Llm

llm

This module defines the llm_router of the Mirumoji API

Attributes:

Name	Type	Description
LOGGER	Logger	Module's logging object
llm_router	APIRouter	The FastAPI router object

breakdown(req) async

Streams the nuance of a focus word within a sentence as Server-Sent Events

The focus word (its stitched token + dictionary data) is emitted once as a focus event, then the LLM explanation streams as data: frames

sys_msg and prompt are independently optional. sys_msg falls back to the default breakdown system message, and prompt to the default template. The template is rendered server-side with the sentence, focus, and context values (see render_breakdown_prompt)

Parameters:

Name	Type	Description	Default
req	BreakdownRequest	The breakdown request	required

Returns:

Type	Description
StreamingResponse	An SSE stream: a focus frame (when a focus is given), the explanation chunks, and a terminal done frame

Raises:

Type	Description
InvalidModelStringError	If the model selector is malformed
LLMProviderUnavailableError	If the requested provider isn't configured

breakdown_preview(req) async

Renders a breakdown prompt template against sample values, no LLM call

Powers the template editor's live preview, so a user sees the assembled prompt (and how a {#context} block appears or vanishes) as they type

Parameters:

Name	Type	Description	Default
req	BreakdownPreviewRequest	The template and sample values	required

Returns:

Type	Description
BreakdownPreviewResponse	The rendered prompt

explain_sentence(req) async

Streams an explanation of a whole sentence as Server-Sent Events

sys_msg and prompt are independently optional and fall back to the default breakdown system message and sentence prompt

Parameters:

Name	Type	Description	Default
req	ExplainSentenceRequest	The explanation request	required

Returns:

Type	Description
StreamingResponse	An SSE stream of explanation chunks ending with a done frame

Raises:

Type	Description
InvalidModelStringError	If the model selector is malformed
LLMProviderUnavailableError	If the requested provider isn't configured
LLMRequestError	If the prompt template is invalid

list_models(provider) async

Lists the available models for a configured LLM provider

Lets the frontend offer a model dropdown for the cloud providers, so users pick a valid model instead of typing one that 404s

Parameters:

Name	Type	Description	Default
provider	str	The provider id (openai, anthropic, gemini, local)	required

Returns:

Type	Description
ModelsResponse	Mapping with a models list of model ids

Raises:

Type	Description
InvalidModelStringError	If the provider id is unknown
LLMProviderUnavailableError	If the provider isn't configured
LLMRequestError	If the provider's models endpoint fails

list_providers() async

Reports which LLM providers are usable in this deployment

Returns:

Type	Description
ProvidersResponse	A providers list of {"provider", "available"} entries, used by the frontend to populate the model picker

stream(req) async

Streams a chat completion as Server-Sent Events

Parameters:

Name	Type	Description	Default
req	ChatRequest	The chat request	required

Returns:

Type	Description
StreamingResponse	An SSE stream of completion chunks ending with a done event

Raises:

Type	Description
InvalidModelStringError	If the model selector is malformed
LLMProviderUnavailableError	If the requested provider isn't configured