Signal Forge

Signal Forge is a lightweight web app for researching stocks and crypto with:

• free market-data integrations for stocks and digital assets
• a multi-analyst debate panel that argues the bull, bear, and quant case
• an optional AI counsel mode with 2 to 5 web-enabled agents, rotating models, and a visible discussion log
• ranked ideas for a quarter, six months, and one year
• direct analysis of a ticker, company name, coin, or token symbol
• an explicit "why this is recommended" explanation for every analysis

Tech stack

• Python 3.13+
• standard-library HTTP server
• static HTML, CSS, and vanilla JavaScript frontend
• free market data from Yahoo Finance and CoinGecko
• optional OpenAI-compatible chat endpoint for LLM-backed debate and AI counsel

Project layout

• stock_analyser/app.py: entrypoint
• stock_analyser/server.py: web server and JSON API routes
• stock_analyser/providers.py: free data-provider integrations
• stock_analyser/analysis.py: scoring, horizon ranking, and recommendation explanations
• stock_analyser/debate.py: rules-based and optional LLM-backed debate orchestration
• static/index.html: web interface
• static/styles.css: UI styling
• static/app.js: client-side interactions

Run locally

1. Create and activate a virtual environment:

python3 -m venv .venv
source .venv/bin/activate

2. Install project requirements:

python -m pip install --upgrade pip
python -m pip install -r requirements.txt

3. Copy the sample environment file:

cp .env.example .env

4. Optional: add OPENAI_API_KEY and OPENAI_MODEL to .env if you want real LLM-backed analyst debate.

Optional AI counsel configuration:

OPENAI_API_KEY=...
OPENAI_MODEL=gpt-4.1-mini
OPENAI_COUNSEL_MODELS=gpt-5,gpt-5-mini,gpt-4.1-mini
OPENAI_COUNSEL_MAX_MEMBERS=5
OPENAI_COUNSEL_TIMEOUT_SECONDS=90

The standard debate uses OPENAI_MODEL. The optional AI counsel can reuse that model or randomly draw from the comma-separated OPENAI_COUNSEL_MODELS pool. OPENAI_COUNSEL_TIMEOUT_SECONDS gives the web-enabled counsel extra time for slower search-heavy runs. The UI keeps AI counsel off by default and lets you pick 2 to 5 agents when you enable it.

5. Start the app:

python -m stock_analyser.app

6. Open:

http://127.0.0.1:8000

API endpoints

• GET /api/health
• GET /api/lookup?query=MSFT&assetType=auto
• POST /api/analyze
• GET /api/suggestions?assetType=stock
• GET /api/suggestions?assetType=crypto

Example analyze payload:

{
  "query": "MSFT",
  "assetType": "stock",
  "horizon": "quarter",
  "aiCounselEnabled": true,
  "counselMembers": 4
}

Notes

• Stock lookups and charts use free Yahoo Finance endpoints.
• Crypto lookups and analytics use CoinGecko.
• The stock fundamentals feed is intentionally lightweight because the app prefers free, low-friction sources over paid APIs.
• The AI counsel path calls the OpenAI Responses API with web search enabled, so it is slower and more expensive than the standard debate.
• requirements.txt is intentionally minimal right now because the app runs on the standard library.
• This app is for research support and education, not personal financial advice.

Tests

Run the small offline test suite with:

python -m unittest discover -s tests