Contents
-
Terminally Deprecated Elements
since 0.36.0, for removal in 1.0.0. A public enum of model constants goes
stale the day Anthropic ships a model it does not list, and removing a
retired constant is itself a breaking change — so offering one at all works
against ever reaching a release with no breaking entry. Name a model by its
string id instead; every request has honoured one since 0.35.0, and
AnthropicProvider.getDefaultModel() still names a sensible default when none is given.
If the model needs a context window other than the fallback, register one
directly with
ModelTokenLimits.register(java.lang.String, int).
since 0.36.0, for removal in 1.0.0. A public enum of model constants goes
stale the day a new model lands in Ollama's library, and removing a retired
constant is itself a breaking change — so offering one at all works against
ever reaching a release with no breaking entry. Name a model by its string id
instead — any tag combination, not only the ones listed here — and
OllamaProvider.getDefaultModel() still names a sensible default when none is given.
If the model needs a context window other than the fallback, register one
directly with
ModelTokenLimits.register(java.lang.String, int).
since 0.36.0, for removal in 1.0.0. A public enum of model constants goes
stale the day OpenAI ships a model it does not list, and removing a retired
constant is itself a breaking change — so offering one at all works against
ever reaching a release with no breaking entry. Name a model by its string id
instead; every request has honoured one since 0.35.0, and
OpenAIProvider.getDefaultModel() still names a sensible default when none is given.
If the model needs a context window other than the fallback, register one
directly with
ModelTokenLimits.register(java.lang.String, int).
since 0.36.0, for removal in 1.0.0. A result with no call id cannot be paired
to the call it answers — Anthropic rejects the request outright, OpenAI
accepts it as an ordinary user message instead of a tool result, and two
parallel calls to the same function are ambiguous by construction. Use
LLMMessage.function(FunctionCall, String), which every caller that still has the
originating
FunctionCall in scope can call instead.
-
Deprecated Enum Classes
since 0.36.0, for removal in 1.0.0. A public enum of model constants goes
stale the day Anthropic ships a model it does not list, and removing a
retired constant is itself a breaking change — so offering one at all works
against ever reaching a release with no breaking entry. Name a model by its
string id instead; every request has honoured one since 0.35.0, and
AnthropicProvider.getDefaultModel() still names a sensible default when none is given.
If the model needs a context window other than the fallback, register one
directly with
ModelTokenLimits.register(java.lang.String, int).
since 0.36.0, for removal in 1.0.0. A public enum of model constants goes
stale the day a new model lands in Ollama's library, and removing a retired
constant is itself a breaking change — so offering one at all works against
ever reaching a release with no breaking entry. Name a model by its string id
instead — any tag combination, not only the ones listed here — and
OllamaProvider.getDefaultModel() still names a sensible default when none is given.
If the model needs a context window other than the fallback, register one
directly with
ModelTokenLimits.register(java.lang.String, int).
since 0.36.0, for removal in 1.0.0. A public enum of model constants goes
stale the day OpenAI ships a model it does not list, and removing a retired
constant is itself a breaking change — so offering one at all works against
ever reaching a release with no breaking entry. Name a model by its string id
instead; every request has honoured one since 0.35.0, and
OpenAIProvider.getDefaultModel() still names a sensible default when none is given.
If the model needs a context window other than the fallback, register one
directly with
ModelTokenLimits.register(java.lang.String, int).
-
Deprecated Methods
since 0.36.0, for removal in 1.0.0. A result with no call id cannot be paired
to the call it answers — Anthropic rejects the request outright, OpenAI
accepts it as an ordinary user message instead of a tool result, and two
parallel calls to the same function are ambiguous by construction. Use
LLMMessage.function(FunctionCall, String), which every caller that still has the
originating
FunctionCall in scope can call instead.