Skip to main content

Interface: ModelTargetInfo

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:149

Information about the target AI model that affects prompt construction. This guides template selection, token limits, and capability-specific formatting. ModelTargetInfo

Properties

audioSupport?

optional audioSupport: object

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:187

Audio input support configuration (more likely for pre-processing than direct model input).

requiresTranscription?

optional requiresTranscription: boolean

supported

supported: boolean


capabilities

capabilities: string[]

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:162

A list of functional capabilities of the model (e.g., 'tool_use', 'vision_input', 'json_mode').


maxContextTokens

maxContextTokens: number

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:155

Maximum context length in tokens supported by the model.


modelId

modelId: string

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:151

Unique identifier of the target model (e.g., "gpt-4o", "ollama/llama3").


optimalContextTokens?

optional optimalContextTokens: number

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:160

Optional: Optimal context length for best performance/cost-efficiency, if different from max. Prompts might be targeted to this length.


optimizationHints?

optional optimizationHints: object

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:192

Model-specific hints for optimizing prompt construction or token budgeting.

optimalHistoryMessages?

optional optimalHistoryMessages: number

preferredSystemPromptLength?

optional preferredSystemPromptLength: number

tokenBudgetingStrategy?

optional tokenBudgetingStrategy: "aggressive_truncate" | "summarize_old" | "balanced"


promptFormatType

promptFormatType: string

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:170

The type of prompt format the model expects. 'openai_chat': Standard OpenAI chat completion format (array of messages). 'anthropic_messages': Anthropic Messages API format (messages array + optional system prompt). 'generic_completion': A single string prompt for older completion-style models. 'custom': A custom format handled by a specific template.


providerId

providerId: string

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:153

Identifier of the provider hosting the model (e.g., "openai", "ollama").


toolSupport

toolSupport: object

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:172

Configuration for tool/function calling support.

format

format: string

Format expected by the model for tool definitions and calls.

maxToolsPerCall?

optional maxToolsPerCall: number

Maximum number of tools that can be defined or called in a single interaction.

supported

supported: boolean


visionSupport?

optional visionSupport: object

Defined in: packages/agentos/src/core/llm/IPromptEngine.ts:180

Vision input support configuration.

maxImageResolution?

optional maxImageResolution: string

maxImages?

optional maxImages: number

supported

supported: boolean

supportedFormats?

optional supportedFormats: string[]