GitHub Actions
Clean sync from GitHub - no large files in history
aca8ab4
# Azure OpenAI Configuration
# Get these from https://portal.azure.com → Your Azure OpenAI Resource
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_API_KEY=your-api-key-here
AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4o-mini
AZURE_OPENAI_API_VERSION=2024-05-01-preview
# ⚠️ CRITICAL: Embedding model deployment name
# This MUST match an existing deployment in your Azure OpenAI resource
# Common deployment names (check Azure Portal → Model deployments):
# - text-embedding-3-small (recommended, most cost-effective)
# - text-embedding-3-large (higher quality, more expensive)
# - text-embedding-ada-002 (legacy, widely compatible)
#
# HOW TO VERIFY:
# 1. Run: python scripts/validate_azure_embeddings.py
# 2. Or check Azure Portal → Your Resource → Model deployments
#
# ⚠️ If this deployment doesn't exist, you'll get a 404 error!
AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME=text-embedding-3-small
# Optional: Cost Pricing Overrides (per 1M tokens)
# These override the prices in config/pricing.json for all models
# Useful for testing or when using custom pricing
# PRICING_INPUT_PER_1M=0.08
# PRICING_OUTPUT_PER_1M=0.32
# PRICING_EMBEDDING_PER_1M=0.02
# MCP (Model Context Protocol) Configuration
# Set to 'true' to use MCP for arXiv access (default: FastMCP)
# Set to 'false' to use direct arXiv API
USE_MCP_ARXIV=false
# Set to 'true' to force legacy MCP instead of FastMCP
# Only applies when USE_MCP_ARXIV=true
USE_LEGACY_MCP=false
# Path where MCP server stores downloaded papers
# Should match the storage path configured in your MCP server
MCP_ARXIV_STORAGE_PATH=./data/mcp_papers/
# FastMCP Configuration
# Port for FastMCP server (auto-started when USE_MCP_ARXIV=true)
FASTMCP_SERVER_PORT=5555
# LangFuse Observability Configuration
# Enable/disable LangFuse tracing (default: true)
LANGFUSE_ENABLED=true
# LangFuse Cloud API Keys (get from https://cloud.langfuse.com)
LANGFUSE_PUBLIC_KEY=pk-lf
LANGFUSE_SECRET_KEY=sk-lf-
# LangFuse Host URL (default: https://cloud.langfuse.com)
# For self-hosted: LANGFUSE_HOST=http://localhost:3000
LANGFUSE_HOST=https://cloud.langfuse.com
# Optional: LangFuse Tracing Settings
# Trace all LLM calls automatically (default: true)
LANGFUSE_TRACE_ALL_LLM=true
# Trace RAG retrieval operations (default: true)
LANGFUSE_TRACE_RAG=true
# Flush observations after N items (default: 15)
LANGFUSE_FLUSH_AT=15
# Flush interval in seconds (default: 10)
LANGFUSE_FLUSH_INTERVAL=10