Merge pull request #48 from premAI-io/perplexity

Perplexity Provider.

e2e.py

@@ -15,6 +15,7 @@
     mistral,
     octoai,
     openai,
+    perplexity,
     prem,
     replicate,
     together,
@@ -71,6 +72,7 @@ def run_single_connector(connector_name: str) -> None:
         "mistralai": (mistral.MistralConnector, "MISTRAL_AI_API_KEY"),
         "prem": (prem.PremConnector, "PREMAI_BEARER_TOKEN"),
         "deepinfra": (openai.OpenAIConnector, "DEEP_INFRA_API_KEY"),
+        "perplexity": (perplexity.PerplexityAIConnector, "PERPLEXITY_API_KEY"),
     }
 
     if connector_name == "deepinfra":
@@ -99,10 +101,12 @@ def run_single_connector(connector_name: str) -> None:
 
     if len(text2text_models) > 0:
         model_object = text2text_models[0]
+
         parameters = {}
         parameters["model"] = model_object["slug"]
-        messages = [{"role": "user", "content": "Hello, how is it going?"}]
-        messages.append({"role": "system", "content": "Behave like Rick Sanchez."})
+
+        messages = [{"role": "system", "content": "Behave like Rick Sanchez."}]
+        messages.append({"role": "user", "content": "Hello, how is it going?"})
         parameters["messages"] = messages
 
         print(f"Testing model {model_object['slug']} from {connector['provider']} connector \n\n\n")

prem_utils/connectors/perplexity.py

@@ -0,0 +1,75 @@
+from collections.abc import Sequence
+from typing import Any
+
+from prem_utils.connectors.openai import OpenAIConnector
+
+
+class PerplexityAIConnector(OpenAIConnector):
+    def __init__(self, api_key: str, base_url: str = "https://api.perplexity.ai", prompt_template: str = None) -> None:
+        super().__init__(prompt_template=prompt_template, base_url=base_url, api_key=api_key)
+
+    def chat_completion(
+        self,
+        model: str,
+        messages: list[dict[str, Any]],
+        max_tokens: int = None,
+        frequency_penalty: float = 0.1,
+        log_probs: int = None,
+        logit_bias: dict[str, float] = None,
+        presence_penalty: float = 0,
+        seed: int | None = None,
+        stop: str | list[str] = None,
+        stream: bool = False,
+        temperature: float = 1,
+        top_p: float = 1,
+        tools: list[dict[str, Any]] = None,
+        tool_choice: dict = None,
+    ):
+        if "perplexity" in model:
+            model = model.replace("perplexity/", "", 1)
+
+        return super().chat_completion(
+            model,
+            messages,
+            max_tokens,
+            frequency_penalty,
+            log_probs,
+            logit_bias,
+            presence_penalty,
+            seed,
+            stop,
+            stream,
+            temperature,
+            top_p,
+            tools,
+            tool_choice,
+        )
+
+    def embeddings(
+        self,
+        model: str,
+        input: str | Sequence[str] | Sequence[int] | Sequence[Sequence[int]],
+        encoding_format: str = "float",
+        user: str = None,
+    ):
+        raise NotImplementedError
+
+    def finetuning(
+        self, model: str, training_data: list[dict], validation_data: list[dict] | None = None, num_epochs: int = 3
+    ) -> str:
+        raise NotImplementedError
+
+    def get_finetuning_job(self, job_id) -> dict[str, Any]:
+        raise NotImplementedError
+
+    def generate_image(
+        self,
+        model: str,
+        prompt: str,
+        size: str = "1024x1024",
+        n: int = 1,
+        quality: str = "standard",
+        style: str = "vivid",
+        response_format: str = "url",
+    ):
+        raise NotImplementedError

prem_utils/models.json

@@ -623,6 +623,57 @@
                     "coming_soon": true
                 }
             ]
+        },
+
+        {
+            "provider": "perplexity",
+            "models": [
+                {
+                    "slug": "perplexity/codellama-34b-instruct",
+                    "model_type": "text2text",
+                    "context_window": 16384
+                },
+                {
+                    "slug": "perplexity/codellama-70b-instruct",
+                    "model_type": "text2text",
+                    "context_window": 16384
+                },
+                {
+                    "slug": "perplexity/llama-2-70b-chat",
+                    "model_type": "text2text",
+                    "context_window": 4096
+                },
+                {
+                    "slug": "perplexity/mistral-7b-instruct",
+                    "model_type": "text2text",
+                    "context_window": 8192
+                },
+                {
+                    "slug": "perplexity/mixtral-8x7b-instruct",
+                    "model_type": "text2text",
+                    "context_window": 8192
+                },
+                {
+                    "slug": "perplexity/pplx-7b-chat",
+                    "model_type": "text2text",
+                    "context_window": 8192
+                },
+                {
+                    "slug": "perplexity/pplx-70b-chat",
+                    "model_type": "text2text",
+                    "context_window": 8192
+                },
+                {
+                    "slug": "perplexity/pplx-7b-online",
+                    "model_type": "text2text",
+                    "context_window": 8192
+                },
+                {
+                    "slug": "perplexity/pplx-70b-online",
+                    "model_type": "text2text",
+                    "context_window": 8192
+                }
+            ]
         }
     ]
 }