scriptoriumadmin/Roo-Code
1
0
Fork 0
mirror of https://github.com/RooVetGit/Roo-Code.git synced 2026-09-05 08:10:14 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Remove unnecessary variable, add JSDocs

Browse source
This commit is contained in:
Nissa Seru 2025-02-02 02:56:30 -05:00
parent fef95e10b8
commit 3492f79f37
1 changed files with 53 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

57
src/core/sliding-window/index.ts
View file

@ -1,6 +1,16 @@
import { Anthropic } from "@anthropic-ai/sdk"
import { ModelInfo } from "../../shared/api"
/**
* Truncates a conversation by removing a fraction of the messages.
*
* The first message is always retained, and a specified fraction (rounded to an even number)
* of messages from the beginning (excluding the first) is removed.
*
* @param {Anthropic.Messages.MessageParam[]} messages - The conversation messages.
* @param {number} fracToRemove - The fraction (between 0 and 1) of messages (excluding the first) to remove.
* @returns {Anthropic.Messages.MessageParam[]} The truncated conversation messages.
*/
export function truncateConversation(
messages: Anthropic.Messages.MessageParam[],
fracToRemove: number,
@ -14,6 +24,18 @@ export function truncateConversation(
return truncatedMessages
}
/**
* Conditionally truncates the conversation messages if the total token count exceeds the model's limit.
*
* Depending on whether the model supports prompt caching, different maximum token thresholds
* and truncation fractions are used. If the current total tokens exceed the threshold,
* the conversation is truncated using the appropriate fraction.
*
* @param {Anthropic.Messages.MessageParam[]} messages - The conversation messages.
* @param {number} totalTokens - The total number of tokens in the conversation.
* @param {ModelInfo} modelInfo - Model metadata including context window size and prompt cache support.
* @returns {Anthropic.Messages.MessageParam[]} The original or truncated conversation messages.
*/
export function truncateConversationIfNeeded(
messages: Anthropic.Messages.MessageParam[],
totalTokens: number,
@ -24,25 +46,52 @@ export function truncateConversationIfNeeded(
? messages
: truncateConversation(messages, getTruncFractionForPromptCachingModels(modelInfo))
} else {
const thresh = getMaxTokensForNonPromptCachingModels(modelInfo)
return totalTokens < thresh
return totalTokens < getMaxTokensForNonPromptCachingModels(modelInfo)
? messages
: truncateConversation(messages, getTruncFractionForNonPromptCachingModels(modelInfo))
}
}
/**
* Calculates the maximum allowed tokens for models that support prompt caching.
*
* The maximum is computed as the greater of (contextWindow - 40000) and 80% of the contextWindow.
*
* @param {ModelInfo} modelInfo - The model information containing the context window size.
* @returns {number} The maximum number of tokens allowed for prompt caching models.
*/
function getMaxTokensForPromptCachingModels(modelInfo: ModelInfo): number {
return Math.max(modelInfo.contextWindow - 40_000, modelInfo.contextWindow * 0.8)
}
/**
* Provides the fraction of messages to remove for models that support prompt caching.
*
* @param {ModelInfo} modelInfo - The model information (unused in current implementation).
* @returns {number} The truncation fraction for prompt caching models (fixed at 0.5).
*/
function getTruncFractionForPromptCachingModels(modelInfo: ModelInfo): number {
return Math.min(80_000, modelInfo.contextWindow * 0.4)
return 0.5
}
/**
* Calculates the maximum allowed tokens for models that do not support prompt caching.
*
* The maximum is computed as the greater of (contextWindow - 40000) and 80% of the contextWindow.
*
* @param {ModelInfo} modelInfo - The model information containing the context window size.
* @returns {number} The maximum number of tokens allowed for non-prompt caching models.
*/
function getMaxTokensForNonPromptCachingModels(modelInfo: ModelInfo): number {
return Math.max(modelInfo.contextWindow - 40_000, modelInfo.contextWindow * 0.8)
}
/**
* Provides the fraction of messages to remove for models that do not support prompt caching.
*
* @param {ModelInfo} modelInfo - The model information (unused in current implementation).
* @returns {number} The truncation fraction for non-prompt caching models (fixed at 0.1).
*/
function getTruncFractionForNonPromptCachingModels(modelInfo: ModelInfo): number {
return Math.min(80_000, modelInfo.contextWindow * 0.4)
return 0.1
}