MediaMetz/gemini-cli
1
0
Fork 0
mirror of https://github.com/google-gemini/gemini-cli.git synced 2026-05-13 05:12:55 -07:00
Code Issues Packages Projects Releases Wiki Activity

refactor(core): Unified context management settings schema (#24391)

Browse Source
This commit is contained in:
joshualitt
2026-04-01 10:24:45 -07:00
committed by GitHub
parent 3a32d9723e
commit 377e834e03
10 changed files with 215 additions and 212 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/cli/settings.md
+10 -11
View File
@@ -157,17 +157,16 @@ they appear in the UI.
### Experimental
| UI Label | Setting | Description | Default |
| -------------------------- | ---------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- | ------- |
| Enable Tool Output Masking | `experimental.toolOutputMasking.enabled` | Enables tool output masking to save tokens. | `true` |
| Enable Git Worktrees | `experimental.worktrees` | Enable automated Git worktree management for parallel work. | `false` |
| Use OSC 52 Paste | `experimental.useOSC52Paste` | Use OSC 52 for pasting. This may be more robust than the default system when using remote terminal sessions (if your terminal is configured to allow it). | `false` |
| Use OSC 52 Copy | `experimental.useOSC52Copy` | Use OSC 52 for copying. This may be more robust than the default system when using remote terminal sessions (if your terminal is configured to allow it). | `false` |
| Model Steering | `experimental.modelSteering` | Enable model steering (user hints) to guide the model during tool execution. | `false` |
| Direct Web Fetch | `experimental.directWebFetch` | Enable web fetch behavior that bypasses LLM summarization. | `false` |
| Memory Manager Agent | `experimental.memoryManager` | Replace the built-in save_memory tool with a memory manager subagent that supports adding, removing, de-duplicating, and organizing memories. | `false` |
| Enable Context Management | `experimental.contextManagement` | Enable logic for context management. | `false` |
| Topic & Update Narration | `experimental.topicUpdateNarration` | Enable the experimental Topic & Update communication model for reduced chattiness and structured progress reporting. | `false` |
| UI Label | Setting | Description | Default |
| ------------------------- | ----------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- | ------- |
| Enable Git Worktrees | `experimental.worktrees` | Enable automated Git worktree management for parallel work. | `false` |
| Use OSC 52 Paste | `experimental.useOSC52Paste` | Use OSC 52 for pasting. This may be more robust than the default system when using remote terminal sessions (if your terminal is configured to allow it). | `false` |
| Use OSC 52 Copy | `experimental.useOSC52Copy` | Use OSC 52 for copying. This may be more robust than the default system when using remote terminal sessions (if your terminal is configured to allow it). | `false` |
| Model Steering | `experimental.modelSteering` | Enable model steering (user hints) to guide the model during tool execution. | `false` |
| Direct Web Fetch | `experimental.directWebFetch` | Enable web fetch behavior that bypasses LLM summarization. | `false` |
| Memory Manager Agent | `experimental.memoryManager` | Replace the built-in save_memory tool with a memory manager subagent that supports adding, removing, de-duplicating, and organizing memories. | `false` |
| Enable Context Management | `experimental.contextManagement` | Enable logic for context management. | `false` |
| Topic & Update Narration | `experimental.topicUpdateNarration` | Enable the experimental Topic & Update communication model for reduced chattiness and structured progress reporting. | `false` |
### Skills
docs/reference/configuration.md
+23 -25
View File
@@ -1587,28 +1587,6 @@ their corresponding top-level category object in your `settings.json` file.
#### `experimental`
- **`experimental.toolOutputMasking.enabled`** (boolean):
- **Description:** Enables tool output masking to save tokens.
- **Default:** `true`
- **Requires restart:** Yes
- **`experimental.toolOutputMasking.toolProtectionThreshold`** (number):
- **Description:** Minimum number of tokens to protect from masking (most
recent tool outputs).
- **Default:** `50000`
- **Requires restart:** Yes
- **`experimental.toolOutputMasking.minPrunableTokensThreshold`** (number):
- **Description:** Minimum prunable tokens required to trigger a masking pass.
- **Default:** `30000`
- **Requires restart:** Yes
- **`experimental.toolOutputMasking.protectLatestTurn`** (boolean):
- **Description:** Ensures the absolute latest turn is never masked,
regardless of token count.
- **Default:** `true`
- **Requires restart:** Yes
- **`experimental.enableAgents`** (boolean):
- **Description:** Enable local and remote subagents.
- **Default:** `true`
@@ -1834,18 +1812,38 @@ their corresponding top-level category object in your `settings.json` file.
- **Default:** `0.25`
- **Requires restart:** Yes
- **`contextManagement.toolDistillation.maxOutputTokens`** (number):
- **Description:** Maximum tokens to show when truncating large tool outputs.
- **`contextManagement.tools.distillation.maxOutputTokens`** (number):
- **Description:** Maximum tokens to show to the model when truncating large
tool outputs.
- **Default:** `10000`
- **Requires restart:** Yes
- **`contextManagement.toolDistillation.summarizationThresholdTokens`**
- **`contextManagement.tools.distillation.summarizationThresholdTokens`**
(number):
- **Description:** Threshold above which truncated tool outputs will be
summarized by an LLM.
- **Default:** `20000`
- **Requires restart:** Yes
- **`contextManagement.tools.outputMasking.protectionThresholdTokens`**
(number):
- **Description:** Minimum number of tokens to protect from masking (most
recent tool outputs).
- **Default:** `50000`
- **Requires restart:** Yes
- **`contextManagement.tools.outputMasking.minPrunableThresholdTokens`**
(number):
- **Description:** Minimum prunable tokens required to trigger a masking pass.
- **Default:** `30000`
- **Requires restart:** Yes
- **`contextManagement.tools.outputMasking.protectLatestTurn`** (boolean):
- **Description:** Ensures the absolute latest turn is never masked,
regardless of token count.
- **Default:** `true`
- **Requires restart:** Yes
#### `admin`
- **`admin.secureModeEnabled`** (boolean):
packages/cli/src/config/config.ts
-1
View File
@@ -983,7 +983,6 @@ export async function loadCliConfig(
},
modelSteering: settings.experimental?.modelSteering,
topicUpdateNarration: settings.experimental?.topicUpdateNarration,
toolOutputMasking: settings.experimental?.toolOutputMasking,
noBrowser: !!process.env['NO_BROWSER'],
summarizeToolOutput: settings.model?.summarizeToolOutput,
ideMode,
packages/cli/src/config/settingsSchema.ts
+66 -65
View File
@@ -1933,58 +1933,6 @@ const SETTINGS_SCHEMA = {
description: 'Setting to enable experimental features',
showInDialog: false,
properties: {
toolOutputMasking: {
type: 'object',
label: 'Tool Output Masking',
category: 'Experimental',
requiresRestart: true,
ignoreInDocs: false,
default: {},
description:
'Advanced settings for tool output masking to manage context window efficiency.',
showInDialog: false,
properties: {
enabled: {
type: 'boolean',
label: 'Enable Tool Output Masking',
category: 'Experimental',
requiresRestart: true,
default: true,
description: 'Enables tool output masking to save tokens.',
showInDialog: true,
},
toolProtectionThreshold: {
type: 'number',
label: 'Tool Protection Threshold',
category: 'Experimental',
requiresRestart: true,
default: 50000,
description:
'Minimum number of tokens to protect from masking (most recent tool outputs).',
showInDialog: false,
},
minPrunableTokensThreshold: {
type: 'number',
label: 'Min Prunable Tokens Threshold',
category: 'Experimental',
requiresRestart: true,
default: 30000,
description:
'Minimum prunable tokens required to trigger a masking pass.',
showInDialog: false,
},
protectLatestTurn: {
type: 'boolean',
label: 'Protect Latest Turn',
category: 'Experimental',
requiresRestart: true,
default: true,
description:
'Ensures the absolute latest turn is never masked, regardless of token count.',
showInDialog: false,
},
},
},
enableAgents: {
type: 'boolean',
label: 'Enable Agents',
@@ -2544,33 +2492,86 @@ const SETTINGS_SCHEMA = {
},
},
},
toolDistillation: {
tools: {
type: 'object',
label: 'Tool Distillation',
label: 'Context Management Tools',
category: 'Context Management',
requiresRestart: true,
default: {},
showInDialog: false,
properties: {
maxOutputTokens: {
type: 'number',
label: 'Max Output Tokens',
distillation: {
type: 'object',
label: 'Tool Distillation',
category: 'Context Management',
requiresRestart: true,
default: 10_000,
description:
'Maximum tokens to show when truncating large tool outputs.',
default: {},
showInDialog: false,
properties: {
maxOutputTokens: {
type: 'number',
label: 'Max Output Tokens',
category: 'Context Management',
requiresRestart: true,
default: 10_000,
description:
'Maximum tokens to show to the model when truncating large tool outputs.',
showInDialog: false,
},
summarizationThresholdTokens: {
type: 'number',
label: 'Tool Summarization Threshold',
category: 'Context Management',
requiresRestart: true,
default: 20_000,
description:
'Threshold above which truncated tool outputs will be summarized by an LLM.',
showInDialog: false,
},
},
},
summarizationThresholdTokens: {
type: 'number',
label: 'Tool Summarization Threshold',
outputMasking: {
type: 'object',
label: 'Tool Output Masking',
category: 'Context Management',
requiresRestart: true,
default: 20_000,
ignoreInDocs: false,
default: {},
description:
'Threshold above which truncated tool outputs will be summarized by an LLM.',
'Advanced settings for tool output masking to manage context window efficiency.',
showInDialog: false,
properties: {
protectionThresholdTokens: {
type: 'number',
label: 'Tool Protection Threshold (Tokens)',
category: 'Context Management',
requiresRestart: true,
default: 50_000,
description:
'Minimum number of tokens to protect from masking (most recent tool outputs).',
showInDialog: false,
},
minPrunableThresholdTokens: {
type: 'number',
label: 'Min Prunable Tokens Threshold',
category: 'Context Management',
requiresRestart: true,
default: 30_000,
description:
'Minimum prunable tokens required to trigger a masking pass.',
showInDialog: false,
},
protectLatestTurn: {
type: 'boolean',
label: 'Protect Latest Turn',
category: 'Context Management',
requiresRestart: true,
default: true,
description:
'Ensures the absolute latest turn is never masked, regardless of token count.',
showInDialog: false,
},
},
},
},
},
packages/core/src/config/config.ts
+46 -42
View File
@@ -206,6 +206,12 @@ export interface OutputSettings {
format?: OutputFormat;
}
export interface ToolOutputMaskingConfig {
protectionThresholdTokens: number;
minPrunableThresholdTokens: number;
protectLatestTurn: boolean;
}
export interface ContextManagementConfig {
enabled: boolean;
historyWindow: {
@@ -217,19 +223,15 @@ export interface ContextManagementConfig {
retainedMaxTokens: number;
normalizationHeadRatio: number;
};
toolDistillation: {
maxOutputTokens: number;
summarizationThresholdTokens: number;
tools: {
distillation: {
maxOutputTokens: number;
summarizationThresholdTokens: number;
};
outputMasking: ToolOutputMaskingConfig;
};
}
export interface ToolOutputMaskingConfig {
enabled: boolean;
toolProtectionThreshold: number;
minPrunableTokensThreshold: number;
protectLatestTurn: boolean;
}
export interface GemmaModelRouterSettings {
enabled?: boolean;
classifier?: {
@@ -711,7 +713,7 @@ export interface ConfigParameters {
experimentalAgentHistorySummarization?: boolean;
memoryBoundaryMarkers?: string[];
topicUpdateNarration?: boolean;
toolOutputMasking?: Partial<ToolOutputMaskingConfig>;
disableLLMCorrection?: boolean;
plan?: boolean;
tracker?: boolean;
@@ -913,7 +915,7 @@ export class Config implements McpContext, AgentLoopContext {
private pendingIncludeDirectories: string[];
private readonly enableHooks: boolean;
private readonly enableHooksUI: boolean;
private readonly toolOutputMasking: ToolOutputMaskingConfig;
private hooks: { [K in HookEventName]?: HookDefinition[] } | undefined;
private projectHooks:
| ({ [K in HookEventName]?: HookDefinition[] } & { disabled?: string[] })
@@ -1162,12 +1164,27 @@ export class Config implements McpContext, AgentLoopContext {
params.contextManagement?.messageLimits?.normalizationHeadRatio ??
0.25,
},
toolDistillation: {
maxOutputTokens:
params.contextManagement?.toolDistillation?.maxOutputTokens ?? 10000,
summarizationThresholdTokens:
params.contextManagement?.toolDistillation
?.summarizationThresholdTokens ?? 20000,
tools: {
distillation: {
maxOutputTokens:
params.contextManagement?.tools?.distillation?.maxOutputTokens ??
10000,
summarizationThresholdTokens:
params.contextManagement?.tools?.distillation
?.summarizationThresholdTokens ?? 20000,
},
outputMasking: {
protectionThresholdTokens:
params.contextManagement?.tools?.outputMasking
?.protectionThresholdTokens ?? DEFAULT_TOOL_PROTECTION_THRESHOLD,
minPrunableThresholdTokens:
params.contextManagement?.tools?.outputMasking
?.minPrunableThresholdTokens ??
DEFAULT_MIN_PRUNABLE_TOKENS_THRESHOLD,
protectLatestTurn:
params.contextManagement?.tools?.outputMasking?.protectLatestTurn ??
DEFAULT_PROTECT_LATEST_TURN,
},
},
};
this.topicUpdateNarration = params.topicUpdateNarration ?? false;
@@ -1176,18 +1193,6 @@ export class Config implements McpContext, AgentLoopContext {
this.isModelSteeringEnabled(),
);
ExecutionLifecycleService.setInjectionService(this.injectionService);
this.toolOutputMasking = {
enabled: params.toolOutputMasking?.enabled ?? true,
toolProtectionThreshold:
params.toolOutputMasking?.toolProtectionThreshold ??
DEFAULT_TOOL_PROTECTION_THRESHOLD,
minPrunableTokensThreshold:
params.toolOutputMasking?.minPrunableTokensThreshold ??
DEFAULT_MIN_PRUNABLE_TOKENS_THRESHOLD,
protectLatestTurn:
params.toolOutputMasking?.protectLatestTurn ??
DEFAULT_PROTECT_LATEST_TURN,
};
this.maxSessionTurns = params.maxSessionTurns ?? -1;
this.acpMode = params.acpMode ?? false;
this.listSessions = params.listSessions ?? false;
@@ -2415,10 +2420,6 @@ export class Config implements McpContext, AgentLoopContext {
return this.modelSteering;
}
getToolOutputMaskingEnabled(): boolean {
return this.toolOutputMasking.enabled;
}
async getToolOutputMaskingConfig(): Promise<ToolOutputMaskingConfig> {
await this.ensureExperimentsLoaded();
@@ -2440,17 +2441,19 @@ export class Config implements McpContext, AgentLoopContext {
: undefined;
return {
enabled: this.toolOutputMasking.enabled,
toolProtectionThreshold:
protectionThresholdTokens:
parsedProtection !== undefined && !isNaN(parsedProtection)
? parsedProtection
: this.toolOutputMasking.toolProtectionThreshold,
minPrunableTokensThreshold:
: this.contextManagement.tools.outputMasking
.protectionThresholdTokens,
minPrunableThresholdTokens:
parsedPrunable !== undefined && !isNaN(parsedPrunable)
? parsedPrunable
: this.toolOutputMasking.minPrunableTokensThreshold,
: this.contextManagement.tools.outputMasking
.minPrunableThresholdTokens,
protectLatestTurn:
remoteProtectLatest ?? this.toolOutputMasking.protectLatestTurn,
remoteProtectLatest ??
this.contextManagement.tools.outputMasking.protectLatestTurn,
};
}
@@ -3301,11 +3304,12 @@ export class Config implements McpContext, AgentLoopContext {
}
getToolMaxOutputTokens(): number {
return this.contextManagement.toolDistillation.maxOutputTokens;
return this.contextManagement.tools.distillation.maxOutputTokens;
}
getToolSummarizationThresholdTokens(): number {
return this.contextManagement.toolDistillation.summarizationThresholdTokens;
return this.contextManagement.tools.distillation
.summarizationThresholdTokens;
}
getNextCompressionTruncationId(): number {
packages/core/src/context/toolOutputMaskingService.test.ts
+4 -5
View File
@@ -45,11 +45,10 @@ describe('ToolOutputMaskingService', () => {
},
getSessionId: () => 'mock-session',
getUsageStatisticsEnabled: () => false,
getToolOutputMaskingEnabled: () => true,
getToolOutputMaskingConfig: async () => ({
enabled: true,
toolProtectionThreshold: 50000,
minPrunableTokensThreshold: 30000,
protectionThresholdTokens: 50000,
minPrunableThresholdTokens: 30000,
protectLatestTurn: true,
}),
} as unknown as Config;
@@ -66,8 +65,8 @@ describe('ToolOutputMaskingService', () => {
it('should respect remote configuration overrides', async () => {
mockConfig.getToolOutputMaskingConfig = async () => ({
enabled: true,
toolProtectionThreshold: 100, // Very low threshold
minPrunableTokensThreshold: 50,
protectionThresholdTokens: 100, // Very low threshold
minPrunableThresholdTokens: 50,
protectLatestTurn: false,
});
packages/core/src/context/toolOutputMaskingService.ts
+7 -7
View File
@@ -53,13 +53,13 @@ export interface MaskingResult {
*
* It implements a "Hybrid Backward Scanned FIFO" algorithm to balance context relevance with
* token savings:
* 1. **Protection Window**: Protects the newest `toolProtectionThreshold` (default 50k) tool tokens
* 1. **Protection Window**: Protects the newest `protectionThresholdTokens` (default 50k) tool tokens
* from pruning. Optionally skips the entire latest conversation turn to ensure full context for
* the model's next response.
* 2. **Global Aggregation**: Scans backwards past the protection window to identify all remaining
* tool outputs that haven't been masked yet.
* 3. **Batch Trigger**: Trigger masking only if the total prunable tokens exceed
* `minPrunableTokensThreshold` (default 30k).
* `minPrunableThresholdTokens` (default 30k).
*
* @remarks
* Effectively, this means masking only starts once the conversation contains approximately 80k
@@ -71,11 +71,11 @@ export class ToolOutputMaskingService {
history: readonly Content[],
config: Config,
): Promise<MaskingResult> {
const maskingConfig = await config.getToolOutputMaskingConfig();
if (!maskingConfig.enabled || history.length === 0) {
if (history.length === 0) {
return { newHistory: history, maskedCount: 0, tokensSaved: 0 };
}
const maskingConfig = await config.getToolOutputMaskingConfig();
let cumulativeToolTokens = 0;
let protectionBoundaryReached = false;
let totalPrunableTokens = 0;
@@ -124,7 +124,7 @@ export class ToolOutputMaskingService {
if (!protectionBoundaryReached) {
cumulativeToolTokens += partTokens;
if (cumulativeToolTokens > maskingConfig.toolProtectionThreshold) {
if (cumulativeToolTokens > maskingConfig.protectionThresholdTokens) {
protectionBoundaryReached = true;
// The part that crossed the boundary is prunable.
totalPrunableTokens += partTokens;
@@ -151,12 +151,12 @@ export class ToolOutputMaskingService {
// Trigger pruning only if we have accumulated enough savings to justify the
// overhead of masking and file I/O (batch pruning threshold).
if (totalPrunableTokens < maskingConfig.minPrunableTokensThreshold) {
if (totalPrunableTokens < maskingConfig.minPrunableThresholdTokens) {
return { newHistory: history, maskedCount: 0, tokensSaved: 0 };
}
debugLogger.debug(
`[ToolOutputMasking] Triggering masking. Prunable tool tokens: ${totalPrunableTokens.toLocaleString()} (> ${maskingConfig.minPrunableTokensThreshold.toLocaleString()})`,
`[ToolOutputMasking] Triggering masking. Prunable tool tokens: ${totalPrunableTokens.toLocaleString()} (> ${maskingConfig.minPrunableThresholdTokens.toLocaleString()})`,
);
// Perform masking and offloading
packages/core/src/core/client.test.ts
+5 -1
View File
@@ -220,8 +220,12 @@ describe('Gemini Client (client.ts)', () => {
getSessionMemory: vi.fn().mockReturnValue(''),
isJitContextEnabled: vi.fn().mockReturnValue(false),
getContextManager: vi.fn().mockReturnValue(undefined),
getToolOutputMaskingEnabled: vi.fn().mockReturnValue(false),
getDisableLoopDetection: vi.fn().mockReturnValue(false),
getToolOutputMaskingConfig: vi.fn().mockReturnValue({
protectionThresholdTokens: 50000,
minPrunableThresholdTokens: 30000,
protectLatestTurn: true,
}),
getSessionId: vi.fn().mockReturnValue('test-session-id'),
getProxy: vi.fn().mockReturnValue(undefined),
packages/core/src/core/client.ts
-3
View File
@@ -1231,9 +1231,6 @@ export class GeminiClient {
* Masks bulky tool outputs to save context window space.
*/
private async tryMaskToolOutputs(history: readonly Content[]): Promise<void> {
if (!this.config.getToolOutputMaskingEnabled()) {
return;
}
const result = await this.toolOutputMaskingService.mask(
history,
this.config,
schemas/settings.schema.json
+54 -52
View File
@@ -2748,44 +2748,6 @@
"default": {},
"type": "object",
"properties": {
"toolOutputMasking": {
"title": "Tool Output Masking",
"description": "Advanced settings for tool output masking to manage context window efficiency.",
"markdownDescription": "Advanced settings for tool output masking to manage context window efficiency.\n\n- Category: `Experimental`\n- Requires restart: `yes`\n- Default: `{}`",
"default": {},
"type": "object",
"properties": {
"enabled": {
"title": "Enable Tool Output Masking",
"description": "Enables tool output masking to save tokens.",
"markdownDescription": "Enables tool output masking to save tokens.\n\n- Category: `Experimental`\n- Requires restart: `yes`\n- Default: `true`",
"default": true,
"type": "boolean"
},
"toolProtectionThreshold": {
"title": "Tool Protection Threshold",
"description": "Minimum number of tokens to protect from masking (most recent tool outputs).",
"markdownDescription": "Minimum number of tokens to protect from masking (most recent tool outputs).\n\n- Category: `Experimental`\n- Requires restart: `yes`\n- Default: `50000`",
"default": 50000,
"type": "number"
},
"minPrunableTokensThreshold": {
"title": "Min Prunable Tokens Threshold",
"description": "Minimum prunable tokens required to trigger a masking pass.",
"markdownDescription": "Minimum prunable tokens required to trigger a masking pass.\n\n- Category: `Experimental`\n- Requires restart: `yes`\n- Default: `30000`",
"default": 30000,
"type": "number"
},
"protectLatestTurn": {
"title": "Protect Latest Turn",
"description": "Ensures the absolute latest turn is never masked, regardless of token count.",
"markdownDescription": "Ensures the absolute latest turn is never masked, regardless of token count.\n\n- Category: `Experimental`\n- Requires restart: `yes`\n- Default: `true`",
"default": true,
"type": "boolean"
}
},
"additionalProperties": false
},
"enableAgents": {
"title": "Enable Agents",
"description": "Enable local and remote subagents.",
@@ -3190,25 +3152,65 @@
},
"additionalProperties": false
},
"toolDistillation": {
"title": "Tool Distillation",
"tools": {
"title": "Context Management Tools",
"markdownDescription": "Description not provided.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `{}`",
"default": {},
"type": "object",
"properties": {
"maxOutputTokens": {
"title": "Max Output Tokens",
"description": "Maximum tokens to show when truncating large tool outputs.",
"markdownDescription": "Maximum tokens to show when truncating large tool outputs.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `10000`",
"default": 10000,
"type": "number"
"distillation": {
"title": "Tool Distillation",
"markdownDescription": "Description not provided.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `{}`",
"default": {},
"type": "object",
"properties": {
"maxOutputTokens": {
"title": "Max Output Tokens",
"description": "Maximum tokens to show to the model when truncating large tool outputs.",
"markdownDescription": "Maximum tokens to show to the model when truncating large tool outputs.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `10000`",
"default": 10000,
"type": "number"
},
"summarizationThresholdTokens": {
"title": "Tool Summarization Threshold",
"description": "Threshold above which truncated tool outputs will be summarized by an LLM.",
"markdownDescription": "Threshold above which truncated tool outputs will be summarized by an LLM.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `20000`",
"default": 20000,
"type": "number"
}
},
"additionalProperties": false
},
"summarizationThresholdTokens": {
"title": "Tool Summarization Threshold",
"description": "Threshold above which truncated tool outputs will be summarized by an LLM.",
"markdownDescription": "Threshold above which truncated tool outputs will be summarized by an LLM.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `20000`",
"default": 20000,
"type": "number"
"outputMasking": {
"title": "Tool Output Masking",
"description": "Advanced settings for tool output masking to manage context window efficiency.",
"markdownDescription": "Advanced settings for tool output masking to manage context window efficiency.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `{}`",
"default": {},
"type": "object",
"properties": {
"protectionThresholdTokens": {
"title": "Tool Protection Threshold (Tokens)",
"description": "Minimum number of tokens to protect from masking (most recent tool outputs).",
"markdownDescription": "Minimum number of tokens to protect from masking (most recent tool outputs).\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `50000`",
"default": 50000,
"type": "number"
},
"minPrunableThresholdTokens": {
"title": "Min Prunable Tokens Threshold",
"description": "Minimum prunable tokens required to trigger a masking pass.",
"markdownDescription": "Minimum prunable tokens required to trigger a masking pass.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `30000`",
"default": 30000,
"type": "number"
},
"protectLatestTurn": {
"title": "Protect Latest Turn",
"description": "Ensures the absolute latest turn is never masked, regardless of token count.",
"markdownDescription": "Ensures the absolute latest turn is never masked, regardless of token count.\n\n- Category: `Context Management`\n- Requires restart: `yes`\n- Default: `true`",
"default": true,
"type": "boolean"
}
},
"additionalProperties": false
}
},
"additionalProperties": false