Configuration

Environment Variables

Variable	Component	Description
EVENT_BUS_URL	All	NATS server URL
DATABASE_URL	API Server	PostgreSQL connection string
INSTANCE_NAME	Channels	Owning Agent name
MEMORY_ENABLED	Agent Runner	Whether persistent memory is active
MAX_TOOL_ITERATIONS	Agent Runner	Maximum LLM round-trips before the agent stops (default: 50). Each round may contain multiple parallel tool calls. Can also be set per-run via spec.env in AgentRun CR.
DETAILED_LOG_PATH	Agent Runner	Directory for untruncated JSONL log files. Empty = disabled. See Detailed Logging.
DETAILED_LOG_MAX_SIZE	Agent Runner	Max size per log file before rotation (default: 50m). Supports m (MB) and g (GB) suffixes.
CONTEXT_TOOL_RESULT_MAX_BYTES	Agent Runner	Per-result cap on tool output entering the conversation (default: 0 = off). See Context Cost.
CONTEXT_HISTORY_BUDGET_BYTES	Agent Runner	Total tool-result bytes that triggers elision of older results (default: 0 = off).
CONTEXT_HISTORY_BUDGET_LOW_BYTES	Agent Runner	Target to drain down to when elision fires (default: half the budget).
CONTEXT_KEEP_RECENT_RESULTS	Agent Runner	Newest tool results never eligible for elision (default: 3).
TELEGRAM_BOT_TOKEN	Telegram	Bot API token
SLACK_BOT_TOKEN	Slack	Bot OAuth token
SLACK_APP_TOKEN	Slack	App-level token for Socket Mode
DISCORD_BOT_TOKEN	Discord	Bot token
WHATSAPP_ACCESS_TOKEN	WhatsApp	Cloud API access token

LLM Providers

Sympozium supports any GenAI provider with an OpenAI-compatible API:

Provider	Base URL	API Key Variable
OpenAI	(default)	OPENAI_API_KEY
Anthropic	(default)	ANTHROPIC_API_KEY
Azure OpenAI	your endpoint	AZURE_OPENAI_API_KEY
Ollama	http://ollama:11434/v1	none
LM Studio	http://localhost:1234/v1	none
llama-server	http://localhost:8080/v1	none
Unsloth	http://localhost:8080/v1	none
Any OpenAI-compatible	custom URL	custom

See the Ollama guide, LM Studio guide, llama-server guide, or Unsloth guide for detailed local LLM setup.