Skip to content

environment-variables.md

Latest commit

 

History

History
215 lines (162 loc) · 9.29 KB

environment-variables.md

File metadata and controls

215 lines (162 loc) · 9.29 KB
Raw

Environment Variables Reference

AutoBot supports comprehensive configuration through environment variables with the AUTOBOT_ prefix. All variables have fallback defaults defined in config/config.yaml.

Backend Configuration

Variable Default Description
AUTOBOT_BACKEND_HOST 127.0.0.1 (or 0.0.0.0 on WSL2) Backend server host — automatically configured by Ansible from backend_host variable (MVA-2418)
AUTOBOT_BACKEND_PORT 8001 Backend server port
AUTOBOT_BACKEND_API_ENDPOINT http://localhost:8001 Full API endpoint URL
AUTOBOT_BACKEND_TIMEOUT 60 Request timeout in seconds
AUTOBOT_BACKEND_MAX_RETRIES 3 Maximum retry attempts
AUTOBOT_BACKEND_STREAMING false Enable streaming responses

LLM Configuration

Variable Default Description
AUTOBOT_OLLAMA_HOST http://localhost:11434 Ollama server endpoint
AUTOBOT_OLLAMA_MODEL qwen3.5:9b Default Ollama model (deprecated, use AUTOBOT_DEFAULT_LLM_MODEL)
AUTOBOT_OLLAMA_ENDPOINT http://localhost:11434/api/generate Ollama API endpoint
AUTOBOT_OLLAMA_SELECTED_MODEL qwen3.5:9b Currently selected Ollama model
AUTOBOT_ORCHESTRATOR_LLM llama3.2:1b LLM model for orchestrator (Routing tier)
AUTOBOT_DEFAULT_LLM ollama_qwen3.5:9b Default LLM identifier
AUTOBOT_TASK_LLM ollama_qwen3.5:9b Task-specific LLM identifier
AUTOBOT_LLM_PROVIDER_TYPE local LLM provider type (local/cloud)

6-Tier Model Configuration

Variable Default Tier Purpose
AUTOBOT_ROUTING_MODEL llama3.2:1b Routing Orchestrator, request routing
AUTOBOT_CLASSIFICATION_MODEL gemma2:2b Classification Intent detection, category assignment
AUTOBOT_LIGHT_PROCESSING_MODEL phi3:mini Light Processing Extraction, formatting
AUTOBOT_INSTRUCTION_MODEL mistral:7b-instruct Instruction RAG, step execution
AUTOBOT_SYSTEM_MODEL dolphin-llama3:8b System Commands, security
AUTOBOT_DEFAULT_LLM_MODEL qwen3.5:9b Quality Chat, research, code

LLM Fallback Chain Configuration (GH#8998)

When the primary model hits a rate limit (429) or quota cap, AutoBot automatically routes to backup models. See LLM Fallback Configuration for the full guide.

Variable Default Description
AUTOBOT_FALLBACK_CHAIN_<MODEL> (none) Comma-separated fallback models for <MODEL>. Use provider:model format to specify provider (e.g. anthropic:claude-sonnet-4,openai:gpt-4o-mini). Model suffix uses uppercase with underscores (hyphens are auto-converted).

Examples:

# Claude Opus → Sonnet → GPT-4o-mini if Anthropic quota exhausted
AUTOBOT_FALLBACK_CHAIN_CLAUDE_OPUS_4=anthropic:claude-sonnet-4,openai:gpt-4o-mini

# GPT-4o → mini → local Ollama as last resort
AUTOBOT_FALLBACK_CHAIN_GPT_4O=openai:gpt-4o-mini,ollama:llama3

LLM Rate Limiter Configuration (GH#8170)

Proactive per-provider token-bucket rate limiter (Redis-backed, shared across all uvicorn workers).

Variable Default Description
AUTOBOT_LLM_RL_<PROVIDER>_RPM (see below) Requests per minute for <PROVIDER> (e.g. AUTOBOT_LLM_RL_ANTHROPIC_RPM=200)
AUTOBOT_LLM_RL_<PROVIDER>_BURST Same as RPM Burst token capacity for <PROVIDER>

Provider defaults: openai=500, anthropic=60, groq=30, ollama=600, vllm=600, huggingface=30, openrouter=60, custom_openai=120. All others default to 60 RPM.

Redis Configuration

Variable Default Description
AUTOBOT_REDIS_HOST localhost Redis server host
AUTOBOT_REDIS_PORT 6379 Redis server port
AUTOBOT_REDIS_ENABLED false Enable Redis for memory storage

Chat Configuration

Variable Default Description
AUTOBOT_CHAT_MAX_MESSAGES 100 Maximum messages per chat
AUTOBOT_CHAT_WELCOME_MESSAGE Hello! How can I assist you today? Default welcome message
AUTOBOT_CHAT_AUTO_SCROLL true Auto-scroll chat interface
AUTOBOT_CHAT_RETENTION_DAYS 30 Chat message retention period

Knowledge Base Configuration

Variable Default Description
AUTOBOT_KB_ENABLED true Enable knowledge base
AUTOBOT_KB_UPDATE_FREQUENCY 7 Update frequency in days
AUTOBOT_KB_DB_PATH `` Knowledge base database path

Logging Configuration

Variable Default Description
AUTOBOT_LOG_LEVEL info Logging level (debug/info/warning/error)
AUTOBOT_LOG_TO_FILE false Enable file logging
AUTOBOT_LOG_FILE_PATH `` Log file path

Developer Configuration

Variable Default Description
AUTOBOT_DEVELOPER_MODE false Enable developer mode
AUTOBOT_DEBUG_LOGGING false Enable debug logging
AUTOBOT_ENHANCED_ERRORS true Show enhanced error messages

UI Configuration

Variable Default Description
AUTOBOT_UI_THEME light UI theme (light/dark)
AUTOBOT_UI_FONT_SIZE medium Font size (small/medium/large)
AUTOBOT_UI_LANGUAGE en Interface language
AUTOBOT_UI_ANIMATIONS true Enable UI animations

Message Display Configuration

Variable Default Description
AUTOBOT_SHOW_THOUGHTS true Show thought messages
AUTOBOT_SHOW_JSON false Show JSON output
AUTOBOT_SHOW_DEBUG false Show debug messages
AUTOBOT_SHOW_PLANNING true Show planning messages
AUTOBOT_SHOW_UTILITY false Show utility messages

Security Configuration

Variable Default Description
AUTOBOT_ENABLE_ENCRYPTION false Enable data encryption
AUTOBOT_FIELD_ENCRYPTION_KEY (required) 32-byte AES-256-GCM encryption key (base64url-encoded) for encrypting sensitive fields like Telegram bot tokens. Required for Telegram bot integration. Generate with: python -c "import secrets,base64; print(base64.urlsafe_b64encode(secrets.token_bytes(32)).decode())" (GH#9606)
AUTOBOT_SESSION_TIMEOUT 30 Session timeout in minutes
AUTOBOT_EMBED_ALLOWED_ORIGINS * Comma-separated list of allowed origins for embed widget. Use * (default) to allow all origins, or specify domains like https://example.com,https://app.acme.io to restrict access (GH#9117)
AUTOBOT_EMBED_TRUSTED_PROXIES `` Comma-separated list of trusted proxy IPs for X-Forwarded-For validation. Only connections from these IPs will have their X-Forwarded-For header trusted for rate limiting. Leave empty (default) to rate-limit by direct connection IP only. Example: 10.0.1.100,10.0.1.101 (GH#9117)

Voice Interface Configuration

Variable Default Description
AUTOBOT_VOICE_ENABLED false Enable voice interface
AUTOBOT_VOICE_RATE 1 Speech rate multiplier
AUTOBOT_VOICE default Voice selection

Legacy Configuration

Variable Default Description
AUTOBOT_USE_LANGCHAIN false Use LangChain orchestrator
AUTOBOT_USE_PHI2 false Use Phi-2 model (deprecated)

Frontend Configuration

The frontend uses Vite environment variables with the VITE_ prefix:

Variable Default Description
VITE_API_BASE_URL http://localhost:8001 Backend API base URL

Usage Examples

Setting Default LLM Model

# Quality tier (default for complex tasks)
export AUTOBOT_DEFAULT_LLM_MODEL="qwen3.5:9b"

# Override specific tiers
export AUTOBOT_ROUTING_MODEL="llama3.2:1b"
export AUTOBOT_INSTRUCTION_MODEL="mistral:7b-instruct"
export AUTOBOT_SYSTEM_MODEL="dolphin-llama3:8b"

Using Different Backend Port

export AUTOBOT_BACKEND_PORT=8002
export AUTOBOT_BACKEND_API_ENDPOINT="http://localhost:8002"
export VITE_API_BASE_URL="http://localhost:8002"

Enable Redis

export AUTOBOT_REDIS_ENABLED=true
export AUTOBOT_REDIS_HOST="redis.example.com"
export AUTOBOT_REDIS_PORT=6380

Developer Mode

export AUTOBOT_DEVELOPER_MODE=true
export AUTOBOT_DEBUG_LOGGING=true
export AUTOBOT_LOG_LEVEL=debug

Telegram Bot Encryption (GH#9606)

# Generate a secure encryption key
export AUTOBOT_FIELD_ENCRYPTION_KEY=$(python -c "import secrets,base64; print(base64.urlsafe_b64encode(secrets.token_bytes(32)).decode())")

# Configure Telegram bot token (will be encrypted automatically)
# Note: Bot tokens are encrypted at rest in Redis using AES-256-GCM

Configuration Priority

Environment variables take precedence over config file values:

  1. Environment Variables (highest priority)
  2. config/config.yaml (fallback defaults)
  3. Hardcoded defaults (last resort, now eliminated)

Validation

The system validates environment variables on startup and logs applied overrides. Invalid values will fall back to config file defaults with warnings logged.