Attributes
- Companion
- class
- Graph
-
- Supertypes
-
class Objecttrait Matchableclass Any
- Self type
Members list
Type members
Classlikes
How a request authenticates.
How a request authenticates.
Attributes
- Supertypes
-
trait Enumtrait Serializabletrait Producttrait Equalsclass Objecttrait Matchableclass AnyShow all
Where an OpenAICompatibleClient sends requests, and how it describes itself.
Where an OpenAICompatibleClient sends requests, and how it describes itself.
Value parameters
- baseUrl
-
API base URL; requests go to
<baseUrl>/chat/completions. - contextWindow
-
the model's total token capacity.
- credential
-
how requests authenticate; Credential.Anonymous sends no
Authorizationheader. - displayName
-
human-readable name used in log lines and the "already closed" error, e.g.
"DeepSeek". - model
-
model identifier sent in every request.
- providerName
-
metrics and exchange-log label, and the provider named in mapped HTTP errors, e.g.
"deepseek". - reserveCompletion
-
tokens held back from the prompt for the reply.
Attributes
- Supertypes
-
trait Serializabletrait Producttrait Equalsclass Objecttrait Matchableclass AnyShow all
The tool calls seen so far in one stream, by their index.
The tool calls seen so far in one stream, by their index.
A streamed tool call is split across deltas: the first carries its id, name and the start of its arguments, and each continuation carries only its index and the next argument fragment. Several calls can be interleaved, told apart by index. Continuations are given the id and name their index was first seen with, so the chunks a caller receives
- and the ones
StreamingAccumulator, which keys calls by id and skips a chunk with none, accumulates - all name their call. (Before #1132 each client defaulted a missing id to"", and every continuation's arguments were dropped.)
Attributes
- Supertypes
-
class Objecttrait Matchableclass Any
Value members
Concrete methods
A client for a generic OpenAI-compatible endpoint: the standard dialect, plus config.headers on every request, asking for streamed usage unless config.streamUsage is off.
A client for a generic OpenAI-compatible endpoint: the standard dialect, plus config.headers on every request, asking for streamed usage unless config.streamUsage is off.
Attributes
The settings the generic openai-compatible provider derives from its config.
The settings the generic openai-compatible provider derives from its config.
Attributes
Concrete fields
The timeout on complete's request: two minutes, what the old MistralClient and CohereClient used, and what OllamaClient, GeminiClient and VertexAIClient use. This is the default: a section's timeouts.request replaces it (#712).
The timeout on complete's request: two minutes, what the old MistralClient and CohereClient used, and what OllamaClient, GeminiClient and VertexAIClient use. This is the default: a section's timeouts.request replaces it (#712).
Attributes
The timeout on streamComplete's request: five minutes, as in the clients this one replaced. A section's timeouts.stream replaces it.
The timeout on streamComplete's request: five minutes, as in the clients this one replaced. A section's timeouts.stream replaces it.