3 |
| SushiSwap V3 | SUSHISWAP_V3 | DEX / LP | sushiswap_v3 |
| TraderJoe V2 | TRADERJOE_V2 | DEX / LP | traderjoe_v2 |
| Aerodrome | AERODROME | DEX / LP | aerodrome |
| Agni Finance | AGNI_FINANCE | DEX / LP | agni_finance |
| Enso | ENSO | Aggregator | enso |
| Pendle | PENDLE | Yield | pendle |
| MetaMorpho | METAMORPHO | Lending | metamorpho |
| LiFi | LIFI | Bridge | lifi |
| BenQi | BENQI | Lending | benqi |
| Joe Lend (DORMANT) | JOE_LEND | Lending — wound down on-chain (VIB-3960); compiler short-circuits. Do NOT use. | joelend |
| Silo V2 | SILO_V2 | Lending | silo_v2 |
| Euler V2 | EULER_V2 | Lending | euler_v2 |
| Vault | VAULT | ERC-4626 | vault |
| Curve | CURVE | DEX / LP | curve |
| Balancer V2 | BALANCER | DEX / LP | balancer_v2 |
| Aave V3 | * | Lending | aave_v3 |
| Morpho Blue | * | Lending | morpho_blue |
| Compound V3 | * | Lending | compound_v3 |
| GMX V2 | * | Perps | gmx_v2 |
| Hyperliquid | * | Perps | hyperliquid |
| Polymarket | * | Prediction | polymarket |
| Kraken | * | CEX | kraken |
| Lido | * | Staking | lido |
| Lagoon | * | Vault | lagoon |
* These protocols do not have a Protocol enum value. Use the string config name (e.g., protocol="aave_v3") in intents. They are resolved by the intent compiler and transaction builder directly.
The connector descriptor registry is the post-VIB-4298 source of truth. Connectors below have full SDK + adapter + receipt-parser implementations and are routable via the protocol="<name>" string in intents, but do not have a top-level Protocol enum entry. Use the canonical name listed in Config Name:
| Connector | Type | Config Name | Chains |
|---|---|---|---|
| Across | Bridge | across | Ethereum, Arbitrum, Optimism, Base, Polygon, Linea |
| Aster Perps | Perp | aster_perps | BNB Chain |
| Camelot | DEX | camelot | Arbitrum |
| Curvance | Lending | curvance | Monad |
| Drift | Perp | drift | Solana |
| Ethena | Staking | ethena | Ethereum |
| Fluid | DEX | fluid | Ethereum, Arbitrum, Base, Polygon |
| Fluid DEX LP | DEX / LP | fluid_dex_lp | Arbitrum |
| Fluid Vault | Lending | fluid_vault | Arbitrum, Base |
| Gimo | Staking | gimo | 0G |
| Jupiter | DEX | jupiter | Solana |
| Jupiter Lend | Lending | jupiter_lend | Solana |
| Kamino | Lending | kamino | Solana |
| Meteora | DEX | meteora | Solana |
| Morpho Vault | Vault | morpho_vault | Ethereum, Base |
| Orca | DEX | orca | Solana |
| PancakeSwap Perps | Perp | pancakeswap_perps | BNB Chain |
| Raydium | DEX | raydium | Solana |
| Spark | Lending | spark | Ethereum |
| Stargate | Bridge | stargate | Ethereum, Arbitrum, Optimism, Base, Polygon, BNB Chain, Avalanche |
For the full machine-readable inventory (every connector, every intent it implements, every chain it ships on), see the generated matrix at docs/api/connectors/index.md.
| Network | Enum Value | Description |
|---|---|---|
| Mainnet | MAINNET | Production chains |
| Anvil | ANVIL | Local fork for testing |
| Sepolia | SEPOLIA | Testnet |
GMX V2 (Perpetuals)
"BTC/USD", "ETH/USD", "LINK/USD" (not dash).Intent.perp_open(), the SDK submits an order creation transaction. A GMX keeper then executes the actual position change in a separate transaction. on_intent_executed(success=True) fires when the order creation TX confirms, not when the keeper executes the position. Strategies should poll position state before relying on it.get_all_positions() may not return positions immediately after opening due to keeper delay. Allow a few seconds before querying.def decide(self, market):
rsi = market.rsi(self.base_token, period=self.rsi_period)
quote_bal = market.balance(self.quote_token)
base_bal = market.balance(self.base_token)
if rsi.is_oversold and quote_bal.balance_usd >= self.trade_size:
return Intent.swap(
from_token=self.quote_token, to_token=self.base_token,
amount_usd=self.trade_size, max_slippage=Decimal("0.005"),
)
if rsi.is_overbought and base_bal.balance_usd >= self.trade_size:
return Intent.swap(
from_token=self.base_token, to_token=self.quote_token,
amount_usd=self.trade_size, max_slippage=Decimal("0.005"),
)
return Intent.hold(reason=f"RSI={rsi.value:.1f} in neutral zone")
Use
pool_price/pool_price_by_pair, nevermarket.price(), for concentrated-LP range bounds or a range-exit test (VIB-exp19).market.price()is a USD valuation oracle — it returns a hardcoded1.0for stablecoins (source: stablecoin_peg) and is not guaranteed to match the pool's actual price for any pair. A range centered on it, or a range-exit check built from it, can silently mint out of range (zero fees, no error) — and for a stable pair pinned at oracle1.0, the range-exit check below becomes a structural no-op (spotnever moves, so it's always "in range") even while the real pool has drifted outside the position.pool_price_by_pairreturns the price in the pool's own on-chain token0/token1 order, which may not match yourbase/quoteorder — verify orientation against a known price before trusting the sign, or usepool_price(pool_address, chain)when you have the address. See Blueprint 03 §"LP_OPEN range-excludes-spot warning" for the compile-time guard that now catches this class of bug at LP_OPEN time.
def decide(self, market):
price = market.pool_price_by_pair(self.base_token, self.quote_token).price
position_id = self._lp_position_id
if position_id:
# Check if price is out of range - close and reopen
if price < self._range_lower or price > self._range_upper:
return Intent.lp_close(position_id=position_id, protocol="uniswap_v3")
# Open new position centered on current pool price
atr = market.atr(self.base_token)
half_range = price * (atr.value_percent / Decimal("100")) * 2 # value_percent is percentage points
return Intent.lp_open(
pool="WETH/USDC",
amount0=Decimal("1"), amount1=Decimal("2000"),
range_lower=price - half_range,
range_upper=price + half_range,
)
def decide(self, market):
return Intent.sequence(
intents=[
Intent.swap(from_token="USDC", to_token="WETH", amount_usd=Decimal("5000")),
Intent.supply(protocol="aave_v3", token="WETH", amount="all"),
Intent.borrow(
protocol="aave_v3",
collateral_token="WETH", collateral_amount=Decimal("0"),
borrow_token="USDC", borrow_amount=Decimal("3000"),
),
],
description="Leverage loop: buy WETH, supply, borrow USDC",
)
Strategies can operate across multiple chains with per-chain wallet addresses.
The primary chain is supported_chains[0]. Intents without an explicit chain=
parameter run on the primary chain.
Decorator:
@almanak_strategy(
name="cross_chain_arb",
supported_chains=["base", "arbitrum"], # base is primary (first in list)
supported_protocols=["uniswap_v3", "across"],
intent_types=["SWAP", "BRIDGE", "HOLD"],
)
class CrossChainArbStrategy(IntentStrategy):
...
config.json:
{
"chains": ["base", "arbitrum"],
"swap_amount_usdc": "100",
"anvil_funding": {
"base": {
"0x833589fcd6edb6e08f4c7c32d4f71b54bda02913": 500
},
"arbitrum": {
"0xaf88d065e77c8cc2239327c5edb3a432268e5831": 500
}
}
}
Use per-chain funding sections for multi-chain strategies. A flat object remains valid for a single chain (or an address intentionally shared across every fork).
Strategy properties:
self.chain # "base" (primary chain = supported_chains[0])
self.chains # ["base", "arbitrum"]
self.wallet_address # default wallet
self.get_wallet_for_chain("arbitrum") # per-chain wallet (if wallet registry configured)
When a gateway wallet registry is configured (ALMANAK_GATEWAY_WALLETS), each
chain can use a different Safe wallet. The framework resolves destination wallets
automatically for bridge intents.
decide() with cross-chain intents:
def decide(self, market: MarketSnapshot):
return Intent.sequence([
# Bridge USDC from Base to Arbitrum
Intent.bridge(
token="USDC",
amount=Decimal("100"),
from_chain="base",
to_chain="arbitrum",
preferred_bridge="across",
max_slippage=Decimal("0.01"),
),
# Swap on Arbitrum (explicit chain= required for non-primary chain)
Intent.swap(
from_token="USDC",
to_token="WETH",
amount=Decimal("50"),
protocol="uniswap_v3",
chain="arbitrum",
),
# Bridge back to primary chain
Intent.bridge(
token="USDC",
amount=Decimal("50"),
from_chain="arbitrum",
to_chain="base",
preferred_bridge="across",
),
], description="Arb USDC across chains")
Multi-chain market data:
For multi-chain strategies, market is a MultiChainMarketSnapshot with
chain-aware queries:
def decide(self, market):
# Chain-specific prices and balances
arb_price = market.price("WETH", chain="arbitrum")
base_price = market.price("WETH", chain="base")
usdc_on_base = market.balance("USDC", chain="base")
# Chain health monitoring
market.healthy_chains # ["base", "arbitrum"]
market.stale_chains # [] (empty if all healthy)
market.all_chains_healthy # True
Key rules:
chain= -- it's implicitchain="arbitrum" etc.from_chain and to_chainIntent.sequence() to order cross-chain operationsamount="all" chaining does not work after bridge intents (bridge receipt
parsers don't extract output amounts) -- use explicit amounts insteadfrom almanak.framework.alerting import AlertManager
class MyStrategy(IntentStrategy):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.alerts = AlertManager.from_config(self.config.get("alerting", {}))
def decide(self, market):
rsi = market.rsi("WETH")
if rsi.value < 20:
self.alerts.send("Extreme oversold: RSI={:.1f}".format(rsi.value), level="warning")
# ... trading logic
All IntentStrategy subclasses must implement two abstract teardown methods:
get_open_positions() and generate_teardown_intents(). Without these, the
strategy class cannot be instantiated.
For strategies that never hold positions, extend StatelessStrategy instead of
IntentStrategy — it provides empty default implementations.
get_open_positions()Returns a TeardownPositionSummary describing all current positions. Must query
on-chain state (not cached) for safety:
def get_open_positions(self):
from datetime import UTC, datetime
from almanak.framework.teardown import PositionInfo, PositionType, TeardownPositionSummary
positions = []
try:
market = self.create_market_snapshot()
base_bal = market.balance(self.base_token)
if base_bal.balance > 0:
positions.append(
PositionInfo(
position_type=PositionType.TOKEN,
position_id=f"{self.base_token}-holding",
chain=self.chain,
protocol="uniswap_v3",
value_usd=base_bal.balance_usd,
details={"asset": self.base_token, "amount": str(base_bal.balance)},
)
)
except Exception:
logger.warning("Unable to fetch balances for teardown position summary")
return TeardownPositionSummary(
deployment_id=getattr(self, "deployment_id", "my_strategy"),
timestamp=datetime.now(UTC),
positions=positions,
)
PositionType values (close in this priority order):
PERP > BORROW > SUPPLY > LP > STAKE > PREDICTION > CEX > TOKEN
For strategies with no positions, return TeardownPositionSummary.empty(self.deployment_id).
Perpetual strategies must preserve the perps scaffold's normalized venue
probe in both teardown methods:
from almanak.framework.strategies import probe_perp_position
probe = probe_perp_position(
market,
protocol=self.protocol,
chain=self.chain,
market_symbol=self.market,
)
The two position models are not interchangeable:
market.perp_positions() returns raw PerpsPositionOnChain. Its
size_in_usd and size_in_tokens are venue-scaled integers intended for
connector/framework code. It does not have notional_usd.probe_perp_position() returns normalized PerpProbePosition rows. Only
these rows expose strategy-facing notional_usd: Decimal | None.Never access position.notional_usd on a raw row, and never publish raw
size_in_usd as PositionInfo.value_usd. Handle the probe's three states:
OPEN uses venue positions and side; FLAT overrides stale cache; UNMEASURED keeps
persisted exposure with value_usd_unknown=True and
valuation_status="no_path". Every full close uses size_usd=None.
Retain a unit test that constructs the real SDK PerpsPositionOnChain and
passes it through get_open_positions() and generate_teardown_intents(). A
MagicMock or SimpleNamespace with an invented notional_usd field cannot catch
this boundary error.
generate_teardown_intents()Returns intents to close all positions, respecting priority order and teardown mode:
def generate_teardown_intents(self, mode, market=None) -> list[Intent]:
from almanak.framework.teardown import TeardownMode
max_slippage = Decimal("0.03") if mode == TeardownMode.HARD else Decimal("0.01")
intents = []
# Close LP positions first (if any)
position_id = self._lp_position_id
if position_id:
intents.append(Intent.lp_close(position_id=position_id))
# Swap all base token back to quote
intents.append(Intent.swap(
from_token=self.base_token, to_token=self.quote_token,
amount="all", max_slippage=max_slippage,
protocol="pancakeswap_v3",
swap_params={"pool": self.approved_pool},
))
return intents
TeardownMode.SOFT = graceful exit (minimize costs), TeardownMode.HARD = emergency (speed over cost).
When delegating to self.teardown_full_close_intents(), describe non-standard token holdings with
canonical address identity and their exact close route. token_address takes precedence over a
display symbol, and close_swap_params is copied into the generated SwapIntent. A pool pin
without protocol is skipped fail-closed rather than routed generically.
PositionInfo(
position_type=PositionType.TOKEN,
position_id="held-XAUT0",
chain="bsc",
protocol="pancakeswap_v3",
value_usd=xaut0_value_usd,
details={
"asset_symbol": "XAUT0",
"token_address": self.base_token_address,
"amount": str(xaut0_balance),
"close_swap_params": {"pool": self.approved_pool},
},
)
def get_open_positions(self):
from datetime import UTC, datetime
from almanak.framework.teardown import PositionInfo, PositionType, TeardownPositionSummary
positions = []
if self._borrowed_amount > 0:
positions.append(PositionInfo(
position_type=PositionType.BORROW,
position_id=f"aave-borrow-{self.borrow_token}",
chain=self.chain, protocol="aave_v3",
value_usd=self._borrowed_amount * self._borrow_price,
details={"asset": self.borrow_token, "amount": str(self._borrowed_amount)},
))
if self._supplied_amount > 0:
positions.append(PositionInfo(
position_type=PositionType.SUPPLY,
position_id=f"aave-supply-{self.collateral_token}",
chain=self.chain, protocol="aave_v3",
value_usd=self._supplied_amount,
details={"asset": self.collateral_token, "amount": str(self._supplied_amount)},
))
return TeardownPositionSummary(
deployment_id=self.deployment_id, timestamp=datetime.now(UTC), positions=positions,
)
def generate_teardown_intents(self, mode, market=None) -> list[Intent]:
intents = []
if self._borrowed_amount > 0:
intents.append(Intent.repay(
protocol="aave_v3", token=self.borrow_token,
amount=self._borrowed_amount, repay_full=True, chain=self.chain,
))
if self._supplied_amount > 0:
intents.append(Intent.withdraw(
protocol="aave_v3", token=self.collateral_token,
amount=self._supplied_amount, withdraw_all=True, chain=self.chain,
))
return intents
Let exceptions propagate from decide(). The framework catches them and feeds
them into its built-in circuit breaker, which tracks consecutive failures and
stops the strategy after a threshold is reached.
def decide(self, market):
rsi = market.rsi("WETH", period=14)
# ... strategy logic — no try/except needed
The framework retries each failed intent up to max_retries (default: 3) with
exponential backoff. However, after all retries are exhausted the strategy
continues running and will attempt the same trade on the next iteration.
Without a circuit breaker, this creates an infinite loop of reverted transactions
that burn gas without any hope of success.
Always track consecutive execution failures in persistent state and stop trading (or enter an extended cooldown) after a threshold is reached:
MAX_CONSECUTIVE_FAILURES = 3 # Stop after 3 rounds of failed intents
FAILURE_COOLDOWN_SECONDS = 1800 # 30-min cooldown before retrying
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.consecutive_failures = 0
self.failure_cooldown_until = 0.0
def decide(self, market):
try:
now = time.time()
# Circuit breaker: skip trading while in cooldown
if now < self.failure_cooldown_until:
remaining = int(self.failure_cooldown_until - now)
return Intent.hold(
reason=f"Circuit breaker active, cooldown {remaining}s remaining"
)
# Circuit breaker: enter cooldown after too many failures
if self.consecutive_failures >= MAX_CONSECUTIVE_FAILURES:
self.failure_cooldown_until = now + FAILURE_COOLDOWN_SECONDS
self.consecutive_failures = 0
logger.warning(
f"Circuit breaker tripped after {MAX_CONSECUTIVE_FAILURES} "
f"consecutive failures, cooling down {FAILURE_COOLDOWN_SECONDS}s"
)
return Intent.hold(reason="Circuit breaker tripped")
# ... normal strategy logic ...
except Exception as e:
logger.exception(f"Error in decide(): {e}")
return Intent.hold(reason=f"Error: {e}")
def on_intent_executed(self, intent, success: bool, result):
if success:
self.consecutive_failures = 0 # Reset on success
else:
self.consecutive_failures += 1
logger.warning(
f"Intent failed ({self.consecutive_failures}/{MAX_CONSECUTIVE_FAILURES})"
)
def get_persistent_state(self) -> dict:
return {
"consecutive_failures": self.consecutive_failures,
"failure_cooldown_until": self.failure_cooldown_until,
}
def load_persistent_state(self, state: dict) -> None:
self.consecutive_failures = int(state.get("consecutive_failures", 0))
self.failure_cooldown_until = float(state.get("failure_cooldown_until", 0))
Important: Only update trade-timing state (e.g. last_trade_ts) inside
on_intent_executed when success=True, not when the intent is created. Setting
it at creation time means a failed trade still resets the interval timer, causing
the strategy to wait before retrying — or worse, to keep retrying on a fixed
schedule with no failure awareness.
Override on_sadflow_enter to react to specific error types during intent
retries. This hook is called before each retry attempt and lets you modify the
transaction (e.g. increase gas or slippage) or abort early:
from almanak.framework.intents.state_machine import SadflowAction
class MyStrategy(IntentStrategy):
def on_sadflow_enter(self, error_type, attempt, context):
# Abort immediately on insufficient funds — retrying won't help
if error_type == "INSUFFICIENT_FUNDS":
return SadflowAction.abort("Insufficient funds, stopping retries")
# Increase gas limit for gas-related errors
if error_type == "GAS_ERROR" and context.action_bundle:
modified = self._increase_gas(context.action_bundle)
return SadflowAction.modify(modified, reason="Increased gas limit")
# For slippage errors ("Too little received"), abort after 1 attempt
# since retrying with the same parameters will produce the same result
if error_type == "SLIPPAGE" and attempt >= 1:
return SadflowAction.abort("Slippage error persists, aborting")
# Default: let the framework retry with backoff
return None
Error types passed to on_sadflow_enter (from _categorize_error in state_machine.py):
GAS_ERROR — gas estimation failed or gas limit exceededINSUFFICIENT_FUNDS — wallet balance too lowSLIPPAGE — "Too little received" or similar DEX revertTIMEOUT — transaction confirmation timed outNONCE_ERROR — nonce mismatch or conflictREVERT — generic transaction revertRATE_LIMIT — RPC or API rate limit hitNETWORK_ERROR — connection or network failureCOMPILATION_PERMANENT — unsupported protocol/chain (non-retriable)None — unclassified errorBefore deploying to mainnet:
--network anvil --once until decide() works correctly--dry-run --once on mainnet to verify compilation without submitting transactionsamount= (token units) for swaps if amount_usd= causes reverts (see swap reference above)get_persistent_state() / load_persistent_state() if your strategy tracks positions or phase statealmanak strat permissions -o permissions.json-s)| Error | Cause | Fix |
|---|---|---|
TokenNotFoundError | Token symbol not in registry | Use exact symbol (e.g., "WETH" not "ETH" for swaps). Check resolver.resolve("TOKEN", "chain"). |
Gateway not available | Gateway not running | Use almanak strat run (auto-starts gateway) or start manually with almanak gateway. |
ALMANAK_PRIVATE_KEY not set | Missing .env | Set your private key in .env (see Configuration section). |
Anvil not found | Foundry not installed | Install Foundry: see getfoundry.sh for instructions. |
RSI data unavailable | Insufficient price history | The gateway needs time to accumulate data. Try a longer timeframe or wait. |
Insufficient balance | Wallet doesn't have enough tokens | For Anvil: add anvil_funding to config.json. For mainnet: fund the wallet. |
Slippage exceeded | Trade too large or pool illiquid | Increase max_slippage or reduce trade size. |
Too little received (repeated reverts) | Placeholder prices used for slippage calculation, or stale price data | Ensure real price feeds are active (not placeholder). Implement on_sadflow_enter to abort on persistent slippage errors. Add a circuit breaker to stop retrying the same failing trade. |
| Transactions keep reverting after max retries | Strategy re-emits the same failing intent on subsequent iterations | Track consecutive_failures in persistent state and enter cooldown after a threshold. See the "Execution Failure Tracking" pattern. |
| Gas wasted on reverted transactions | No circuit breaker; framework retries 3x per intent, then strategy retries next iteration indefinitely | Implement on_intent_executed callback to count failures and on_sadflow_enter to abort non-recoverable errors early. |
| Intent compilation fails | Wrong parameter types | Ensure amounts are Decimal, not float. Use Decimal(str(value)). |
--verbose flag for detailed logging: almanak strat run --once --verbose--dry-run to test decide() without submitting transactions--log-file out.json for machine-readable JSON logsself.state persists between iterationsalmanak strat backtest paper start -s my_strategy runs real execution on AnvilFrequently asked questions
You are helping a quant build DeFi strategies using the Almanak SDK. Strategies are Python classes that return Intent objects. The framework handles compilation to transactions, execution, and state management.
The source record exposes this install command: npx skills add https://github.com/almanak-co/sdk --skill "almanak/skills/almanak-strategy-builder". Inspect the command and pinned source before running it.
Static rules flagged exec-script, network, write-files in the source; the page lists the matching lines and excerpts.
Alternatives
oaslananka/kicad-mcp-pro
Use this skill for GitHub Copilot pull request and code reviews in oaslananka/kicad-mcp-pro. Review Python MCP server changes, KiCad adapter and tool-contract changes, tests, npm/package wrappers, Tauri/Rust desktop code, GitHub Actions, security controls, documentation, generated metadata, and compatibility/release surfaces. Use it whenever reviewing a PR or diff in this repository, especially changes under src/, tests/, packages/, src-tauri/, .github/workflows/, or public MCP metadata/configur
brucesongs/kali-claw
Physical penetration testing covering mechanical lock bypass (pin-tubular/wafer), RFID/NFC badge cloning (Proxmark3/ESP-RFID-Tool/Walrus), HID iCLASS/Mifare duplication, drop box deployment (LAN Turtle/Packet Squirrel), USB weapons (Rubber Ducky/Bash Bunny), hidden camera placement, and on-site engagement operations including tailgating pretext preparation and physical-docs legal templates.
kdeldycke/repomatic
Monitor CI tests, lint, autofix, docs, and Nuitka binary-build workflows, diagnose failures, fix code, commit, and loop until all stable jobs pass. Ignores unstable failures.
UiPath/skills
Always invoke for `.xaml` or `.cs` workflow files. UiPath RPA — create, edit, build, run, debug `.cs` coded workflows and `.xaml` workflows. UI automation with Object Repository selectors, test case authoring, Integration Service connector calls. Live desktop/browser UI exploration and control. Deploy via `.uipx`→uipath-solution. Non-solution Orchestrator ops→uipath-platform. Test reports→uipath-test. Agents→uipath-agents.