Documentation
¶
Index ¶
Constants ¶
View Source
const ( // Default authentication tokens DefaultUserToken = "tingly-box-user-token" DefaultModelToken = "tingly-box-model-token" DefaultVirtualModelToken = "tingly-virtual-model-token" // Default mode name DefaultModeName = "tingly" )
View Source
const ( LogDirName = "log" // DebugLogFileName is the name of the debug log file DebugLogFileName = "bad_requests.log" // DefaultRequestTimeout is the default timeout for HTTP requests in seconds DefaultRequestTimeout = 1800 // DefaultMaxTimeout in seconds DefaultMaxTimeout = 30 * 60 // ModelFetchTimeout is the timeout for fetching models from provider API in seconds ModelFetchTimeout = 30 // DefaultMaxTokens is the default max_tokens value for API requests DefaultMaxTokens = 8192 )
View Source
const ( DefaultLatencyThresholdMs = int64(2000) // Default latency threshold in milliseconds DefaultLatencySampleWindow = 100 // Default number of latency samples to keep DefaultLatencyPercentile = 0.95 // Default percentile for latency comparison (0.95 = p95) DefaultLatencyComparisonMode = "avg" // Default comparison mode: "avg", "p50", "p95", "p99" )
Latency-based routing defaults
View Source
const ( DefaultMinSpeedSamples = 5 // Minimum samples required before making speed-based decisions DefaultSpeedThresholdTps = 50.0 // Minimum acceptable tokens per second DefaultSpeedSampleWindow = 100 // Default number of speed samples to keep )
Token speed-based routing defaults
View Source
const ( DefaultTTFTThresholdMs = int64(500) // Default TTFT threshold in milliseconds DefaultTTFTSampleWindow = 100 // Default number of TTFT samples to keep DefaultTTFTPercentile = 0.95 // Default percentile for TTFT comparison (0.95 = p95) DefaultTTFTComparisonMode = "p95" // Default comparison mode: "avg", "p50", "p95", "p99" )
TTFT (Time To First Token) based routing defaults
View Source
const ( DefaultMinCacheHitRate = 0.3 // Minimum acceptable cache hit rate (30%) DefaultMinCacheSamples = 10 // Minimum samples before using cache data )
Cache-aware routing defaults
View Source
const ( DefaultMaxCostTokens = int64(50000) // Default max token cost per window DefaultCostWindowSize = 3600 // Default cost window in seconds (1 hour) )
Cost-optimized routing defaults
View Source
const ( DefaultLatencyWeight = 0.25 // Weight for total latency in adaptive scoring DefaultTTFTWeight = 0.20 // Weight for TTFT in adaptive scoring DefaultCacheWeight = 0.15 // Weight for cache hit rate in adaptive scoring DefaultSpeedWeight = 0.15 // Weight for token speed in adaptive scoring DefaultTokenWeight = 0.15 // Weight for token usage in adaptive scoring DefaultHealthWeight = 0.10 // Weight for health status in adaptive scoring DefaultScoringMode = "weighted_sum" // Default scoring mode: "weighted_sum", "multiplicative", "rank_based" )
Adaptive routing defaults (updated with new dimensions)
View Source
const ConfigDirName = ".tingly-box"
View Source
const DBDirName = "db"
View Source
const DBFileName = "tingly.db" // Unified SQLite database file
View Source
const DefaultTokenThreshold = int64(10000) // Default token threshold for token-based tactics
Load balancing threshold defaults
View Source
const MemoryDirName = "memory"
Variables ¶
This section is empty.
Functions ¶
func GetMemoryDir ¶
GetMemoryDir returns the memory directory path
func GetTinglyConfDir ¶
func GetTinglyConfDir() string
GetTinglyConfDir returns the config directory path (default: ~/.tingly-box)
Types ¶
This section is empty.
Click to show internal directories.
Click to hide internal directories.