Configuration

LLM Interceptor uses a single YAML configuration file. All settings are optional — sensible defaults are provided for every value.

Minimal Config

listen: "127.0.0.1:8080"
upstream: "https://api.anthropic.com"

Server

Field	Default	Description
listen	127.0.0.1:8080	Bind address and port
upstream	https://api.anthropic.com	Upstream LLM base URL
metric_prefix	llm_proxy.	Metrics prefix for OTel and Prometheus
shutdown_timeout_sec	10	Graceful shutdown timeout

Admin

admin:
  username: "admin"
  password: ""          # bcrypt hashed; auto-generated if empty on first run
  jwt_secret: ""        # auto-generated if empty on first run

Log

log:
  format: json           # json or text
  level: info            # debug, info, warn, error
  output: stdout         # stdout or file path
  request_body: false    # log request bodies (can be large)
  response_body: false   # log response bodies (can be large)

Storage

storage:
  type: sqlite
  sqlite:
    path: "~/.llm-interceptor/data.db"
  # type: postgres
  # postgres:
  #   connection_string: "postgres://user:pass@localhost:5432/llmproxy?sslmode=disable"
  compression:
    enabled: true
    algorithm: gzip
    min_size: 1024

State Store

state_store:
  type: memory
  # type: redis
  # redis:
  #   url: "redis://:password@localhost:6379/0"

Plugins

plugins:
  otel-exporter:
    enabled: false
    endpoint: "localhost:4318"
  cost-tracker:
    enabled: true
  budget:
    max_cost_per_session: 0.50
    max_cost_per_day: 0
  rate-limit:
    requests_per_minute: 60
    tokens_per_minute: 0
  tool-policy:
    blocked_tools: []
    allowed_tools: []

Router

router:
  enabled: false
  providers:
    - name: anthropic
      base_url: "https://api.anthropic.com"
      model_glob: "claude-*"
      api_key: "${ANTHROPIC_API_KEY}"
    - name: openai
      base_url: "https://api.openai.com"
      model_glob: "gpt-*"
      api_key: "${OPENAI_API_KEY}"

Alerting

# alerting:
#   slack_webhook_url: "https://hooks.slack.com/services/..."
#   rules:
#     - name: "high-error-rate"
#       metric: "error_rate"
#       threshold: 0.1
#       duration: "5m"
#       channels: ["slack"]
#       severity: "critical"

Runtime Overlay

Config values stored in the runtime_config database table override YAML at startup. Use the Admin API (PUT /api/admin/config/{key}) to update settings without restarting.