| Package |
Purpose |
docling |
PDF and HTML parsing with OCR support |
openai-agents |
OpenAI Agents SDK for multi-agent orchestration |
openai |
OpenAI-compatible API client (used with DeepSeek) |
onnxruntime-gpu |
GPU-accelerated OCR inference for Docling |
pydantic-settings |
Settings management from .env |
pymupdf |
Fallback PDF text extraction |
torch |
Required by Docling's ML models |
All tunable parameters live in rfpilot/config/settings.py and are overridable via .env:
| Setting |
Default |
Description |
DEEPSEEK_API_KEY |
— |
Your DeepSeek API key (required) |
DEEPSEEK_BASE_URL |
https://api.deepseek.com |
API base URL |
DEEPSEEK_MODEL |
deepseek-v4-flash |
Model name |
MAX_CONCURRENCY |
7 |
Max parallel LLM calls (raise to 10–15 for speed) |
MAX_CHUNK_SIZE |
1200 |
Max characters per prose chunk |
QA_GROUP_SIZE |
5 |
Q&A pairs grouped per extraction call |