fix: Avoid BadRequestError due to invalid max_tokens (#3667)

This patch ensures if max tokens is not defined, then is set to None
instead of 0 when calling openai_chat_completion. This way some
providers (like gemini) that cannot handle the `max_tokens = 0` will not
fail

Issue: #3666

Author: luis5tb

client-sdks/stainless/openapi.yml

@@ -11600,7 +11600,6 @@ components:
           description: The sampling strategy.
         max_tokens:
           type: integer
-          default: 0
           description: >-
             The maximum number of tokens that can be generated in the completion.
             The token count of your prompt plus max_tokens cannot exceed the model's

docs/static/deprecated-llama-stack-spec.html

@@ -3901,7 +3901,6 @@
                     },
                     "max_tokens": {
                         "type": "integer",
-                        "default": 0,
                         "description": "The maximum number of tokens that can be generated in the completion. The token count of your prompt plus max_tokens cannot exceed the model's context length."
                     },
                     "repetition_penalty": {

docs/static/deprecated-llama-stack-spec.yaml

@@ -2862,7 +2862,6 @@ components:
           description: The sampling strategy.
         max_tokens:
           type: integer
-          default: 0
           description: >-
             The maximum number of tokens that can be generated in the completion.
             The token count of your prompt plus max_tokens cannot exceed the model's

docs/static/experimental-llama-stack-spec.html

@@ -2376,7 +2376,6 @@
                     },
                     "max_tokens": {
                         "type": "integer",
-                        "default": 0,
                         "description": "The maximum number of tokens that can be generated in the completion. The token count of your prompt plus max_tokens cannot exceed the model's context length."
                     },
                     "repetition_penalty": {

docs/static/experimental-llama-stack-spec.yaml

@@ -1695,7 +1695,6 @@ components:
           description: The sampling strategy.
         max_tokens:
           type: integer
-          default: 0
           description: >-
             The maximum number of tokens that can be generated in the completion.
             The token count of your prompt plus max_tokens cannot exceed the model's

docs/static/stainless-llama-stack-spec.html

@@ -15452,7 +15452,6 @@
                     },
                     "max_tokens": {
                         "type": "integer",
-                        "default": 0,
                         "description": "The maximum number of tokens that can be generated in the completion. The token count of your prompt plus max_tokens cannot exceed the model's context length."
                     },
                     "repetition_penalty": {

docs/static/stainless-llama-stack-spec.yaml

@@ -11600,7 +11600,6 @@ components:
           description: The sampling strategy.
         max_tokens:
           type: integer
-          default: 0
           description: >-
             The maximum number of tokens that can be generated in the completion.
             The token count of your prompt plus max_tokens cannot exceed the model's

llama_stack/apis/inference/inference.py

@@ -97,7 +97,7 @@ class SamplingParams(BaseModel):
 
     strategy: SamplingStrategy = Field(default_factory=GreedySamplingStrategy)
 
-    max_tokens: int | None = 0
+    max_tokens: int | None = None
     repetition_penalty: float | None = 1.0
     stop: list[str] | None = None