lionagi 0.12.2__py3-none-any.whl → 0.12.3__py3-none-any.whl

lionagi/service/providers/perplexity_/models.py

@@ -1,5 +1,144 @@
-from khive.providers.perplexity_ import (
-    PerplexityChatRequest as PerplexityChatCompletionRequest,
-)
+from enum import Enum
+from typing import Any
 
-__all__ = ("PerplexityChatCompletionRequest",)
+from pydantic import BaseModel, Field, model_validator
+
+
+class PerplexityRole(str, Enum):
+    """Roles allowed in Perplexity's messages."""
+
+    system = "system"
+    user = "user"
+    assistant = "assistant"
+
+
+class PerplexityMessage(BaseModel):
+    """
+    A single message in the conversation.
+    `role` can be 'system', 'user', or 'assistant'.
+    `content` is the text for that conversation turn.
+    """
+
+    role: PerplexityRole = Field(
+        ...,
+        description="The role of the speaker. Must be system, user, or assistant.",
+    )
+    content: str = Field(..., description="The text content of this message.")
+
+
+class PerplexityChatCompletionRequest(BaseModel):
+    """
+    Represents the request body for Perplexity's Chat Completions endpoint.
+    Endpoint: POST https://api.perplexity.ai/chat/completions
+    """
+
+    model: str = Field(
+        "sonar",
+        description="The model name, e.g. 'sonar', (the only model available at the time when this request model was updated, check doc for latest info).",
+    )
+    messages: list[PerplexityMessage] = Field(
+        ..., description="A list of messages forming the conversation so far."
+    )
+
+    # Optional parameters
+    frequency_penalty: float | None = Field(
+        default=None,
+        gt=0,
+        description=(
+            "Multiplicative penalty > 0. Values > 1.0 penalize repeated tokens more strongly. "
+            "Value=1.0 means no penalty. Incompatible with presence_penalty."
+        ),
+    )
+    presence_penalty: float | None = Field(
+        default=None,
+        ge=-2.0,
+        le=2.0,
+        description=(
+            "Penalizes tokens that have appeared so far (range -2 to 2). "
+            "Positive values encourage talking about new topics. Incompatible with frequency_penalty."
+        ),
+    )
+    max_tokens: int | None = Field(
+        default=None,
+        description=(
+            "Maximum number of completion tokens. If omitted, model generates tokens until it "
+            "hits stop or context limit."
+        ),
+    )
+    return_images: bool | None = Field(
+        default=None,
+        description="If True, attempt to return images (closed beta feature).",
+    )
+    return_related_questions: bool | None = Field(
+        default=None,
+        description="If True, attempt to return related questions (closed beta feature).",
+    )
+    search_domain_filter: list[Any] | None = Field(
+        default=None,
+        description=(
+            "List of domains to limit or exclude in the online search. Example: ['example.com', '-twitter.com']. "
+            "Supports up to 3 entries. (Closed beta feature.)"
+        ),
+    )
+    search_recency_filter: str | None = Field(
+        default=None,
+        description=(
+            "Returns search results within a specified time interval: 'month', 'week', 'day', or 'hour'."
+        ),
+    )
+    stream: bool | None = Field(
+        default=None,
+        description=(
+            "If True, response is returned incrementally via Server-Sent Events (SSE)."
+        ),
+    )
+    temperature: float | None = Field(
+        default=None,
+        ge=0.0,
+        lt=2.0,
+        description=(
+            "Controls randomness of sampling, range [0, 2). Higher => more random. "
+            "Defaults to 0.2."
+        ),
+    )
+    top_k: int | None = Field(
+        default=None,
+        ge=0,
+        le=2048,
+        description=(
+            "Top-K filtering. 0 disables top-k filtering. If set, only the top K tokens are considered. "
+            "We recommend altering either top_k or top_p, but not both."
+        ),
+    )
+    top_p: float | None = Field(
+        default=None,
+        ge=0.0,
+        le=1.0,
+        description=(
+            "Nucleus sampling threshold. We recommend altering either top_k or top_p, but not both."
+        ),
+    )
+
+    @model_validator(mode="before")
+    def validate_penalties(cls, values):
+        """
+        Disallow using both frequency_penalty != 1.0 and presence_penalty != 0.0 at once,
+        since the docs say they're incompatible.
+        """
+        freq_pen = values.get("frequency_penalty", 1.0)
+        pres_pen = values.get("presence_penalty", 0.0)
+
+        # The doc states frequency_penalty is incompatible with presence_penalty.
+        # We'll enforce that if presence_penalty != 0, frequency_penalty must be 1.0
+        # or vice versa. Adjust logic as needed.
+        if pres_pen != 0.0 and freq_pen != 1.0:
+            raise ValueError(
+                "presence_penalty is incompatible with frequency_penalty. "
+                "Please use only one: either presence_penalty=0 with freq_pen !=1, "
+                "or presence_penalty!=0 with freq_pen=1."
+            )
+        return values
+
+    def to_dict(self) -> dict:
+        """Return a dict suitable for JSON serialization and sending to Perplexity API."""
+        return self.model_dump(exclude_none=True)