Configuration¶

DualLens Analytics uses a two-layer configuration system:

Secrets – loaded from .env via python-dotenv
Application parameters – loaded from conf/config.yaml via Hydra / OmegaConf

Environment Variables (`.env`)¶

Variable	Required	Description
`API_KEY`	Yes	OpenAI-compatible API key
`OPENAI_API_BASE`	Yes	Base URL for the LLM endpoint

API_KEY=gl-xxxxx
OPENAI_API_BASE=https://aibe.mygreatlearning.com/openai/v1

Warning

Never commit .env to version control. It is listed in .gitignore.

Hydra Config (`conf/config.yaml`)¶

The YAML file is divided into logical sections:

LLM¶

llm:
  model: gpt-4o-mini
  temperature: 0.3
  max_tokens: 1024
  top_p: 1.0
  frequency_penalty: 0.0

Embedding¶

embedding:
  model: text-embedding-ada-002

Chunking¶

Controls how PDF documents are split before embedding:

chunking:
  chunk_size: 1000
  chunk_overlap: 200
  encoding_name: cl100k_base

Retriever¶

retriever:
  k: 5
  search_type: similarity

Vector Store¶

vector_store:
  collection_name: ai_initiatives

Companies & Financial¶

companies:
  - Google
  - Microsoft
  - Amazon
  - Meta

stock:
  period: 1y

financial_metrics:
  - revenue
  - market_cap
  - total_debt
  - free_cash_flow
  - operating_income

Overriding via CLI (Hydra)¶

Any config value can be overridden from the command line:

poetry run python -m duallens_analytics.cli llm.temperature=0.7 retriever.k=10

See the Hydra documentation for advanced override syntax.