Skip to content

AI-Parrot Configuration Guide

This document describes the most important configuration values for AI-Parrot. All configuration values are loaded from .env files in the env/ folder using navconfig.

Table of Contents

LLM API Keys

Google GenAI

Configuration for Google's Generative AI (Gemini models).

  • GOOGLE_API_KEY: API key for Google GenAI services
  • VERTEX_PROJECT_ID: Google Cloud project ID for Vertex AI
  • VERTEX_REGION: Region for Vertex AI deployment

OpenAI

Configuration for OpenAI models (GPT-4, GPT-3.5, etc.).

  • OPENAI_API_KEY: API key for OpenAI services
  • OPENAI_ORGANIZATION: Optional organization ID for OpenAI

Anthropic Claude

Configuration for Claude models.

  • ANTHROPIC_API_KEY: API key for Claude/Anthropic services

Groq

Configuration for Groq's fast inference platform.

  • GROQ_API_KEY: API key for Groq services
  • DEFAULT_GROQ_MODEL: Default model to use (default: qwen/qwen3-32b)

HuggingFace

Configuration for HuggingFace models and embeddings.

  • HUGGINGFACEHUB_API_TOKEN: API token for HuggingFace Hub access

Database Configuration

PostgreSQL

Primary database for structured data and PgVector.

  • DBHOST: PostgreSQL host (default: localhost)
  • DBUSER: Database username
  • DBPWD: Database password
  • DBNAME: Database name (default: navigator)
  • DBPORT: PostgreSQL port (default: 5432)

The connection string is automatically constructed as: 

postgresql+asyncpg://{DBUSER}:{DBPWD}@{DBHOST}:{DBPORT}/{DBNAME}

ScyllaDB

NoSQL database for high-performance distributed storage.

  • SCYLLADB_DRIVER: Driver type (default: scylladb)
  • SCYLLADB_HOST: ScyllaDB host (default: localhost)
  • SCYLLADB_PORT: ScyllaDB port (default: 9042)
  • SCYLLADB_USERNAME: Username (default: navigator)
  • SCYLLADB_PASSWORD: Password (default: navigator)
  • SCYLLADB_KEYSPACE: Keyspace name (default: navigator)

BigQuery

Google BigQuery configuration for analytics.

  • BIGQUERY_CREDENTIALS: Path to BigQuery credentials JSON file
  • BIGQUERY_PROJECT_ID: GCP project ID (default: navigator)
  • BIGQUERY_DATASET: Default dataset name (default: navigator)

Cache & Redis

Redis is used extensively for caching, conversation history, job management, and knowledge base storage.

Core Redis Configuration

  • CACHE_HOST: Redis host (inherited from Navigator config)
  • CACHE_PORT: Redis port (inherited from Navigator config)

Redis History

Used for storing conversation history and agent memory.

  • REDIS_HISTORY_DB: Database number for conversation history (default: 3)
  • REDIS_HISTORY_URL: Automatically constructed as redis://{CACHE_HOST}:{CACHE_PORT}/{REDIS_HISTORY_DB}

Usage: The RedisKnowledgeBase class uses this for storing agent conversation memory and user preferences. JobManager and RQ (Redis Queue) use the base CACHE_HOST and CACHE_PORT for background task processing.

Vector Databases

Vector stores are used for RAG (Retrieval-Augmented Generation) and semantic search.

Milvus

High-performance vector database with advanced features.

  • MILVUS_HOST: Milvus host (default: localhost)
  • MILVUS_PROTOCOL: Protocol (default: http)
  • MILVUS_PORT: Milvus port (default: 19530)
  • MILVUS_URL: Complete Milvus URL (overrides host/port if set)
  • MILVUS_TOKEN: Authentication token
  • MILVUS_USER: Username for authentication
  • MILVUS_PASSWORD: Password for authentication
  • MILVUS_SECURE: Enable secure connection (default: false)

TLS/SSL Configuration: - MILVUS_SERVER_NAME: Server name for TLS - MILVUS_CA_CERT: Path to CA certificate - MILVUS_SERVER_CERT: Path to server certificate - MILVUS_SERVER_KEY: Path to server key - MILVUS_USE_TLSv2: Use TLS v1.2 (default: false)

Qdrant

Alternative vector database with good performance.

  • QDRANT_PROTOCOL: Protocol (default: http)
  • QDRANT_HOST: Qdrant host (default: localhost)
  • QDRANT_PORT: Qdrant port (default: 6333)
  • QDRANT_USE_HTTPS: Enable HTTPS (default: false)
  • QDRANT_URL: Complete Qdrant URL (overrides other settings)
  • QDRANT_CONN_TYPE: Connection type - server or cloud (default: server)

ChromaDB

Lightweight vector database for development.

  • CHROMADB_HOST: ChromaDB host (default: localhost)
  • CHROMADB_PORT: ChromaDB port (default: 8000)

Microsoft Integration

MS Teams Toolkit

Configuration for MS Teams bot and messaging integration. Used by the MSTeamsToolkit class for sending messages, creating chats, and managing Teams resources.

  • MS_TEAMS_TENANT_ID: Azure AD tenant ID (required)
  • MS_TEAMS_CLIENT_ID: Azure AD application client ID (required)
  • MS_TEAMS_CLIENT_SECRET: Azure AD application secret (required for app-only auth)
  • MS_TEAMS_USERNAME: Username for delegated authentication (required if as_user=True)
  • MS_TEAMS_PASSWORD: Password for delegated authentication (required if as_user=True)

Default Team/Channel: - MS_TEAMS_DEFAULT_TEAMS_ID: Default Teams team ID for notifications - MS_TEAMS_DEFAULT_CHANNEL_ID: Default channel ID for notifications

Note: Teams authentication supports both application-only permissions (using client secret) and delegated user permissions (using username/password).

Teams Notifications (Legacy)

Older notification system configuration.

  • TEAMS_NOTIFY_TENANT_ID: Tenant ID for notifications
  • TEAMS_NOTIFY_CLIENT_ID: Client ID for notifications
  • TEAMS_NOTIFY_CLIENT_SECRET: Client secret for notifications
  • TEAMS_NOTIFY_USERNAME: Username for notification service
  • TEAMS_NOTIFY_PASSWORD: Password for notification service

Office 365

General Office 365 integration.

  • O365_CLIENT_ID: Office 365 application client ID
  • O365_CLIENT_SECRET: Office 365 application secret
  • O365_TENANT_ID: Office 365 tenant ID

SharePoint

SharePoint-specific configuration.

  • SHAREPOINT_APP_ID: SharePoint app ID
  • SHAREPOINT_APP_SECRET: SharePoint app secret
  • SHAREPOINT_TENANT_ID: SharePoint tenant ID
  • SHAREPOINT_TENANT_NAME: SharePoint tenant name
  • SHAREPOINT_SITE_ID: Default site ID
  • SHAREPOINT_DEFAULT_HOST: Default SharePoint host

Cloud Services

Amazon AWS

AWS credentials for S3 and other services.

  • AWS_REGION: AWS region (default: us-east-1)
  • AWS_BUCKET: S3 bucket name (default: static-files)
  • AWS_KEY: AWS access key ID
  • AWS_SECRET: AWS secret access key

Google Cloud Services

Additional Google services beyond GenAI.

  • GOOGLE_SEARCH_API_KEY: API key for Google Custom Search
  • GOOGLE_SEARCH_ENGINE_ID: Custom Search Engine ID
  • GOOGLE_PLACES_API_KEY: API key for Google Places
  • GOOGLE_CREDENTIALS_FILE: Path to Google service account credentials (default: env/google/key.json)

Google Text-to-Speech: - GOOGLE_TTS_SERVICE: Path to TTS service credentials (default: env/google/tts-service.json)

Google Analytics: - GA_SERVICE_ACCOUNT_NAME: Service account filename (default: google.json) - GA_SERVICE_PATH: Path to service account files (default: env/google/)

LLM Defaults

Model Selection

  • DEFAULT_LLM_MODEL: Default model across the application (default: gemini-2.5-flash)
  • LLM_MODEL_NAME: Alias for model name (default: gemini-2.5-pro)
  • LLM_TEMPERATURE: Default temperature for generation (default: 0.1)

Model-Specific Defaults

  • DEFAULT_GROQ_MODEL: Default Groq model (default: qwen/qwen3-32b)

Embedding Configuration

Settings for text embeddings used in RAG and semantic search.

Device & Performance

  • EMBEDDING_DEVICE: Device for embedding inference - cpu or cuda (default: cpu)
  • CUDA_DEFAULT_DEVICE: CUDA device type (default: cpu)
  • CUDA_DEFAULT_DEVICE_NUMBER: CUDA device number (default: 0)
  • MAX_VRAM_AVAILABLE: Maximum VRAM in MB (default: 20000)
  • RAM_AVAILABLE: Available RAM in MB (default: 819200)
  • MAX_BATCH_SIZE: Maximum batch size for processing (default: 2048)

Models

  • EMBEDDING_DEFAULT_MODEL: Default embedding model from HuggingFace (default: sentence-transformers/all-MiniLM-L12-v2)
  • KB_DEFAULT_MODEL: Default model for knowledge base operations (default: sentence-transformers/paraphrase-MiniLM-L3-v2)

Storage

  • HUGGINGFACE_EMBEDDING_CACHE_DIR: Cache directory for downloaded models (default: {BASE_DIR}/model_cache/huggingface)

Directory Paths

Core application directories. Most paths are relative to BASE_DIR (project root).

Application Directories

  • PLUGINS_DIR: Plugins directory (default: {BASE_DIR}/plugins)
  • STATIC_DIR: Static files directory (default: {BASE_DIR}/static)
  • AGENTS_DIR: Agent definitions directory (default: {BASE_DIR}/agents)
  • AGENTS_BOTS_PROMPT_DIR: Agent prompts directory (default: {AGENTS_DIR}/prompts)
  • MCP_SERVER_DIR: MCP (Model Context Protocol) servers directory (default: {BASE_DIR}/mcp_servers)

Static Content

  • BASE_STATIC_URL: Base URL for static files (default: http://localhost:5000/static)

HTTP Client

Configuration for async HTTP operations.

  • HTTPCLIENT_MAX_SEMAPHORE: Maximum concurrent HTTP requests (default: 5)
  • HTTPCLIENT_MAX_WORKERS: Worker threads for HTTP operations (default: 1)

Additional Tools

Weather

  • OPENWEATHER_APPID: OpenWeatherMap API key
  • SERPAPI_API_KEY: SerpAPI key for web search functionality

Azure Bot

  • ENABLE_AZURE_BOT: Enable Azure Bot Framework integration (default: true)

Ethics

  • ETHICAL_PRINCIPLE: Ethical guidelines for AI responses (default: The model should only talk about ethical and legal things.)

Example Configuration File

Here's a minimal example .env file for the env/ folder:

# LLM API Keys
GOOGLE_API_KEY=your_google_api_key
OPENAI_API_KEY=your_openai_api_key
ANTHROPIC_API_KEY=your_anthropic_api_key
GROQ_API_KEY=your_groq_api_key

# Database
DBHOST=localhost
DBUSER=parrot_user
DBPWD=secure_password
DBNAME=parrot_db
DBPORT=5432

# Redis (from Navigator)
CACHE_HOST=localhost
CACHE_PORT=6379
REDIS_HISTORY_DB=3

# Vector Store (Milvus)
MILVUS_HOST=localhost
MILVUS_PORT=19530
MILVUS_USER=root
MILVUS_PASSWORD=milvus_password

# MS Teams Integration
MS_TEAMS_TENANT_ID=your-tenant-id
MS_TEAMS_CLIENT_ID=your-client-id
MS_TEAMS_CLIENT_SECRET=your-client-secret

# LLM Defaults
DEFAULT_LLM_MODEL=gemini-2.5-flash
LLM_TEMPERATURE=0.1

# Embeddings
EMBEDDING_DEVICE=cpu
EMBEDDING_DEFAULT_MODEL=sentence-transformers/all-MiniLM-L12-v2

Notes

  1. Configuration Priority: Values in environment files override default values in conf.py.

  2. Path Resolution: File paths can be absolute or relative. Relative paths are resolved from BASE_DIR.

  3. Security: Never commit .env files with sensitive credentials to version control.

  4. Dependencies: Some features require specific configuration:

  5. JobManager & RQ: Requires CACHE_HOST and CACHE_PORT for Redis
  6. MS Teams Toolkit: Requires tenant ID, client ID, and either client secret (app-only) or username/password (delegated)

  7. Vector RAG: Requires at least one vector database configured (PostgreSQL with PgVector, FAISS, Milvus, Qdrant, or ChromaDB)

  8. navconfig Integration: Parrot uses navconfig for configuration management, which loads from env/ folder by convention.