Class MistralChatCapabilityImpl

Adapts Mistral chat completions into ProviderPlaneAI's normalized chat response and stream chunk shapes.

Supports text and image inputs, batches streamed text deltas using chatStreamBatchSize, and emits a terminal error chunk for non-abort streaming failures while keeping Mistral SDK event details local to the adapter.

Implements

Index

Constructors

constructor

Methods

chat chatStream

Constructors

constructor

new MistralChatCapabilityImpl(
provider: BaseProvider,
client: Mistral,
): MistralChatCapabilityImpl
Creates a new Mistral chat capability adapter.
Parameters
- provider: BaseProvider
  Owning provider instance used for initialization checks and merged config access.
- client: Mistral
  Initialized official Mistral SDK client.
Returns MistralChatCapabilityImpl
- Defined in src/providers/mistral/capabilities/MistralChatCapabilityImpl.ts:57

Methods

chat

chat(
    request: AIRequest<ClientChatRequest>,
    _ctx?: MultiModalExecutionContext,
    signal?: AbortSignal,
): Promise<AIResponse<NormalizedChatMessage>>
Executes a non-streaming Mistral chat request.

Responsibilities:
- validate request inputs
- resolve merged model/runtime options
- execute chat.complete through the official SDK
- flatten Mistral content into normalized assistant text
- attach provider/model/token metadata for observability
Parameters
- request: AIRequest<ClientChatRequest>
  Unified chat request envelope.
- Optional_ctx: MultiModalExecutionContext
  Optional execution context. Unused directly in this adapter.
- Optionalsignal: AbortSignal
  Optional cancellation signal.
Returns Promise<AIResponse<NormalizedChatMessage>>
Provider-normalized assistant message response.

Throws
When input messages are empty or the request is already aborted.
Implementation of ChatCapability.chat
- Defined in src/providers/mistral/capabilities/MistralChatCapabilityImpl.ts:78

chatStream

chatStream(
    request: AIRequest<ClientChatRequest>,
    _ctx?: MultiModalExecutionContext,
    signal?: AbortSignal,
): AsyncGenerator<AIResponseChunk<NormalizedChatMessage>>
Executes a streaming Mistral chat request.

Responsibilities:
- execute chat.stream through the official SDK
- accumulate provider deltas into larger buffered chunks
- expose both delta and accumulated output in normalized chunk form
- preserve final token usage when Mistral includes it late in the stream
- emit one terminal error chunk on provider/runtime failures
Parameters
- request: AIRequest<ClientChatRequest>
  Unified streaming chat request envelope.
- Optional_ctx: MultiModalExecutionContext
  Optional execution context. Unused directly in this adapter.
- Optionalsignal: AbortSignal
  Optional cancellation signal.
Returns AsyncGenerator<AIResponseChunk<NormalizedChatMessage>>
Async stream of normalized chat chunks.

Throws
When input messages are empty.
Implementation of ChatStreamCapability.chatStream
- Defined in src/providers/mistral/capabilities/MistralChatCapabilityImpl.ts:142

Class MistralChatCapabilityImpl

Implements

Index

Constructors

Methods

Constructors

constructor

Parameters

Returns MistralChatCapabilityImpl

Methods

chat

Parameters

Returns Promise<AIResponse<NormalizedChatMessage>>

Throws

chatStream

Parameters

Returns AsyncGenerator<AIResponseChunk<NormalizedChatMessage>>

Throws

Settings

On This Page