Open source file-system and knowledge based agent template. Build AI agents that stay up to date with your knowledge base
export interface SavoirConfig {
/** Base URL of your Savoir API (e.g., 'https://chat.example.com') */
apiUrl: string
/** API key for authentication. Create one from the admin panel (/admin/api-keys) or user settings (/settings/api-keys). */
apiKey?: string
/** Extra headers to include in every request (e.g. forwarding cookies for internal server-to-server calls). */
headers?: Record<string, string>
/** Reuse an existing sandbox session instead of creating a new one. */
sessionId?: string
/** Usage source identifier (e.g. 'github-bot'). Defaults to 'sdk'. */
source?: string
/** Optional tracking ID (e.g. 'issue-123'). Can be overridden per reportUsage() call. */
sourceId?: string
}
export interface SearchResult {
path: string
lineNumber: number
content: string
}
export interface FileContent {
path: string
content: string
matchCount?: number
totalLines?: number
}
export interface SearchAndReadResponse {
sessionId: string
matches: SearchResult[]
files: FileContent[]
}
export interface ReadResponse {
sessionId: string
files: FileContent[]
}
export interface ShellResponse {
sessionId: string
stdout: string
stderr: string
exitCode: number
}
export interface ShellCommandResult {
command: string
stdout: string
stderr: string
exitCode: number
}
export interface ShellBatchResponse {
sessionId: string
results: ShellCommandResult[]
}
export interface ApiErrorResponse {
statusCode: number
message: string
error?: string
}
export interface SyncSource {
id: string
type: 'github' | 'youtube'
label: string
repo?: string | null
branch?: string | null
contentPath?: string | null
outputPath?: string | null
readmeOnly?: boolean | null
channelId?: string | null
handle?: string | null
maxVideos?: number | null
}
export interface SyncOptions {
/** Clear all content before sync (default: false) */
reset?: boolean
/** Push to snapshot repo after sync (default: true) */
push?: boolean
/** Sources to sync (passed from chat app DB) */
sources?: SyncSource[]
/** Filter to sync only a specific source by ID */
sourceFilter?: string
}
export interface SyncResponse {
status: 'started'
message: string
options: {
reset: boolean
push: boolean
}
}
export interface SnapshotResponse {
status: 'started'
message: string
}
export interface GitHubSource {
id: string
label: string
type: 'github'
repo: string
branch: string
outputPath: string
readmeOnly: boolean
}
export interface YouTubeSource {
id: string
label: string
type: 'youtube'
channelId: string
handle: string
}
export interface SourcesResponse {
total: number
github: {
count: number
sources: GitHubSource[]
}
youtube: {
count: number
sources: YouTubeSource[]
}
}
export interface SyncSourceResponse {
status: 'started'
message: string
source: string
options: {
reset: boolean
push: boolean
sourceFilter: string
}
}
/** Duck-typed AI SDK generate result — works with GenerateTextResult/StreamTextResult without importing ai. */
export interface GenerateResult {
totalUsage: { inputTokens?: number, outputTokens?: number }
response: { modelId?: string }
}
export interface ReportUsageOptions {
/** Override the sourceId set in config. */
sourceId?: string
/** Start time (Date.now()) to compute durationMs automatically. */
startTime?: number
/** Explicit duration in ms (takes precedence over startTime). */
durationMs?: number
/** Extra metadata to attach to the usage record. */
metadata?: Record<string, unknown>
}
/** Admin-defined settings that customize agent behavior (fetched from /api/agent-config/public). */
export interface AgentConfig {
id: string
name: string
additionalPrompt: string | null
responseStyle: 'concise' | 'detailed' | 'technical' | 'friendly'
language: string
defaultModel: string | null
/** Multiplier applied to router-determined maxSteps (0.5-3.0) */
maxStepsMultiplier: number
temperature: number
searchInstructions: string | null
citationFormat: 'inline' | 'footnote' | 'none'
isActive: boolean
}