InteractionContextSettings

How an interaction keeps its conversation within the model’s context window, and how many rounds of tool calls an agent may make in one turn.

Kind: an object, used in requests and responses. In a request it is called InteractionContextSettingsInput.

In a request

Field	Type	Required	Description
mode	ContextTrimmingMode		What happens when the conversation no longer fits the model’s context window.
targetUtilization	number		How full the context window may get, as a fraction from 0 to 1, before the conversation is shortened.
toolResultTrimmingMode	ToolResultTrimmingMode		When the results of earlier tool calls are replaced, to save space.
toolResultTrimmingMaxCycles	integer | null		With toolResultTrimmingMode set to AfterCycles or Both: once an agent has had this many turns, every tool result but the latest is replaced.
toolResultTrimmingStyle	ToolResultTrimmingStyle| null		What a replaced tool result becomes. Null: Placeholder.
toolResultTrimmingReplacement	string		The text a replaced tool result becomes, with the Placeholder style.
compactionType	CompactionType| null		Whether, and how, older parts of a long conversation are summarised to make room. Suits long, many-turn conversations such as chats. Null: no summaries (None).
lowWaterUtilization	number | null		With compaction on: how far a forced trim shortens the conversation, as a fraction of the context window from 0 to 1, so that it is not trimmed again on the very next turn. Null: 0.2. Never more than targetUtilization.
compactAtUtilization	number		How full the context window gets, as a fraction from 0 to 1, before the conversation is summarised (0.85 is 85% full).
compactKeepRecentTokens	integer		The most recent part of the conversation kept word for word when it is summarised, in tokens.
compactTargetTokens	integer		The length the summary aims for, in tokens. compactMaxTokens is the hard limit.
maxToolCallLoops	integer | null		The most rounds of tool calls an agent may make in one turn. After that its tools are withheld and it is told to answer. 0 means no limit; null means 10.
responseMaxTokens	integer | null		The most the model may write in one reply, in tokens, including any reasoning; that much of the context window is kept free for the reply. Null or 0 allows the model’s maximum. A value above the model’s maximum is lowered to it.
compactMaxTokens	integer | null		The most a summary may be, in tokens, when compaction is on. Null: 40,000. A value above the model’s maximum is lowered to it.

{
  "mode": "None",
  "targetUtilization": 0,
  "toolResultTrimmingMode": "Never",
  "toolResultTrimmingMaxCycles": 0,
  "toolResultTrimmingStyle": "Placeholder",
  "toolResultTrimmingReplacement": "...",
  "compactionType": "SameAgent",
  "lowWaterUtilization": 0,
  "compactAtUtilization": 0,
  "compactKeepRecentTokens": 0,
  "compactTargetTokens": 0,
  "maxToolCallLoops": 0,
  "responseMaxTokens": 0,
  "compactMaxTokens": 0
}

In a response

Every field is present in a response; empty ones are null.

Field	Type	Description
mode	ContextTrimmingMode	What happens when the conversation no longer fits the model’s context window.
targetUtilization	number	How full the context window may get, as a fraction from 0 to 1, before the conversation is shortened.
toolResultTrimmingMode	ToolResultTrimmingMode	When the results of earlier tool calls are replaced, to save space.
toolResultTrimmingMaxCycles	integer | null	With toolResultTrimmingMode set to AfterCycles or Both: once an agent has had this many turns, every tool result but the latest is replaced.
toolResultTrimmingStyle	ToolResultTrimmingStyle| null	What a replaced tool result becomes. Null: Placeholder.
effectiveToolResultTrimmingStyle	ToolResultTrimmingStyle	The style actually used: toolResultTrimmingStyle, or Placeholder when that is null. Set by Maitento.
toolResultTrimmingReplacement	string	The text a replaced tool result becomes, with the Placeholder style.
compactionType	CompactionType| null	Whether, and how, older parts of a long conversation are summarised to make room. Suits long, many-turn conversations such as chats. Null: no summaries (None).
effectiveCompactionType	CompactionType	The compaction actually used: compactionType, or None when that is null. Set by Maitento.
lowWaterUtilization	number | null	With compaction on: how far a forced trim shortens the conversation, as a fraction of the context window from 0 to 1, so that it is not trimmed again on the very next turn. Null: 0.2. Never more than targetUtilization.
effectiveLowWaterUtilization	number	The low-water mark actually used: lowWaterUtilization when it is above 0 and at most 1, otherwise 0.2; never more than targetUtilization. Set by Maitento.
compactAtUtilization	number	How full the context window gets, as a fraction from 0 to 1, before the conversation is summarised (0.85 is 85% full).
compactKeepRecentTokens	integer	The most recent part of the conversation kept word for word when it is summarised, in tokens.
compactTargetTokens	integer	The length the summary aims for, in tokens. compactMaxTokens is the hard limit.
maxToolCallLoops	integer | null	The most rounds of tool calls an agent may make in one turn. After that its tools are withheld and it is told to answer. 0 means no limit; null means 10.
effectiveMaxToolCallLoops	integer	The limit actually applied: maxToolCallLoops, or 10 when that is null. 0 means no limit. Set by Maitento.
responseMaxTokens	integer | null	The most the model may write in one reply, in tokens, including any reasoning; that much of the context window is kept free for the reply. Null or 0 allows the model’s maximum. A value above the model’s maximum is lowered to it.
compactMaxTokens	integer | null	The most a summary may be, in tokens, when compaction is on. Null: 40,000. A value above the model’s maximum is lowered to it.

{
  "mode": "None",
  "targetUtilization": 0,
  "toolResultTrimmingMode": "Never",
  "toolResultTrimmingMaxCycles": 0,
  "toolResultTrimmingStyle": "Placeholder",
  "effectiveToolResultTrimmingStyle": "Placeholder",
  "toolResultTrimmingReplacement": "...",
  "compactionType": "SameAgent",
  "effectiveCompactionType": "SameAgent",
  "lowWaterUtilization": 0,
  "effectiveLowWaterUtilization": 0,
  "compactAtUtilization": 0,
  "compactKeepRecentTokens": 0,
  "compactTargetTokens": 0,
  "maxToolCallLoops": 0,
  "effectiveMaxToolCallLoops": 0,
  "responseMaxTokens": 0,
  "compactMaxTokens": 0
}

Where it’s used

• Inside ManagedInteractionCreateRequest
• Inside ManagedInteractionUpdateRequest
• Inside ManagedInteractionVersion
• Inside OneShotInteractionCreateRequest
• Inside OneShotInteractionUpdateRequest
• Inside OneShotInteractionVersion
• Inside RoundRobinInteractionCreateRequest
• Inside RoundRobinInteractionUpdateRequest
• Inside RoundRobinInteractionVersion