Struct Capabilities 

pub struct Capabilities {

    pub native_tools: bool,
    pub message_wire_format: WireDialect,
    pub native_tool_wire_format: String,
    pub defer_loading: bool,
    pub tool_search: Vec<String>,
    pub responses_api: bool,
    pub hosted_tools: Vec<String>,
    pub remote_mcp: bool,
    pub conversation_state: bool,
    pub compaction: bool,
    pub background_mode: bool,
    pub batch_api: bool,
    pub batch_wire_format: Option<String>,
    pub batch_input_mode: Option<String>,
    pub batch_discount_percent: Option<u32>,
    pub batch_turnaround_hours: Option<u32>,
    pub tool_approval_policy: Option<String>,
    pub max_tools: Option<u32>,
    pub prompt_caching: bool,
    pub cache_breakpoint_style: String,
    pub vision: bool,
    pub audio: bool,
    pub pdf: bool,
    pub video: bool,
    pub files_api_supported: bool,
    pub file_upload_wire_format: Option<String>,
    pub structured_output: Option<String>,
    pub json_schema: Option<String>,
    pub prefers_xml_scaffolding: bool,
    pub reserved_tool_call_token: bool,
    pub prefers_markdown_scaffolding: bool,
    pub structured_output_mode: String,
    pub supports_assistant_prefill: bool,
    pub prefers_role_developer: bool,
    pub prefers_xml_tools: bool,
    pub thinking_block_style: String,
    pub emits_inline_reasoning: bool,
    pub thinking_modes: Vec<String>,
    pub interleaved_thinking_supported: bool,
    pub anthropic_beta_features: Vec<String>,
    pub vision_supported: bool,
    pub image_url_input_supported: bool,
    pub preserve_thinking: bool,
    pub server_parser: String,
    pub honors_chat_template_kwargs: bool,
    pub chat_template_options_field: Option<String>,
    pub requires_completion_tokens: bool,
    pub chat_completions_unsupported: bool,
    pub requires_streaming: bool,
    pub reasoning_effort_supported: bool,
    pub reasoning_effort_levels: Vec<String>,
    pub reasoning_none_supported: bool,
    pub max_thinking_budget: Option<i64>,
    pub reasoning_disable_supported: bool,
    pub reasoning_required_for_tools: bool,
    pub reasoning_text_promotable: bool,
    pub reasoning_wire_format: Option<String>,
    pub seed_supported: bool,
    pub top_k_supported: bool,
    pub temperature_supported: bool,
    pub top_p_supported: bool,
    pub frequency_penalty_supported: bool,
    pub presence_penalty_supported: bool,
    pub allowed_tool_choice_modes: Vec<String>,
    pub requires_tool_result_adjacency: bool,
    pub supports_parallel_tool_calls: bool,
    pub tools_exclude_response_format: bool,
    pub recommended_endpoint: Option<String>,
    pub text_tool_wire_format_supported: bool,
    pub preferred_tool_format: Option<String>,
    pub tool_mode_parity: Option<String>,
    pub tool_mode_parity_notes: Option<String>,
    pub thinking_disable_directive: Option<String>,
    pub auto_reasoning_overrides: BTreeMap<String, String>,
    pub provider_route_denylist: Vec<String>,
    pub openrouter_provider_order: Vec<String>,
    pub serving_precision: String,
    pub computer_use_style: Option<ComputerUseStyle>,
    pub screenshot_scaling: Option<ScreenshotScaling>,
    pub safety_ack_flow: bool,
}

Resolved capabilities for a (provider, model) pair. Unset rule fields resolve to false / empty / None so callers never have to unwrap an Option<bool> for what are really boolean gates.

Fields

native_tools: bool

message_wire_format: WireDialect

native_tool_wire_format: String

defer_loading: bool

tool_search: Vec<String>

responses_api: bool

hosted_tools: Vec<String>

remote_mcp: bool

conversation_state: bool

compaction: bool

background_mode: bool

batch_api: bool

batch_wire_format: Option<String>

batch_input_mode: Option<String>

batch_discount_percent: Option<u32>

batch_turnaround_hours: Option<u32>

tool_approval_policy: Option<String>

max_tools: Option<u32>

prompt_caching: bool

cache_breakpoint_style: String

vision: bool

audio: bool

pdf: bool

video: bool

files_api_supported: bool

file_upload_wire_format: Option<String>

structured_output: Option<String>

json_schema: Option<String>

Legacy mirror for CLI display and older callers.

prefers_xml_scaffolding: bool

reserved_tool_call_token: bool

See ProviderRule::reserved_tool_call_token.

prefers_markdown_scaffolding: bool

structured_output_mode: String

supports_assistant_prefill: bool

prefers_role_developer: bool

prefers_xml_tools: bool

thinking_block_style: String

emits_inline_reasoning: bool

Whether this route emits its reasoning INLINE in the text channel as <think>...</think> blocks (local Ollama/llama.cpp reasoning models, Qwen3 via vLLM, Kimi) rather than in a separate provider reasoning field. When true, the llm_call envelope builder splits those blocks out of text/prose/visible_text and folds them into the reasoning channel, mirroring how hosted providers surface a dedicated thinking field. Derived from thinking_block_style == "inline" — the same population that represents reasoning as inline <think> in prompt context is the one that emits it that way in responses.

thinking_modes: Vec<String>

interleaved_thinking_supported: bool

anthropic_beta_features: Vec<String>

vision_supported: bool

image_url_input_supported: bool

preserve_thinking: bool

server_parser: String

honors_chat_template_kwargs: bool

chat_template_options_field: Option<String>

requires_completion_tokens: bool

chat_completions_unsupported: bool

True when the route is served ONLY by the provider Responses API and rejects /v1/chat/completions (OpenAI *-codex models). Harn routes such calls through the Responses provider automatically.

requires_streaming: bool

reasoning_effort_supported: bool

reasoning_effort_levels: Vec<String>

reasoning_none_supported: bool

max_thinking_budget: Option<i64>

See ProviderRule::max_thinking_budget. None means the model uses the provider’s own default ceiling.

reasoning_disable_supported: bool

reasoning_required_for_tools: bool

See ProviderRule::reasoning_required_for_tools.

reasoning_text_promotable: bool

reasoning_wire_format: Option<String>

seed_supported: bool

top_k_supported: bool

temperature_supported: bool

top_p_supported: bool

frequency_penalty_supported: bool

presence_penalty_supported: bool

allowed_tool_choice_modes: Vec<String>

requires_tool_result_adjacency: bool

supports_parallel_tool_calls: bool

tools_exclude_response_format: bool

recommended_endpoint: Option<String>

text_tool_wire_format_supported: bool

preferred_tool_format: Option<String>

tool_mode_parity: Option<String>

tool_mode_parity_notes: Option<String>

thinking_disable_directive: Option<String>

auto_reasoning_overrides: BTreeMap<String, String>

Per-task auto-policy reasoning-level overrides for this route. See ProviderRule::auto_reasoning_overrides.

provider_route_denylist: Vec<String>

OpenRouter upstream provider names to exclude from routing for this row. See ProviderRule::provider_route_denylist. Empty means “no route restriction”.

openrouter_provider_order: Vec<String>

OpenRouter upstream provider names this row is PINNED to (allowlist), in preference order. See ProviderRule::openrouter_provider_order. Empty means “no pin” (free OpenRouter routing).

serving_precision: String

Serving-quality / precision trust verdict for this route. See ProviderRule::serving_precision. "unverified" when unset.

computer_use_style: Option<ComputerUseStyle>

How the neutral computer tool projects onto this route’s native computer-use surface. None means the route exposes no computer-use surface. See ComputerUseStyle.

screenshot_scaling: Option<ScreenshotScaling>

Screenshot downscaling policy applied before the image reaches the model. None means unset. See ScreenshotScaling.

safety_ack_flow: bool

Whether this route requires echoing acknowledged safety checks on the computer-use follow-up turn (OpenAI Responses pending_safety_checks → acknowledged_safety_checks). See ProviderRule::safety_ack_flow.

Trait Implementations

impl Clone for Capabilities

fn clone(&self) -> Capabilities

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Capabilities

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Capabilities

fn default() -> Self

Returns the “default value” for a type.

impl Eq for Capabilities

impl PartialEq for Capabilities

fn eq(&self, other: &Capabilities) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for Capabilities