Configuration
LLM Configuration
Configure the LLM model in config/agent/llm.yaml:
# config/agent/llm.yaml
llm:
provider: openai
streaming:
enabled: true
thinking:
enabled: true
budget_tokens: 1000
models:
- name: model-1
api_key: ${LLM_MODELS_API_KEY}
base_url: ${LLM_MODELS_BASE_URL}
model: ${LLM_MODELS_MODEL}
temperature: 0.7
timeout: 120
vision: true
current_model: model-1The built-in model registry supports 40+ models, auto-detecting context window, tool support, and max output.
Agent Configuration
Configure agent behavior in config/agent/settings.yaml:
# config/agent/agent.yaml
agent:
section: agent
max_iterations: 10
enforce_iteration_limit: false
request_timeout: 300
max_sessions: 100
default_session_id: default
flash_mode: false
shell_cwd: workspaceServer Config
# config/agent/server.yaml
server:
host: "0.0.0.0"
port: 18000
cors: trueMemory Config
# config/agent/memory.yaml
memory:
section: memory
memory_dir: memory
short_term:
max_history: 50
session_compact:
enabled: true
token_threshold: 100000
tool_call_threshold: 50
keep_recent_messages: 20
max_notes_length: 2000
notes_dir: memory/notes
long_term:
db_file: memory/memory_index.db
hybrid_enabled: true
embedding:
enabled: false
model: ${EMBEDDING_MODEL}
api_key: ${EMBEDDING_API_KEY}
base_url: ${EMBEDDING_BASE_URL}
allow_sensitive_store: false
default_schema: generalLogging Config
# config/agent/logging.yaml
logging:
level: "INFO"
format: "%(asctime)s [%(levelname)s] %(name)s: %(message)s"
console: true
max_size: 5000
backup_count: 5Channel Configuration
Configure external channels in config/agent/channels.yaml:
# config/agent/channels.yaml
channels:
feishu:
app_id: ''
app_secret: ''
auto_start: true
enabled: true
whitelist_users: []
qq:
app_id: ''
app_secret: ''
auto_start: true
enabled: true
intents: 1107296256
telegram:
auto_start: true
bot_token: ''
enabled: true
whitelist_user_ids: []
whitelist_usernames: []
weixin:
auto_start: true
bot_agent: Cellium Agent
enabled: true
state_dir: data/weixin