Skip to content

Latest commit

 

History

History
197 lines (148 loc) · 10 KB

File metadata and controls

197 lines (148 loc) · 10 KB

Configuration Reference

Generated from the config domain definitions in packages/synergy/src/config/domain.ts and the Zod schema in packages/synergy/src/config/schema.ts. Concept and layout guidance lives in Configuration layout.

Domains

Domain File Merge policy
general 00-general.jsonc merge
models 10-models.jsonc merge
providers 20-providers.jsonc merge
library 30-library.jsonc merge
mcp 40-mcp.jsonc merge
plugins 50-plugins.jsonc merge
skills 55-skills.jsonc merge
agents 60-agents.jsonc merge
commands 70-commands.jsonc merge
permissions 80-permissions.jsonc merge
channels 90-channels.jsonc merge
holos 100-holos.jsonc merge
email 110-email.jsonc merge
github 115-github.jsonc merge
runtime 120-runtime.jsonc merge

General

File: 00-general.jsonc · Merge: merge

Key Type Description
$schema string (optional) JSON schema reference for configuration validation
theme string (optional) Theme name to use for the interface
keybinds Keybinds.optional (optional) Custom keybind configurations
toast "info" | "success" | "warning" | "error" (optional) Toast notification preferences
logLevel Log.Level.optional (optional) Log level
snapshot boolean (optional)
compactReasoning boolean (optional) Show live reasoning in a compact single-line viewport
username string (optional) Custom username to display in conversations instead of system username
locale "system" | "en" | "zh-CN" (optional) UI locale (system = follow OS, default: system)
activityDisplay "full" | "balanced" | "minimal" (optional) How much activity detail to show in the interface: full = everything, balanced = semantic activity grouping, minimal = only essential activity (default: balanced)
defaultSessionWorkspace "main" | "worktree" (optional) Default workspace for new sessions started from the Web composer: main = run in the main checkout, worktree = start each new session in an isolated git worktree (default: main). Programmatic session creation (API, channels, Cortex) always uses the main checkout.
layout Layout.optional (optional) @deprecated Always uses stretch layout.
embedding EmbeddingConfig
rerank RerankConfig

Models

File: 10-models.jsonc · Merge: merge

Key Type Description
model string (optional) Default model in the format of provider/model, eg anthropic/claude-sonnet-4-5
nano_model string (optional) Cheapest model for trivial extraction tasks like title generation, in the format of provider/model. Falls back to mini_model → mid_model → model.
mini_model string (optional) Lightweight model for simple tasks like intent extraction, in the format of provider/model. Falls back to mid_model → model.
mid_model string (optional) Mid-tier model for internal agents that need moderate reasoning (script extraction, reward evaluation, code exploration), in the format of provider/model. Falls back to the default model.
thinking_model string (optional) Deep thinking model for complex reasoning and architecture tasks, in the format of provider/model. Falls back to the default model if not set.
long_context_model string (optional) Model with extra-large context window for processing very long inputs, in the format of provider/model. Falls back to the default model if not set.
creative_model string (optional) Model for creative and visual tasks (UI design, writing, artistry), in the format of provider/model. Falls back to the default model if not set.
vision_model string (optional) Model for separate image analysis via the look_at tool, in the format of provider/model. If not set, look_at is disabled. Direct current-model image context uses view_image based on the active model capability.
role_variant string (optional) Default variant (e.g. low, medium, high, xhigh) applied per model role. Requires the resolved model to support the named variant.
quick_switcher QuickSwitcher.optional (optional) Quick switcher model visibility preferences

Providers

File: 20-providers.jsonc · Merge: merge

Key Type Description
provider string (optional) Custom provider configurations and model overrides
providerCatalog ProviderCatalog.optional (optional) Signed remote provider catalog configuration
enabled_providers string (optional) When non-empty, ONLY these providers will be enabled. Empty arrays are ignored in each config layer, preserving lower-priority filters
disabled_providers string (optional) Disable providers that are loaded automatically. Empty arrays are ignored in each config layer, preserving lower-priority filters

Library

File: 30-library.jsonc · Merge: merge

Key Type Description
library LibraryConfig

MCP

File: 40-mcp.jsonc · Merge: merge

Key Type Description
mcp object (optional) MCP (Model Context Protocol) server configurations
mcpDefaults McpDefaults.optional (optional)

Plugins

File: 50-plugins.jsonc · Merge: merge

Key Type Description
plugin string (optional)
pluginConfig string (optional) Per-plugin configuration namespaces. Keys are plugin IDs, values are plugin-specific config.
pluginRuntimePolicy PluginRuntimePolicy.optional (optional) Plugin runtime isolation policy configuration
pluginMarketplace PluginMarketplace.optional (optional) Public plugin marketplace registry configuration

Skills

File: 55-skills.jsonc · Merge: merge

Key Type Description
skills SkillsConfig

Agents

File: 60-agents.jsonc · Merge: merge

Key Type Description
default_agent string (optional) Default agent to use when none is specified. Must be a primary agent. Falls back to 'synergy' if not set or if the specified agent is invalid.
agent object (optional) Agent configuration
external_agent string (optional) External agent configurations (e.g. codex, claude-code)
instructions string (optional) Additional instruction files or patterns to include
project_doc_fallback_filenames string (optional) Ordered fallback instruction filenames to try when AGENTS.md is missing in a directory
project_doc_max_bytes number (optional) Maximum bytes to include from each automatically discovered instruction file (default: 32768; 0 disables automatic discovery)
category string (optional) Custom category configurations for background tasks. Categories define model and prompt presets.

Commands

File: 70-commands.jsonc · Merge: merge

Key Type Description
command string

Permissions

File: 80-permissions.jsonc · Merge: merge

Key Type Description
permission Permission.optional (optional)
tools string (optional)
controlProfile ControlProfileId.optional (optional) Default control profile applied to all agents
sandbox SandboxConfig.optional (optional) Sandbox configuration for workspace boundary enforcement
smartAllow boolean (optional) Use the SmartAllow internal agent to auto-allow high-confidence safe asks in guarded mode and eligible false-positive denies in autonomous mode using metadata or redacted evidence only; full_access does not need SmartAllow

Channels

File: 90-channels.jsonc · Merge: merge

Key Type Description
channel string (optional) Channel configurations for messaging platform integrations

Holos

File: 100-holos.jsonc · Merge: merge

Key Type Description
holos Holos.optional (optional) Holos platform configuration
enterprise object (optional) Enterprise URL

Email

File: 110-email.jsonc · Merge: merge

Key Type Description
email Email.optional (optional) Outgoing email configuration

GitHub

File: 115-github.jsonc · Merge: merge

Key Type Description
github Github.optional (optional) GitHub integration settings (git identity sync, agenda watch)

Runtime

File: 120-runtime.jsonc · Merge: merge

Key Type Description
server Server.optional (optional) Server configuration for synergy serve and web commands
timeout number (optional) Seconds before unanswered questions auto-expire (0 = no timeout, default 3600 = 1h)
cortex object (optional) Cortex task scheduling configuration
execution object (optional) Maximum queued Agent turns waiting for a worker (default: 256)
watcher object (optional)
formatter object (optional)
lsp object (optional)
lspWriteDiagnostics boolean (optional) Include LSP diagnostics after file-writing tools complete (default: true)
lspDiagnostics "error" | "warning" (optional) Severity and scope policy for diagnostics returned after file-writing tools
question object (optional) Seconds before unanswered questions auto-expire (0 = no timeout, default 3600 = 1h)
compaction object (optional) Maximum number of historical images to send as base64 per request (older images replaced with text placeholders). Default: 8.
experimental object (optional) Enable Runtime Boss Mode: auto-provision a home-scope runtime boss session and route all Feishu messages to it
observability ObservabilityConfig.optional (optional) Local logs, indexed telemetry, and diagnostics settings