377 lines
13 KiB
Python
377 lines
13 KiB
Python
# SPDX-License-Identifier: MIT
|
|
# Copyright (c) 2026 Rug Munch Media LLC
|
|
#
|
|
# Part of Pry — https://git.rugmunch.io/RugMunchMedia/pryscraper
|
|
# Licensed under MIT. See LICENSE.
|
|
"""Concrete LLM provider implementations."""
|
|
|
|
from __future__ import annotations
|
|
|
|
import logging
|
|
import os
|
|
|
|
from llm_providers.base import LLMProvider, LLMResponse
|
|
from llm_providers.registry import LLMRegistry
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class OpenAIProvider(LLMProvider):
|
|
name = "openai"
|
|
cost_per_1k_input = 0.00015 # gpt-4o-mini
|
|
cost_per_1k_output = 0.0006
|
|
referral_url = "https://platform.openai.com/signup?via=pry"
|
|
|
|
def __init__(self, api_key: str = ""):
|
|
self.api_key = api_key or os.getenv("OPENAI_API_KEY", "")
|
|
|
|
async def complete(
|
|
self, prompt, system="", max_tokens=1024, temperature=0.7, model="gpt-4o-mini"
|
|
):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
messages = []
|
|
if system:
|
|
messages.append({"role": "system", "content": system})
|
|
messages.append({"role": "user", "content": prompt})
|
|
resp = await client.post(
|
|
"https://api.openai.com/v1/chat/completions",
|
|
json={
|
|
"model": model,
|
|
"messages": messages,
|
|
"max_tokens": max_tokens,
|
|
"temperature": temperature,
|
|
},
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=60,
|
|
)
|
|
data = resp.json()
|
|
choice = data["choices"][0]
|
|
return LLMResponse(
|
|
text=choice["message"]["content"],
|
|
model=model,
|
|
provider=self.name,
|
|
input_tokens=data["usage"]["prompt_tokens"],
|
|
output_tokens=data["usage"]["completion_tokens"],
|
|
raw=data,
|
|
)
|
|
|
|
async def embed(self, text, model="text-embedding-3-small"):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
resp = await client.post(
|
|
"https://api.openai.com/v1/embeddings",
|
|
json={"input": text, "model": model},
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=30,
|
|
)
|
|
return resp.json()["data"][0]["embedding"]
|
|
|
|
|
|
class AnthropicProvider(LLMProvider):
|
|
name = "anthropic"
|
|
cost_per_1k_input = 0.00025 # claude-3-haiku
|
|
cost_per_1k_output = 0.00125
|
|
referral_url = "https://console.anthropic.com/?ref=pry"
|
|
|
|
def __init__(self, api_key: str = ""):
|
|
self.api_key = api_key or os.getenv("ANTHROPIC_API_KEY", "")
|
|
|
|
async def complete(
|
|
self, prompt, system="", max_tokens=1024, temperature=0.7, model="claude-3-haiku-20240307"
|
|
):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
body = {
|
|
"model": model,
|
|
"max_tokens": max_tokens,
|
|
"temperature": temperature,
|
|
"messages": [{"role": "user", "content": prompt}],
|
|
}
|
|
if system:
|
|
body["system"] = system
|
|
resp = await client.post(
|
|
"https://api.anthropic.com/v1/messages",
|
|
json=body,
|
|
headers={"x-api-key": self.api_key, "anthropic-version": "2023-06-01"},
|
|
timeout=60,
|
|
)
|
|
data = resp.json()
|
|
return LLMResponse(
|
|
text=data["content"][0]["text"],
|
|
model=model,
|
|
provider=self.name,
|
|
input_tokens=data["usage"]["input_tokens"],
|
|
output_tokens=data["usage"]["output_tokens"],
|
|
raw=data,
|
|
)
|
|
|
|
async def embed(self, text, model=""):
|
|
raise NotImplementedError("Anthropic doesn't have a public embedding API yet")
|
|
|
|
|
|
class GoogleProvider(LLMProvider):
|
|
name = "google"
|
|
cost_per_1k_input = 0.000125 # gemini-flash
|
|
cost_per_1k_output = 0.000375
|
|
referral_url = "https://aistudio.google.com/?utm_source=pry"
|
|
|
|
def __init__(self, api_key: str = ""):
|
|
self.api_key = api_key or os.getenv("GOOGLE_API_KEY", "")
|
|
|
|
async def complete(
|
|
self, prompt, system="", max_tokens=1024, temperature=0.7, model="gemini-1.5-flash"
|
|
):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
url = f"https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={self.api_key}"
|
|
contents = [{"role": "user", "parts": [{"text": prompt}]}]
|
|
body = {
|
|
"contents": contents,
|
|
"generationConfig": {"maxOutputTokens": max_tokens, "temperature": temperature},
|
|
}
|
|
if system:
|
|
body["systemInstruction"] = {"parts": [{"text": system}]}
|
|
resp = await client.post(url, json=body, timeout=60)
|
|
data = resp.json()
|
|
text = data["candidates"][0]["content"]["parts"][0]["text"]
|
|
usage = data.get("usageMetadata", {})
|
|
return LLMResponse(
|
|
text=text,
|
|
model=model,
|
|
provider=self.name,
|
|
input_tokens=usage.get("promptTokenCount", 0),
|
|
output_tokens=usage.get("candidatesTokenCount", 0),
|
|
raw=data,
|
|
)
|
|
|
|
async def embed(self, text, model="text-embedding-004"):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
url = f"https://generativelanguage.googleapis.com/v1beta/models/{model}:embedContent?key={self.api_key}"
|
|
resp = await client.post(url, json={"content": {"parts": [{"text": text}]}}, timeout=30)
|
|
return resp.json()["embedding"]["values"]
|
|
|
|
|
|
class CohereProvider(LLMProvider):
|
|
name = "cohere"
|
|
cost_per_1k_input = 0.0001
|
|
cost_per_1k_output = 0.0004
|
|
referral_url = "https://dashboard.cohere.com/welcome?ref=pry"
|
|
|
|
def __init__(self, api_key: str = ""):
|
|
self.api_key = api_key or os.getenv("COHERE_API_KEY", "")
|
|
|
|
async def complete(
|
|
self, prompt, system="", max_tokens=1024, temperature=0.7, model="command-r"
|
|
):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
body = {
|
|
"model": model,
|
|
"message": prompt,
|
|
"max_tokens": max_tokens,
|
|
"temperature": temperature,
|
|
}
|
|
if system:
|
|
body["preamble"] = system
|
|
resp = await client.post(
|
|
"https://api.cohere.ai/v1/chat",
|
|
json=body,
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=60,
|
|
)
|
|
data = resp.json()
|
|
return LLMResponse(text=data["text"], model=model, provider=self.name, raw=data)
|
|
|
|
async def embed(self, text, model="embed-english-v3.0"):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
resp = await client.post(
|
|
"https://api.cohere.ai/v1/embed",
|
|
json={"texts": [text], "model": model, "input_type": "search_document"},
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=30,
|
|
)
|
|
return resp.json()["embeddings"][0]
|
|
|
|
|
|
class MistralProvider(LLMProvider):
|
|
name = "mistral"
|
|
cost_per_1k_input = 0.0002
|
|
cost_per_1k_output = 0.0006
|
|
referral_url = "https://console.mistral.ai/?ref=pry"
|
|
|
|
def __init__(self, api_key: str = ""):
|
|
self.api_key = api_key or os.getenv("MISTRAL_API_KEY", "")
|
|
|
|
async def complete(
|
|
self, prompt, system="", max_tokens=1024, temperature=0.7, model="mistral-small-latest"
|
|
):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
messages = []
|
|
if system:
|
|
messages.append({"role": "system", "content": system})
|
|
messages.append({"role": "user", "content": prompt})
|
|
resp = await client.post(
|
|
"https://api.mistral.ai/v1/chat/completions",
|
|
json={
|
|
"model": model,
|
|
"messages": messages,
|
|
"max_tokens": max_tokens,
|
|
"temperature": temperature,
|
|
},
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=60,
|
|
)
|
|
data = resp.json()
|
|
return LLMResponse(
|
|
text=data["choices"][0]["message"]["content"],
|
|
model=model,
|
|
provider=self.name,
|
|
input_tokens=data["usage"]["prompt_tokens"],
|
|
output_tokens=data["usage"]["completion_tokens"],
|
|
raw=data,
|
|
)
|
|
|
|
async def embed(self, text, model="mistral-embed"):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
resp = await client.post(
|
|
"https://api.mistral.ai/v1/embeddings",
|
|
json={"model": model, "input": [text]},
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=30,
|
|
)
|
|
return resp.json()["data"][0]["embedding"]
|
|
|
|
|
|
class OllamaProvider(LLMProvider):
|
|
name = "ollama"
|
|
cost_per_1k_input = 0.0 # Self-hosted, free
|
|
cost_per_1k_output = 0.0
|
|
referral_url = "https://ollama.com" # No referral program, just self-hosted
|
|
|
|
def __init__(self, base_url: str = "", model: str = "llama3.2"):
|
|
self.base_url = base_url or os.getenv("PRY_OLLAMA_URL", "http://localhost:11434")
|
|
self.default_model = model
|
|
|
|
async def complete(self, prompt, system="", max_tokens=1024, temperature=0.7, model=""):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
model = model or self.default_model
|
|
body = {
|
|
"model": model,
|
|
"prompt": prompt,
|
|
"stream": False,
|
|
"options": {"temperature": temperature, "num_predict": max_tokens},
|
|
}
|
|
if system:
|
|
body["system"] = system
|
|
resp = await client.post(f"{self.base_url}/api/generate", json=body, timeout=300)
|
|
data = resp.json()
|
|
return LLMResponse(
|
|
text=data["response"],
|
|
model=model,
|
|
provider=self.name,
|
|
input_tokens=data.get("prompt_eval_count", 0),
|
|
output_tokens=data.get("eval_count", 0),
|
|
raw=data,
|
|
)
|
|
|
|
async def embed(self, text, model=""):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
model = model or "nomic-embed-text"
|
|
resp = await client.post(
|
|
f"{self.base_url}/api/embeddings", json={"model": model, "prompt": text}, timeout=60
|
|
)
|
|
return resp.json()["embedding"]
|
|
|
|
|
|
class OpenRouterProvider(LLMProvider):
|
|
name = "openrouter"
|
|
cost_per_1k_input = 0.0 # Free models available
|
|
cost_per_1k_output = 0.0
|
|
referral_url = "https://openrouter.ai/?ref=pry"
|
|
|
|
def __init__(self, api_key: str = ""):
|
|
self.api_key = api_key or os.getenv("OPENROUTER_API_KEY", "")
|
|
|
|
async def complete(
|
|
self,
|
|
prompt,
|
|
system="",
|
|
max_tokens=1024,
|
|
temperature=0.7,
|
|
model="meta-llama/llama-3.2-3b-instruct:free",
|
|
):
|
|
from client import get_client
|
|
|
|
client = await get_client()
|
|
messages = []
|
|
if system:
|
|
messages.append({"role": "system", "content": system})
|
|
messages.append({"role": "user", "content": prompt})
|
|
resp = await client.post(
|
|
"https://openrouter.ai/api/v1/chat/completions",
|
|
json={
|
|
"model": model,
|
|
"messages": messages,
|
|
"max_tokens": max_tokens,
|
|
"temperature": temperature,
|
|
},
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
timeout=60,
|
|
)
|
|
data = resp.json()
|
|
usage = data.get("usage", {})
|
|
return LLMResponse(
|
|
text=data["choices"][0]["message"]["content"],
|
|
model=model,
|
|
provider=self.name,
|
|
input_tokens=usage.get("prompt_tokens", 0),
|
|
output_tokens=usage.get("completion_tokens", 0),
|
|
raw=data,
|
|
)
|
|
|
|
async def embed(self, text, model=""):
|
|
raise NotImplementedError("OpenRouter focuses on chat; use dedicated embedding providers")
|
|
|
|
|
|
def register_default_providers(registry: LLMRegistry) -> None:
|
|
"""Register all providers whose API keys are set in environment."""
|
|
api_key_map = {
|
|
"openai": os.getenv("OPENAI_API_KEY"),
|
|
"anthropic": os.getenv("ANTHROPIC_API_KEY"),
|
|
"google": os.getenv("GOOGLE_API_KEY"),
|
|
"cohere": os.getenv("COHERE_API_KEY"),
|
|
"mistral": os.getenv("MISTRAL_API_KEY"),
|
|
"openrouter": os.getenv("OPENROUTER_API_KEY"),
|
|
}
|
|
provider_classes = {
|
|
"openai": OpenAIProvider,
|
|
"anthropic": AnthropicProvider,
|
|
"google": GoogleProvider,
|
|
"cohere": CohereProvider,
|
|
"mistral": MistralProvider,
|
|
"openrouter": OpenRouterProvider,
|
|
}
|
|
for name, cls in provider_classes.items():
|
|
key = api_key_map.get(name)
|
|
if key:
|
|
registry.register(cls(api_key=key))
|
|
# Ollama is always available if running locally
|
|
registry.register(OllamaProvider())
|