feat: change predefined signal creation to REPL-based dynamic creation (#16)

Author: kimtth

.env.example

@@ -13,9 +13,9 @@ INVESTMENT_INITIAL_CAPITAL=10000
 # Set true only to launch the Agent Framework Dev UI instead of running the workflow.
 LAUNCH_DEV_UI=false
 
-# Azure OpenAI (API key auth — alternative to Entra ID above)
-# Uncomment and set these if using api_key= instead of credential=
+# Optional direct Azure OpenAI connection for the Semantic Kernel workflow.
+# Without these values, it reuses AZURE_AI_PROJECT_ENDPOINT and the Foundry model above.
 # AZURE_OPENAI_ENDPOINT=https://<resource>.openai.azure.com/
-# AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME=gpt-4o
+# AZURE_OPENAI_CHAT_DEPLOYMENT_NAME=gpt-5.6-terra
 # AZURE_OPENAI_API_VERSION=2025-01-01-preview
 # AZURE_OPENAI_API_KEY=

README.md

@@ -8,7 +8,7 @@
 
 # 💸 Investment Agent Patterns
 
-This project shows how to use Microsoft Agent Framework for stock-market research. It downloads past prices, creates signals, tests simple strategies, and saves the results.
+This project shows how to use Microsoft Agent Framework for stock-market research. Agents download past prices, write and execute a technical-analysis signal script, test the resulting strategy, and save the results.
 
 > [!NOTE]  
 > Recommended first: Microsoft Agent Framework. It combines ideas from AutoGen and Semantic Kernel.  
@@ -43,50 +43,56 @@ On PowerShell, use `Copy-Item .env.example .env` to copy the settings file. Then
 |---|---|
 | `AZURE_AI_PROJECT_ENDPOINT` | Address of the Azure AI Foundry project used by the main workflow. |
 | `AZURE_AI_MODEL_DEPLOYMENT_NAME` | Name of the chat model deployed in that project. |
+| `AZURE_OPENAI_ENDPOINT` | Optional direct Azure OpenAI endpoint for the Semantic Kernel workflow; otherwise it uses the Foundry project endpoint. |
+| `AZURE_OPENAI_CHAT_DEPLOYMENT_NAME` | Optional direct Azure OpenAI deployment for the Semantic Kernel workflow. |
 | `INVESTMENT_TICKER` | Stock symbol to study. The default is `MSFT`. |
 | `INVESTMENT_START_DATE`, `INVESTMENT_END_DATE` | First and last dates for the past-price data. |
 | `INVESTMENT_INITIAL_CAPITAL` | Pretend starting amount for the backtest. |
 
 ## Optional: Semantic Kernel variant
 
-Use this command only when comparing the plugin-based variant. It creates the same type of charts, metrics, spreadsheets, and CSV files as the primary Agent Framework workflow, in its own output folder:
+Use this command only when comparing the plugin-based agent variant. It reuses the configured Foundry project and model by default, then creates charts, metrics, spreadsheets, CSVs, and the generated signal script in its own output folder:
 
 ```bash
 uv run python -m semantic_kernel.main
 ```
 
 ## Sample research output
 
-These files were created by the main Agent Framework workflow for `MSFT`, using the default date range and a pretend $10,000 starting balance. They are examples only. Results can change when the workflow is run again because market data changes.
+The following Agent Framework sample uses `MSFT` from 2020-01-01 through 2026-07-01 with a simulated $10,000 starting balance. It is historical research only; a later run can generate a different strategy and result.
 
-### Input query
+### Input
 
 ```text
-Analyze MSFT from 2020-01-01 to 2026-07-01; generate MACD/RSI signals, backtest $10000, and report CAGR, total return, final value, drawdown, and Sharpe ratio.
+Analyze MSFT from 2020-01-01 to 2026-07-01. Develop one transparent technical-analysis signal strategy as Python code, execute it, backtest $10000, and report CAGR, total return, final value, drawdown, and Sharpe ratio.
 ```
 
-### Performance graph
+### Generated signal
 
-The top line shows how the pretend portfolio value changed over time. The shaded lower chart shows drawdown: how far the portfolio fell from its highest value at each point.
+The sample agent generated a long-only trend-and-momentum signal using MSFT closing prices:
 
-<img src="output/agent_framework/trix_ultimateoscillator_reversal/stock_plot.png" alt="TRIX and Ultimate Oscillator cumulative returns and drawdown" width="500">
+- **Buy:** the 50-day simple moving average is above the 200-day simple moving average and RSI(14) is above 50, after that combined condition was previously false.
+- **Sell:** the combined condition becomes false: the 50-day average is at or below the 200-day average, or RSI(14) is at or below 50.
+- **Hold:** no change in the combined condition.
+- **Execution convention:** a close-based condition affects the next session's return, reducing same-close look-ahead bias.
 
-### Backtest metrics
+### Performance plot
+
+The top panel shows cumulative strategy returns. The lower panel shows portfolio drawdown from its previous peak.
 
-| Strategy | Cumulative return | CAGR | Maximum drawdown | Sharpe ratio | Final value |
-|---|---:|---:|---:|---:|---:|
-| [MACD and RSI momentum](output/agent_framework/macd_rsi_momentum/backtest_metrics.txt) | -2.82% | -0.91% | -26.51% | 0.02 | $9,717.67 |
-| [Moving-average trend](output/agent_framework/movingaverage_trend/backtest_metrics.txt) | -2.00% | -0.64% | -28.29% | 0.02 | $9,799.54 |
-| [TRIX and Ultimate Oscillator reversal](output/agent_framework/trix_ultimateoscillator_reversal/backtest_metrics.txt) | 9.60% | 2.96% | -25.73% | 0.29 | $10,959.96 |
+<img src="output/agent_framework/stock_plot.png" alt="Sample cumulative returns and drawdown for the Agent Framework research run" width="600">
 
-- **Cumulative return** is the total percentage gained or lost over the whole test.
-- **CAGR** is the average yearly growth rate.
-- **Maximum drawdown** is the largest drop from a previous high point.
-- **Sharpe ratio** is one simple measure of return compared with day-to-day ups and downs; it is not a guarantee of quality.
+### Backtest metrics
 
-### Summary example
+| Metric | Sample value |
+|---|---:|
+| [Cumulative return](output/agent_framework/backtest_metrics.txt) | 27.96% |
+| CAGR | 3.88% |
+| Maximum drawdown | -21.03% |
+| Sharpe ratio | 0.36 |
+| Final value | $12,796.05 |
 
-> In this example run, the TRIX and Ultimate Oscillator strategy did better than the other two examples. A pretend $10,000 grew to $10,959.96, a 9.60% total gain. However, the portfolio also fell as much as 25.73% from an earlier high. One past result is not proof that a strategy will work in the future. Test it over other time periods and include trading fees, price changes during trades, and data checks before trusting the result.
+See the full [backtest workbook](output/agent_framework/backtest_results.xlsx), [generated signal script](output/agent_framework/generated_signal_strategy.py), and [validated signals](output/agent_framework/stock_signals.csv).
 
 ## Repository layout
 
@@ -96,7 +102,7 @@ The top line shows how the pretend portfolio value changed over time. The shaded
 | [semantic_kernel](semantic_kernel) | The Semantic Kernel version of the research workflow. |
 | [autogen](autogen) | The AutoGen reference version, with its own Poetry environment. |
 | [agent_framework_patterns](agent_framework_patterns) | Thirty small Agent Framework examples for investment research. |
-| [tests](tests) | Offline tests for the Agent Framework patterns. |
+| [tests](tests) | Offline tests for Agent Framework patterns and the REPL contracts. |
 | [output](output) | Saved charts, metrics, and example pattern responses. |
 | [docs](docs) | Guides for each implementation and framework comparison. |
 
@@ -106,12 +112,13 @@ The Agent Framework pattern tests run without Azure credentials or live market-d
 
 ```bash
 uv run ruff check agent_framework semantic_kernel agent_framework_patterns tests
-uv run pytest tests/agent_framework_patterns -q
+uv run pytest tests -q
 ```
 
 ## Safety and limitations
 
 - These results are for learning and research, not financial advice or a real trading system.
+- Both workflows execute model-authored Python to create signals. Their validation is not a security sandbox; run them only in an isolated development environment without credentials or production data.
 - Good results from the past do not mean the same strategy will work in the future.
 - The examples use public market data and simple rules. They leave out trading fees, price changes that happen while a trade is being made, taxes, careful handling of stock splits and dividends, and checks for an individual investor's needs.
 - Review AI responses, connected tools, and data licences before using this project outside a learning or research setting.

agent_framework/main.py

@@ -15,7 +15,11 @@ def task() -> str:
     start = os.getenv("INVESTMENT_START_DATE", "2020-01-01")
     end = os.getenv("INVESTMENT_END_DATE", "2026-07-01")
     capital = os.getenv("INVESTMENT_INITIAL_CAPITAL", "10000")
-    return f"Analyze {ticker} from {start} to {end}; generate MACD/RSI signals, backtest ${capital}, and report CAGR, total return, final value, drawdown, and Sharpe ratio."
+    return (
+        f"Analyze {ticker} from {start} to {end}. Develop one transparent technical-analysis "
+        f"signal strategy as Python code, execute it, backtest ${capital}, and report CAGR, "
+        "total return, final value, drawdown, and Sharpe ratio."
+    )
 
 
 async def build() -> tuple[QuantInvestWorkflow, object]:

agent_framework/models.py

@@ -1,31 +1,6 @@
-"""Data contracts and strategy catalog for the Agent Framework implementation."""
+"""Data contracts for the Agent Framework research tools."""
 
-from pathlib import Path
-from typing import Literal
-
-from pydantic import BaseModel, Field, field_validator
-
-
-class StrategyIdea(BaseModel):
-    name: str
-    indicator: Literal["macd_rsi", "moving_average", "trix_uo"]
-    description: str
-    rationale: str
-    investing_conditions: str
-    expected_outcome: str
-
-
-class WorkflowRequest(BaseModel):
-    ticker: str = Field(min_length=1, max_length=12)
-    start_date: str
-    end_date: str
-    initial_capital: float = Field(default=10_000, gt=0)
-    strategies: list[StrategyIdea] | None = None
-
-    @field_validator("ticker")
-    @classmethod
-    def normalize_ticker(cls, value: str) -> str:
-        return value.upper().strip()
+from pydantic import BaseModel
 
 
 class BacktestMetrics(BaseModel):
@@ -34,48 +9,3 @@ class BacktestMetrics(BaseModel):
     maximum_drawdown: float
     sharpe_ratio: float
     final_value: float
-
-
-class StrategyRun(BaseModel):
-    strategy: StrategyIdea
-    metrics: BacktestMetrics
-    output_directory: Path
-    stock_data_file: Path
-    signals_file: Path
-    results_file: Path
-    metrics_file: Path
-    plot_file: Path
-
-
-class WorkflowResult(BaseModel):
-    ticker: str
-    strategy_ideas_file: Path
-    runs: list[StrategyRun]
-
-
-IDEAS = (
-    StrategyIdea(
-        name="MACD RSI momentum",
-        indicator="macd_rsi",
-        description="MACD crossover confirmation with RSI filtering.",
-        rationale="Pairs trend momentum with an overbought guardrail.",
-        investing_conditions="Enter on a bullish crossover below RSI 70; exit on a bearish crossover or RSI above 75.",
-        expected_outcome="Research hypothesis for sustained momentum.",
-    ),
-    StrategyIdea(
-        name="Moving-average trend",
-        indicator="moving_average",
-        description="20-day and 60-day moving-average crossover.",
-        rationale="Transparent trend-following baseline.",
-        investing_conditions="Enter and exit on the corresponding crossover.",
-        expected_outcome="Research benchmark for persistent trends.",
-    ),
-    StrategyIdea(
-        name="TRIX ultimate-oscillator reversal",
-        indicator="trix_uo",
-        description="TRIX momentum with the Ultimate Oscillator.",
-        rationale="Tests reversal timing across lookback periods.",
-        investing_conditions="Enter on positive TRIX with UO below 50; exit on negative TRIX or UO above 70.",
-        expected_outcome="Research hypothesis for momentum transitions.",
-    ),
-)