Skip to content

Configuration Reference Open-Core

Environment variables, workspace settings, and proxy configurations for the Intutic platform.


1. Environment Variables

Local Client & Proxy (Open Source)

These variables configure the local client utilities, sync daemon, and proxy gateway running on developer workstations.

VariableRequired?Default / PlaceholderDescription
VALKEY_URLredis://127.0.0.1:6379Local Valkey cache connection URL
INTUTIC_API_KEYVirtual API key (vk_*) used to authenticate and sync workspace rules
CONTROL_PLANE_URLhttps://your-control-plane.exampleControl plane base URL for remote SOP and policy syncing
CONFIG_CAPTURE_INTERVAL5Sync loop cycles between local configuration snapshots
MCP_DAEMON_SOCKET~/.intutic/mcp-proxy.sockDaemon Unix IPC socket path
INTUTIC_WASM_DIR~/.intutic/wasmOverrides the local WASM rule directory; takes precedence over the intutic_settings.wasm_local_dir config file value

Enterprise Control Plane (SaaS / Private VPC)

These variables are used exclusively in the backend control plane deployment to orchestrate registries, databases, and multi-channel notifications.

VariableRequired?Default / PlaceholderDescription
DATABASE_URLpostgresql://...Postgres database connection string (Drizzle ledger)
JWT_SECRETchangemeSecret key for signing dashboard JWT tokens
ENCRYPTION_KEYchangeme32-byte hex key for encrypting credentials and tokens
PORT3001HTTP port for the control plane API service
LITELLM_ADMIN_BASE_URLhttp://litellm:4000LiteLLM helper admin URL
LITELLM_MASTER_KEYLiteLLM helper API token

2. Local Sandbox Development Stack

The open-source workstation components run against a lightweight, local Valkey cache service.

bash
# Start the local development Valkey service
docker compose up -d

# Valkey port mapping:
# Valkey: 6379

Test stack

To execute the unit and integration tests under isolated conditions, spin up the test compose file:

bash
# Start isolated test services
docker compose -f docker-compose.test.yml up -d

# Valkey test port mapping:
# Valkey-test: 6380

3. Workspace Settings

Workspaces are the top-level organizational unit. Each workspace has:

SettingTypeDescription
Workspace IDwk_*Auto-generated unique identifier
NamestringHuman-readable workspace name
Planenumfree_trial, pro, team, enterprise
Budget tiersobjectPer-role budget limits

Budget Tiers

TierIntended for
JUNIORJunior developers — lowest budget ceiling
SENIORSenior developers
STAFFStaff engineers
PRINCIPALPrincipal engineers — highest budget ceiling

Model Routing Tiers

TierUsage
frontierLatest, most capable models (e.g., Claude 4, GPT-4.5)
economyCost-effective models for routine tasks
localLocally-hosted models for maximum privacy

Execution Modes

ModeDescription
STANDARDNormal operation — enforcement actions applied
PLAN_ONLYGenerate execution plan without running
SHADOWRun enforcement in shadow mode (log only, don't block)
AUTONOMOUSFully autonomous — minimal human oversight

ID Conventions

All IDs in Intutic use prefixed nanoid format:

PrefixEntity
wk_Workspace
mb_Member
tr_Trace
sp_SOP
vk_Virtual API Key
ev_Event
tc_Tool Call
in_Incident
an_Anomaly
se_Session

Never use raw UUIDs or Date.now() directly. Use newId(prefix) and newIso() from @intutic/id.


4. Proxy config.yaml (intutic_settings)

The proxy reads a LiteLLM-compatible config.yaml. Intutic-specific options live under the top-level intutic_settings key, alongside model_list and general_settings.

SettingTypeDefaultDescription
wasm_local_dirstring~/.intutic/wasmDirectory scanned for locally installed WASM rules. The INTUTIC_WASM_DIR env var takes precedence over this value

Model Routing (intutic_settings.routing)

Contextual bandit routing picks a model per request via Thompson sampling over a candidate pool.

SettingTypeDefaultDescription
enabledbooleanEnables bandit routing in standalone mode. Unset defers entirely to the control plane
candidate_modelsstring[]claude-3-5-sonnet, gpt-4o, gemini-2.0-flashCandidate model pool for Thompson sampling. Requests for models outside this pool bypass the bandit entirely. Names must match model_list entries so provider resolution and pricing stay accurate
anthropic_model_overridestringWhen set, any Anthropic-bound model is rewritten to this ID after routing. Unset leaves the routed model untouched

Precedence: when a control plane manages the workspace, the Valkey ff_bandit_routing feature flag is authoritative. routing.enabled applies only to standalone deployments where no control plane manages the workspace.

Reward Loop (intutic_settings.routing.reward)

Rewards are computed only from signals the proxy already observes — upstream success, latency versus SLO, token anomaly, and cost ratio. No LLM judge is involved.

SettingTypeDefaultDescription
enabledbooleantrueEnables the local deterministic reward loop
latency_slo_msnumber30000Latency SLO; responses slower than this are penalised proportionally. Includes full stream duration for streaming responses
latency_penaltynumber0.3Maximum reward deduction for exceeding the latency SLO
token_anomaly_penaltynumber0.2Reward deduction when the token-anomaly detector fires
cost_penaltynumber0.2Maximum reward deduction when the routed model costs more than the requested model would have
yaml
intutic_settings:
  # Directory scanned for locally installed WASM rules
  # (INTUTIC_WASM_DIR takes precedence over this value).
  wasm_local_dir: "~/.intutic/wasm"

  routing:
    enabled: true
    candidate_models: ["claude-3-5-sonnet", "gpt-4o", "gemini-2.0-flash"]
    reward:
      enabled: true
      latency_slo_ms: 30000
      latency_penalty: 0.3
      token_anomaly_penalty: 0.2
      cost_penalty: 0.2

The circuit breaker for AI agents