Skip to content

configuration.md

Latest commit

 

History

History
120 lines (91 loc) · 3.78 KB

configuration.md

File metadata and controls

120 lines (91 loc) · 3.78 KB

Configuration Reference

All configuration is managed through environment variables (.env file) and/or Streamlit sidebar controls. Sidebar values override .env defaults.

LLM Configuration

Variable Default Description
LLM_PROVIDER ollama LLM provider: ollama, lmstudio, or openai
OLLAMA_MODEL llama3 Model name for Ollama
OLLAMA_BASE_URL http://localhost:11434 Ollama API endpoint
LMSTUDIO_BASE_URL http://localhost:1234/v1 LM Studio API endpoint (OpenAI-compatible)
OPENAI_API_KEY (empty) OpenAI API key (required for openai provider)

Provider Details

Ollama (LLM_PROVIDER=ollama)

  • Runs models locally via Ollama
  • Endpoint: http://localhost:11434/api/generate
  • Popular models: llama3, mistral, phi3, gemma

LM Studio (LLM_PROVIDER=lmstudio)

  • Runs any GGUF model locally via LM Studio
  • Exposes an OpenAI-compatible API at http://localhost:1234/v1
  • Set model name to match what LM Studio reports, or use local-model

OpenAI (LLM_PROVIDER=openai)

  • Uses OpenAI's cloud API
  • Requires OPENAI_API_KEY
  • Popular models: gpt-4o, gpt-4o-mini, gpt-3.5-turbo

Trading Configuration

Variable Default Description
MAX_PARALLEL_AGENTS 2 Number of analyst agents to run in parallel (1-4). 1 = sequential (safest for slow machines), 4 = full parallel (best for cloud APIs or powerful hardware)

These settings are also configurable via the Streamlit sidebar:

Setting Default Range
Risk Tolerance moderate conservative, moderate, aggressive
Max Debate Rounds 2 1-5
Analysis Period (days) 90 30-365

Data Source Configuration

All data sources are free. Toggle them on/off via .env:

Variable Default Description
ENABLE_GOOGLE_NEWS true Enable Google News article fetching
ENABLE_SEC_EDGAR true Enable SEC EDGAR filing retrieval
ENABLE_GOOGLE_TRENDS true Enable Google Trends interest data
GOOGLE_NEWS_PERIOD 7d Google News lookback period (1d, 7d, 30d)
GOOGLE_NEWS_MAX_RESULTS 10 Max Google News articles per query
SEC_EDGAR_MAX_FILINGS 5 Max SEC filings to retrieve per ticker
GOOGLE_TRENDS_TIMEFRAME today 3-m Google Trends timeframe (today 1-m, today 3-m, today 12-m)

Disabling a Data Source

If a data source is causing issues (network errors, rate limits, etc.), disable it:

ENABLE_GOOGLE_NEWS=false
ENABLE_SEC_EDGAR=false
ENABLE_GOOGLE_TRENDS=false

The system gracefully degrades — agents will work with whatever data is available.

Example .env File

# === LLM Provider ===
LLM_PROVIDER=lmstudio
OLLAMA_MODEL=llama3
OLLAMA_BASE_URL=http://localhost:11434
LMSTUDIO_BASE_URL=http://localhost:1234/v1
OPENAI_API_KEY=

# === Parallel Execution ===
MAX_PARALLEL_AGENTS=2

# === Data Sources ===
ENABLE_GOOGLE_NEWS=true
ENABLE_SEC_EDGAR=true
ENABLE_GOOGLE_TRENDS=true
GOOGLE_NEWS_PERIOD=7d
GOOGLE_NEWS_MAX_RESULTS=10
SEC_EDGAR_MAX_FILINGS=5
GOOGLE_TRENDS_TIMEFRAME=today 3-m

Recommended Settings

For Local LLMs (Ollama / LM Studio)

LLM_PROVIDER=lmstudio        # or ollama
MAX_PARALLEL_AGENTS=2         # balance speed vs stability
ENABLE_GOOGLE_TRENDS=false    # pytrends can be flaky

For Cloud APIs (OpenAI)

LLM_PROVIDER=openai
OPENAI_API_KEY=sk-...
MAX_PARALLEL_AGENTS=4         # cloud APIs handle concurrency well

For Slow Machines / Debugging

MAX_PARALLEL_AGENTS=1         # run analysts sequentially
ENABLE_GOOGLE_NEWS=false
ENABLE_SEC_EDGAR=false
ENABLE_GOOGLE_TRENDS=false