feat: yaml files for llm reasoning
This commit is contained in:
Danny
parent
b0211b944d
commit
77260c3e08
|
|
@ -14,12 +14,14 @@ logger = logging.getLogger(__name__)
|
|||
|
||||
Intent = Literal["codebase", "general", "clarification"]
|
||||
|
||||
INTENT_PROMPT = """Classify this user message into one category:
|
||||
- "codebase": Questions about trading system code, architecture, files, methods, execution, strategies, exchanges, risk management, order handling, or technical implementation
|
||||
- "general": Greetings, meta-questions, off-topic ("How are you?", "What can you do?", "Hello")
|
||||
- "clarification": Follow-ups that rely on conversation context, not new retrieval ("Tell me more", "What did you mean?", "Can you explain that?")
|
||||
INTENT_PROMPT = """You are classifying questions for a Tyndale trading system documentation assistant.
|
||||
|
||||
IMPORTANT: If the user is asking about specific code, files, classes, methods, or system behavior, classify as "codebase".
|
||||
Classify this user message into one category:
|
||||
- "codebase": ANY question about trading, strategies, exchanges, orders, positions, risk, execution, hedging, market making, P&L, or how the system works. Also includes questions with "our", "the system", "this", or references to specific functionality.
|
||||
- "general": ONLY greetings or completely off-topic questions ("How are you?", "What's the weather?", "Hello")
|
||||
- "clarification": Follow-ups that reference previous answers ("Tell me more", "What did you mean?", "Can you explain that?")
|
||||
|
||||
DEFAULT TO "codebase" if uncertain. This is a trading system assistant - assume trading questions are about the codebase.
|
||||
|
||||
Respond with ONLY the category name, nothing else."""
|
||||
|
||||
|
|
@ -75,7 +77,7 @@ class IntentClassifier:
|
|||
|
||||
# Validate intent
|
||||
if raw_intent in ("codebase", "general", "clarification"):
|
||||
logger.debug(f"Classified intent: {raw_intent}")
|
||||
logger.info(f"Intent classified: '{message[:50]}...' -> {raw_intent}")
|
||||
return raw_intent
|
||||
|
||||
# Default to codebase for ambiguous cases (safer for RAG)
|
||||
|
|
|
|||
|
|
@ -0,0 +1,69 @@
|
|||
source_file: ./live.py
|
||||
schema_version: v1.1
|
||||
|
||||
factual_summary: >
|
||||
Provides the runtime framework for executing a market making strategy
|
||||
in either paper or live trading modes. Responsible for initializing
|
||||
exchanges, loading symbol universes, configuring execution components,
|
||||
processing real-time market data, and coordinating the continuous
|
||||
operation of the trading system.
|
||||
|
||||
methods:
|
||||
main() -> None:
|
||||
description: >
|
||||
Asynchronous entry point that orchestrates system initialization,
|
||||
strategy and trader setup, exchange connectivity, optional logging
|
||||
and metrics configuration, and continuous processing of incoming
|
||||
market data events.
|
||||
|
||||
interpretive_summary: >
|
||||
Acts as the top-level execution driver for the trading system, wiring
|
||||
together strategy logic, execution components, and exchange data feeds
|
||||
into a single long-running process. Designed to support both simulated
|
||||
and live trading environments while enabling observability and
|
||||
operational control during runtime.
|
||||
|
||||
invariants:
|
||||
- Trading mode must be explicitly selected before execution begins.
|
||||
- Exchanges must be initialized prior to processing market data.
|
||||
- Strategy and trader instances must be fully constructed before
|
||||
entering the main processing loop.
|
||||
- Market data processing must run continuously until the process
|
||||
terminates.
|
||||
- Logging and metrics configuration must not interfere with execution.
|
||||
|
||||
tags:
|
||||
domain:
|
||||
- market_data
|
||||
- strategy_execution
|
||||
- order_execution
|
||||
- exchange_integration
|
||||
|
||||
trading_function:
|
||||
- data_ingest
|
||||
- entry_logic
|
||||
- exit_logic
|
||||
|
||||
strategy_layer:
|
||||
- execution
|
||||
|
||||
system_layer:
|
||||
- worker
|
||||
|
||||
intent:
|
||||
- orchestration
|
||||
- isolation
|
||||
- safety
|
||||
|
||||
data_type:
|
||||
- signals
|
||||
- orders
|
||||
- exchanges
|
||||
|
||||
risk:
|
||||
- latency
|
||||
- data_corruption
|
||||
- capital_loss
|
||||
|
||||
maturity:
|
||||
- production
|
||||
|
|
@ -9,73 +9,73 @@ factual_summary: >
|
|||
backtesting modes.
|
||||
|
||||
methods:
|
||||
_getBracket(exch: str, sym: str) -> List[]:
|
||||
"_getBracket(exch: str, sym: str) -> List[]":
|
||||
description: >
|
||||
Retrieves the current bid and ask information for a given symbol
|
||||
on a specified exchange using maintained order book state.
|
||||
|
||||
startBidding(exch: str, sym: str, ts: float, entryPrice: float or None) -> None:
|
||||
"startBidding(exch: str, sym: str, ts: float, entryPrice: float or None) -> None":
|
||||
description: >
|
||||
Used in the Strat class to update the entry price of a symbol on a particular exchange. Logic for updating the
|
||||
entry price is contained there.
|
||||
|
||||
stopBidding(exch: str, sym: str, ts: float) -> None:
|
||||
"stopBidding(exch: str, sym: str, ts: float) -> None":
|
||||
description: >
|
||||
Halts the bidding by removing the order from the limitBuy dictionary.
|
||||
|
||||
updateBracket(exch: str, sym: str, bidQuote: float, askQuote: float) -> None:
|
||||
"updateBracket(exch: str, sym: str, bidQuote: float, askQuote: float) -> None":
|
||||
description: >
|
||||
Updates the bracketed strings in our brackets dictionary.
|
||||
|
||||
startOffering(exch: str, sym: str, ts: float, entryPrice: float or None) -> None:
|
||||
"startOffering(exch: str, sym: str, ts: float, entryPrice: float or None) -> None":
|
||||
description: >
|
||||
Executes the trades when we are going short.
|
||||
|
||||
stopOffering(exch: str, sym: str, ts: float) -> None:
|
||||
"stopOffering(exch: str, sym: str, ts: float) -> None":
|
||||
description: >
|
||||
Stops the short position.
|
||||
|
||||
logIntradayPnl(val: float) -> None:
|
||||
"logIntradayPnl(val: float) -> None":
|
||||
description: >
|
||||
Updates the value of our account to our pnl database
|
||||
|
||||
fpnl() -> None:
|
||||
"fpnl() -> None":
|
||||
description: >
|
||||
Returns a list of floating profit and loss summaries
|
||||
|
||||
posCounts() -> dict:
|
||||
"posCounts() -> dict":
|
||||
description: >
|
||||
Returns a python dictionary of the total positions we have on each exchange.
|
||||
|
||||
logFee(fee: float) -> None:
|
||||
"logFee(fee: float) -> None":
|
||||
description: >
|
||||
Updates our fees in our database
|
||||
|
||||
posValueAtExchOld(exch: str, sym: str) -> int:
|
||||
"posValueAtExchOld(exch: str, sym: str) -> int":
|
||||
description: >
|
||||
Calculates the values of the long and short positions.
|
||||
|
||||
fpnlnotional() -> int:
|
||||
"fpnlnotional() -> int":
|
||||
description: >
|
||||
This method calculates the profit/loss (P&L) based on the notional value of long and short positions for each
|
||||
unique symbol across different exchanges. It computes the average price of long positions, sums the notional
|
||||
values of both long and short positions, and then uses these to compute the P&L.
|
||||
|
||||
fillLong(exch: str, sym: str, price: float, quan: float, ts: int, tag: str, feeType: str) -> None:
|
||||
"fillLong(exch: str, sym: str, price: float, quan: float, ts: int, tag: str, feeType: str) -> None":
|
||||
description: >
|
||||
The fillLong method handles the process of filling a buy order for a specified asset on an exchange. It updates
|
||||
the trader's position with the new quantity and calculates the average price if necessary. If there was a short
|
||||
position, it also calculates the profit/loss from buying back some or all of the shorted amount. Additionally,
|
||||
it logs the trade, applies fees, and ensures data integrity by asserting non-negative values.
|
||||
|
||||
fillShort(exch: str, sym: str, price: float, quan: float, ts: int, tag: str, feeType: str) -> None:
|
||||
"fillShort(exch: str, sym: str, price: float, quan: float, ts: int, tag: str, feeType: str) -> None":
|
||||
description: >
|
||||
The fillShort method handles the process of filling a sell order for a specified asset on an exchange. It updates
|
||||
the trader's position with the new quantity and calculates the average price if necessary. If there was a long
|
||||
position, it also calculates the profit/loss from selling some or all of the held amount. Additionally, it logs
|
||||
the trade, applies fees, and ensures data integrity by asserting non-negative values.
|
||||
|
||||
hedgeLong(exch: str, sym: str, price: float, quan: float, ts: int, force: bool) -> Exchange:
|
||||
"hedgeLong(exch: str, sym: str, price: float, quan: float, ts: int, force: bool) -> Exchange":
|
||||
description: >
|
||||
The hedgeLong method finds an appropriate exchange to hedge a long position by executing a corresponding sell
|
||||
order. It iterates over possible exchanges, calculates adjusted bid prices based on offsets, and selects the
|
||||
|
|
@ -83,7 +83,7 @@ methods:
|
|||
ensures that the hedge does not exceed maximum allowed positions and logs relevant information throughout the
|
||||
process.
|
||||
|
||||
hedgeShort(exch: str, sym: str, price: float, quan: float, ts: int, force: bool) -> Exchange:
|
||||
"hedgeShort(exch: str, sym: str, price: float, quan: float, ts: int, force: bool) -> Exchange":
|
||||
description: >
|
||||
The hedgeShort method finds an appropriate exchange to hedge a short position by executing a corresponding buy
|
||||
order. It iterates over possible exchanges, calculates adjusted ask prices based on offsets, and selects the
|
||||
|
|
@ -91,7 +91,7 @@ methods:
|
|||
ensures that the hedge does not exceed maximum allowed positions and logs relevant information throughout the
|
||||
process.
|
||||
|
||||
processTrade(exch: str, sym: str, price: float, tradeDir: list) -> None:
|
||||
"processTrade(exch: str, sym: str, price: float, tradeDir: list) -> None":
|
||||
description: >
|
||||
The first half of the processTrade method handles the scenario where a market sell order results in hitting a
|
||||
limit buy order. It processes the fill, hedges the position, and logs various portfolio and holdings-related
|
||||
|
|
|
|||
|
|
@ -9,22 +9,22 @@ factual_summary: >
|
|||
runtime strategy state used for order placement and monitoring.
|
||||
|
||||
methods:
|
||||
registerExchange(exch: str, exchInst: Exchange) -> None:
|
||||
"registerExchange(exch: str, exchInst: Exchange) -> None":
|
||||
description: >
|
||||
Registers an exchange instance under a named key for use by the
|
||||
strategy during execution.
|
||||
|
||||
setMinQty(exch: str, qtyMap: dict) -> None:
|
||||
"setMinQty(exch: str, qtyMap: dict) -> None":
|
||||
description: >
|
||||
Updates minimum and maximum trade quantity constraints for a given
|
||||
exchange using a provided quantity mapping.
|
||||
|
||||
printFloating(ts: float) -> None:
|
||||
"printFloating(ts: float) -> None":
|
||||
description: >
|
||||
Emits a formatted runtime summary of strategy activity for logging
|
||||
and monitoring purposes.
|
||||
|
||||
bestEdgeBid(exch: str, sym: str) -> List[bool, float, bool, str, float]:
|
||||
"bestEdgeBid(exch: str, sym: str) -> List[bool, float, bool, str, float]":
|
||||
description: >
|
||||
This method is designed to determine whether and where a bid should be placed based on several factors
|
||||
related to market conditions and strategy rules. Its primary goal of is to calculate the optimal bid price for a
|
||||
|
|
@ -32,13 +32,13 @@ methods:
|
|||
price precision, and fee structures. Returns a collection of decision outputs including bid eligibility status,
|
||||
relative pricing divergence, long-position participation, selected hedge exchange, and the computed bid price.
|
||||
|
||||
bestEdgeOffer(exch: str, sym: str) -> list:
|
||||
"bestEdgeOffer(exch: str, sym: str) -> list":
|
||||
description: >
|
||||
The primary goal of bestEdgeOffer is to calculate the optimal offer (ask) price for a given symbol on a specific
|
||||
exchange, considering market positions, hedging strategies, trading limits, price precision, fee structures.
|
||||
Returns almost identical items as bestEdgeBid, but on the other side of the order book
|
||||
|
||||
updateTob(exch: str, sym: str, tob: dict, ts: float, force: bool) -> None:
|
||||
"updateTob(exch: str, sym: str, tob: dict, ts: float, force: bool) -> None":
|
||||
description: >
|
||||
The primary goal of the updateTob function is to update the top of the order book for quick access to perform
|
||||
market making operations (bidding and selling) using the highest bid and lowest ask prices.
|
||||
|
|
|
|||
Loading…
Reference in New Issue