Module llm 

Fluent LLM-mocking builders for the MockServer Rust client.

This module is an idiomatic Rust port of the Java/Node/Python client LLM builders (org.mockserver.client.Llm / LlmMockBuilder / LlmConversationBuilder / LlmFailoverBuilder / TurnBuilder) and the underlying model classes (Completion, ToolUse, Usage, StreamingPhysics, EmbeddingResponse). The builders produce exactly the same expectation wire JSON the other clients emit, so a mock scripted from Rust is byte-for-byte equivalent (modulo JSON key order, which the server ignores) to one scripted from Java.

The expectation action is carried in the httpLlmResponse field of an expectation (a sibling of httpRequest, scenarioName, scenarioState, newScenarioState, times, timeToLive, httpResponse). Null/unset fields are omitted from the JSON, matching the server’s NON_NULL serialization.

Example

use mockserver_client::llm::{llm_mock, completion, Provider};

let expectation = llm_mock("/v1/chat/completions")
    .with_provider(Provider::OPENAI)
    .with_model("gpt-4o")
    .responding_with(completion().with_text("Hello!"))
    .build();

assert_eq!(expectation["httpLlmResponse"]["provider"], "OPENAI");

Structs

Completion

A mocked LLM chat/completion response, provider-agnostic (mirrors org.mockserver.model.Completion).

EmbeddingResponse

A mocked embedding response — vector shape and determinism (mirrors org.mockserver.model.EmbeddingResponse).

IsolationSource

Where to read the per-session isolation key from an inbound request (mirrors org.mockserver.llm.IsolationSource). Encodes as "<kind>:<name>".

LlmConversationBuilder

Builder for multi-turn LLM conversation mocks with scenario advancement (mirrors org.mockserver.client.LlmConversationBuilder).

LlmFailoverBuilder

Builder for provider failover/retry scenarios (mirrors org.mockserver.client.LlmFailoverBuilder).

LlmMockBuilder

Fluent builder for a single LLM mock expectation (mirrors org.mockserver.client.LlmMockBuilder).

Provider

LLM provider names. Serialized on the wire as the upper-case enum name (mirrors org.mockserver.model.Provider).

Role

Parsed-message roles (mirrors org.mockserver.llm.ParsedMessage.Role).

StreamingPhysics

Controls the timing physics of a streamed (SSE) completion (mirrors org.mockserver.model.StreamingPhysics).

ToolUse

A single tool/function call emitted by the assistant (mirrors org.mockserver.model.ToolUse).

TurnBuilder

Sub-builder configuring one turn of a conversation mock (mirrors org.mockserver.client.TurnBuilder).

Usage

Token usage counts for a completion (mirrors org.mockserver.model.Usage).

Enums

LlmResponseBody

Either a Completion or an EmbeddingResponse, accepted by responding_with builder methods.

Functions

completion

Factory mirroring Completion.completion().

conversation

Entry point mirroring LlmConversationBuilder.conversation().

cookie

Isolate conversations by a cookie.

default_error_body

The default error body for the given status code (mirrors the Java/Node/Python defaultErrorBody). Exposed for parity/testing.

embedding

Factory mirroring EmbeddingResponse.embedding().

header

Isolate conversations by a request header.

input_tokens

Factory mirroring Usage.inputTokens(n).

jitter

Factory mirroring StreamingPhysics.jitter(j).

llm_failover

Entry point mirroring LlmFailoverBuilder.llmFailover().

llm_mock

Entry point mirroring LlmMockBuilder.llmMock(path).

output_tokens

Factory mirroring Usage.outputTokens(n).

query_parameter

Isolate conversations by a query parameter.

streaming_physics

Factory mirroring StreamingPhysics.streamingPhysics().

tokens_per_second

Factory mirroring StreamingPhysics.tokensPerSecond(n).

tool_use

Factory mirroring ToolUse.toolUse(name).

turn

Entry point mirroring TurnBuilder construction (used with LlmConversationBuilder::turn).

usage

Factory mirroring Usage.usage().