Agents API

Tools and utilities for integrating financial data with AI agents and LLMs.

---

fs.agents.summarize(df, max_tokens=None)

Generate an LLM-ready text summary of a financial DataFrame.

df = fs.load("AAPL", start="2024-01-01")
df = fs.features.add_all(df)

summary = fs.agents.summarize(df)

Example output:

AAPL | 2024-01-02 to 2025-03-20 (302 trading days)
Current price: $247.99
Range: $163.51 - $285.92
Last day: -0.39%
Last 5 days: -0.85%
Last 21 days: -4.83%
Volatility (annualized): 24.1%
Indicators: RSI(14): 27.4 (oversold) | Price is below 50-day SMA ($261.03) | MACD: bearish crossover
Volume: 87,981,315 (0.7x avg)

Parameters:

Parameter	Type	Default	Description
df	pl.DataFrame	required	DataFrame with finasys standard columns
max_tokens	int	None	Approximate token budget (1 token ~ 4 chars)

---

fs.agents.tools(symbols=None)

Generate tool definitions in OpenAI function-calling format.

tools = fs.agents.tools(symbols=["AAPL", "GOOGL", "MSFT"])

# Plug directly into OpenAI
# response = client.chat.completions.create(tools=tools, ...)

Returns a list of 4 tool definitions:

Tool Name	Description
lookup_price	Get price data for a symbol
get_technical_indicators	Compute RSI, MACD, Bollinger, etc.
compare_symbols	Compare performance across symbols
get_summary	Get comprehensive analysis summary

---

fs.agents.execute_tool(tool_name, arguments)

Execute a tool call and return the result as a string.

result = fs.agents.execute_tool("get_summary", {"symbol": "AAPL", "days": 60})
result = fs.agents.execute_tool("lookup_price", {"symbol": "GOOGL", "start": "2024-06-01"})

---

fs.agents.context(df, query, max_tokens=500, format="markdown")

Extract relevant context from a DataFrame based on a natural language query. Intelligently selects columns and rows based on query keywords.

# Momentum query -> selects RSI, MACD, SMA columns
ctx = fs.agents.context(df, "What is the recent momentum?")

# Volatility query -> selects ATR, rolling_std columns
ctx = fs.agents.context(df, "How volatile has it been?")

# Volume query -> selects volume column
ctx = fs.agents.context(df, "What is the trading volume?")

Parameters:

Parameter	Type	Default	Description
df	pl.DataFrame	required	DataFrame with features
query	str	required	Natural language query
max_tokens	int	500	Token budget for output
format	str	"markdown"	Output format: "markdown", "json", or "text"

---

fs.agents.schema(df)

Generate a human-readable schema description for LLM system prompts.

schema = fs.agents.schema(df)

Example output:

DataFrame with 302 rows and 37 columns.
Time range: 2024-01-02 to 2025-03-20.
Symbols: AAPL.
Columns:
  timestamp (Date)
  open (Float64)
  high (Float64)
  ...
  rsi_14 (Float64)
  macd_line (Float64)

---

LangChain Integration

Requires: pip install finasys[langchain]

from finasys.agents.langchain import get_tools

tools = get_tools(symbols=["AAPL", "GOOGL"])
# Returns List[BaseTool] ready for any LangChain agent