Functions

aggregate_completion

def aggregate_completion(raw: Iterator[tuple[int, dict[int, float]]], tokenizer: TokenizerAdapter, request: CompletionRequest, prompt_ids: Sequence[int], eos_token_ids: Sequence[int] | None=None) -> Iterator[CompletionStreamChunk]

Map raw source positions into the former word-level completion stream.

Source: src/invera_sdk/completions.py:431

create_client_with_tokenizer

def create_client_with_tokenizer(options: InveraClientOptions, tokenizer: TokenizerAdapter) -> InveraClient

Create a high-level client with an advanced host-managed tokenizer.

Source: src/invera_sdk/completions.py:389

create_hugging_face_tokenizer

def create_hugging_face_tokenizer(profile: ModelProfile) -> TokenizerAdapter

Create the default Transformers-backed tokenizer for a release-pinned profile.

Source: src/invera_sdk/completions.py:180

decode_raw_attribution_chunk

def decode_raw_attribution_chunk(payload: Any) -> tuple[int, dict[int, float]]

Decode one raw NDJSON object into a token id and summed source-position scores.

Source: src/invera_sdk/completions.py:398

parse_qwen_tool_calls

def parse_qwen_tool_calls(chunks: Iterator[CompletionStreamChunk]) -> Iterator[CompletionStreamChunk]

Parse Qwen XML tool-call markers into typed compatibility chunks.

Source: src/invera_sdk/completions.py:467

stream_evidence

def stream_evidence(chunks: Iterator[CompletionStreamChunk], request: CompletionRequest, options: EvidenceOptions | None=None) -> Iterator[EvidenceStreamEvent]

Add claim and tool-call evidence to an attributed completion stream.

Source: src/invera_sdk/evidence.py:229