Requests

requests

Defines Pydantic request models for the API's endpoints

Model selection

• LLM requests have a model selector in provider:model form (e.g. "openai:gpt-4.1-mini")

LLM Request Defaults

• sys_msg / prompt are optional

• When omitted the server falls back to its default system message and prompt

SubmitBatchRequest = Annotated[BatchGenerateSrtJobRequest | BatchTranscribeJobRequest | BatchConvertJobRequest | BatchFixSrtJobRequest, Field(discriminator='type')] module-attribute

The body of a POST /jobs/batch request

A discriminated union keyed by the parent job's type attribute, so that each batch operation's shared opts are validated against their own model. The parent fans the shared options out to one child job per file_id

SubmitJobRequest = Annotated[GenerateSrtJobRequest | TranscribeJobRequest | ConvertJobRequest | FixSrtJobRequest, Field(discriminator='type')] module-attribute

The body of a POST /jobs request

A discriminated union keyed by a persisted job's type attribute, so that each operation's opts are validated against their own model

BatchConvertJobRequest

Bases: BatchSubmitBase

Submission for a batch_convert job

Parameters:

Name	Type	Description	Default
type	Literal['batch_convert']	The operation discriminator	required
opts	ConvertVideoRequest	Conversion options shared by every file in the batch	required

child_params(file_id)

Builds one child job's params, matching the single-op shape

Parameters:

Name	Type	Description	Default
file_id	str	The profile file this child operates on	required

Returns:

Type	Description
dict[str, Any]	The file reference and shared conversion options

child_type()

The single-op type stored on each per-file child job

Returns:

Type	Description
str	The discriminator the child shares with its single-op handler

to_params()

Builds the persisted parent job's params

Returns:

Type	Description
dict[str, Any]	The file references and shared conversion options

BatchFixSrtJobRequest

Bases: BatchSubmitBase

Submission for a batch_fix_srt job

Parameters:

Name	Type	Description	Default
type	Literal['batch_fix_srt']	The operation discriminator	required
model	str	LLM model selector in provider:model form	required
sys_msg	str | None	Optional LLM system message shared by every file	required

child_params(file_id)

Builds one child job's params, matching the single-op shape

Parameters:

Name	Type	Description	Default
file_id	str	The profile file this child operates on	required

Returns:

Type	Description
dict[str, Any]	The file reference, model selector, and optional system message

child_type()

The single-op type stored on each per-file child job

Returns:

Type	Description
str	The discriminator the child shares with its single-op handler

to_params()

Builds the persisted parent job's params

Returns:

Type	Description
dict[str, Any]	The file references, model selector, and optional system message

BatchGenerateSrtJobRequest

Bases: BatchSubmitBase

Submission for a batch_generate_srt job

Parameters:

Name	Type	Description	Default
type	Literal['batch_generate_srt']	The operation discriminator	required
opts	GenerateSrtRequest	Curated transcription options shared by every file in the batch	required

child_params(file_id)

Builds one child job's params, matching the single-op shape

Parameters:

Name	Type	Description	Default
file_id	str	The profile file this child operates on	required

Returns:

Type	Description
dict[str, Any]	The file reference and shared transcription options

child_type()

The single-op type stored on each per-file child job

Returns:

Type	Description
str	The discriminator the child shares with its single-op handler

to_params()

Builds the persisted parent job's params

Returns:

Type	Description
dict[str, Any]	The file references and shared transcription options

BatchSubmitBase

Bases: BaseModel

Shared fields for every batch job submission request

Parameters:

Name	Type	Description	Default
file_ids	list[str]	The profile files the batch operates on, one child job per file (at least one)	required

BatchTranscribeJobRequest

Bases: BatchSubmitBase

Submission for a batch_transcribe job

Parameters:

Name	Type	Description	Default
type	Literal['batch_transcribe']	The operation discriminator	required
opts	TranscribeAudioRequest	clean_audio + curated transcription options shared by every file in the batch	required

child_params(file_id)

Builds one child job's params, matching the single-op shape

Parameters:

Name	Type	Description	Default
file_id	str	The profile file this child operates on	required

Returns:

Type	Description
dict[str, Any]	The file reference and shared transcription options

child_type()

The single-op type stored on each per-file child job

Returns:

Type	Description
str	The discriminator the child shares with its single-op handler

to_params()

Builds the persisted parent job's params

Returns:

Type	Description
dict[str, Any]	The file references and shared transcription options

BreakdownRequest

Bases: BaseModel

Request for the /llm/breakdown endpoint

Parameters:

Name	Type	Description	Default
sentence	str	The sentence containing the focus word	required
focus	str | None	The word to explain in context	required
model	str	Model selector in provider:model form	required
sys_msg	str | None	Optional custom system message	required
prompt	str | None	Optional custom prompt template ({0} = sentence, {1} = focus)	required

ChatRequest

Bases: BaseModel

Request for the streaming /llm/stream chat endpoint

Parameters:

Name	Type	Description	Default
prompt	str	The user prompt	required
model	str	Model selector in provider:model form	required
system_message	str | None	Optional system message	required

ConvertJobRequest

Bases: SubmitJobBase

Submission for a convert job

Parameters:

Name	Type	Description	Default
type	Literal['convert']	The operation discriminator	required
opts	ConvertVideoRequest	Conversion options	required

to_params()

Builds the persisted job's stored params

Returns:

Type	Description
dict[str, Any]	The file reference and normalised conversion options

ConvertVideoRequest

Bases: BaseModel

Options for /video/convert_to_mp4

Parameters:

Name	Type	Description	Default
resolution	str	Maximum output size WxH (aspect preserved)	required
target_bitrate	str	Video bitrate ceiling	required
preset	Literal	Encoder speed / quality preset, one of performance, balanced, quality	required

ExplainSentenceRequest

Bases: BaseModel

Request for the /llm/explain_sentence endpoint

Parameters:

Name	Type	Description	Default
sentence	str	The sentence to explain	required
model	str	Model selector in provider:model form	required
sys_msg	str | None	Optional custom system message	required
prompt	str | None	Optional custom prompt template ({0} = sentence)	required

FixSrtJobRequest

Bases: SubmitJobBase

Submission for a fix_srt job

Parameters:

Name	Type	Description	Default
type	Literal	The operation discriminator	required
model	str	LLM model selector in provider:model form	required
sys_msg	str | None	Optional LLM system message	required

to_params()

Builds the persisted job's params attribute

Returns:

Type	Description
dict[str, Any]	The file reference, model selector, and optional system message

GenerateSrtJobRequest

Bases: SubmitJobBase

Submission for a generate_srt job

Parameters:

Name	Type	Description	Default
type	Literal['generate_srt']	The operation discriminator	required
opts	GenerateSrtRequest	Curated transcription options	required

to_params()

Builds the persisted job's params attribute

Returns:

Type	Description
dict[str, Any]	The file reference and normalised transcription options

GenerateSrtRequest

Bases: TranscribeOptions

Options for /video/generate_srt

Inherits the curated transcription options unchanged

LlmTemplateRequest

Bases: BaseModel

Request for upserting a profile's LLM template

Parameters:

Name	Type	Description	Default
sys_msg	str	System message	required
prompt	str	Prompt template	required
model	str	Model selector in provider:model form	required
srt_sys_msg	str	Subtitle-fix system message (empty = server default)	required
srt_model	str	Subtitle-fix model override (empty = use model)	required

SaveSubtitlesRequest

Bases: BaseModel

Request for persisting SRT content to the active profile (POST /profiles/subtitles)

Replace vs Create

• When file_id references an existing SRT file owned by the profile, that file's content is overwritten in place (same id / path)

• Otherwise a new SRT file is created under the profile

Parameters:

Name	Type	Description	Default
content	str	The SRT content to store	required
file_id	str | None	Existing SRT file id to overwrite, if any	required
name	str | None	Optional file name for a newly created file	required

SubmitJobBase

Bases: BaseModel

Shared fields for every job submission request

Parameters:

Name	Type	Description	Default
file_id	str	The profile file that the job operates on	required

TokenizeBatchRequest

Bases: BaseModel

Request for the batch tokenize endpoint (POST /dict/tokenize)

Parameters:

Name	Type	Description	Default
sentences	list[str]	The sentences to tokenize, in order	required
mode	BundleMode	How aggressively to group tokens into words	required

TranscribeAudioRequest

Bases: TranscribeOptions

Options for /audio/transcribe

Parameters:

Name	Type	Description	Default
clean_audio	bool	Apply a band-pass + loudness filter before transcription	required

TranscribeJobRequest

Bases: SubmitJobBase

Submission for a transcribe job

Parameters:

Name	Type	Description	Default
type	Literal['transcribe']	The operation discriminator	required
opts	TranscribeAudioRequest	clean_audio + curated transcription options	required

to_params()

Builds the persisted job's params attribute

Returns:

Type	Description
dict[str, Any]	The file reference and normalised transcription options

TranscribeOptions

Bases: BaseModel

Curated, query-param-friendly transcription options for faster_whisper

Curation

• Only scalar faster_whisper transcribe options that tune quality and filtering are exposed

• Options that would change the segment structure (.start / .end / .text) or the assumption that media is transcribed (not translated) in the chosen language are deliberately omitted (e.g. task, without_timestamps)

• Unset (None) fields fall back to mirumoji.server.processing.whisper.DEFAULT_TRANSCRIBE_OPTS

Parameters:

Name	Type	Description	Default
language	str	Language code to transcribe in	required
beam_size	int | None	Beam size for decoding	required
best_of	int | None	Number of candidates when sampling	required
patience	float | None	Beam-search patience factor	required
length_penalty	float | None	Exponential length penalty	required
temperature	float | None	Sampling temperature	required
compression_ratio_threshold	float | None	Gzip compression-ratio threshold above which output is treated as failed	required
log_prob_threshold	float | None	Average log-probability threshold below which output is treated as failed	required
no_speech_threshold	float | None	No-speech probability threshold	required
condition_on_previous_text	bool | None	Feed prior text as context	required
initial_prompt	str | None	Optional text to bias the first window	required
vad_filter	bool | None	Apply voice-activity-detection filtering	required