Ollama

export OLLAMA_BASE_URL="http://localhost:11434"  # Optional, this is the default

# macOS
brew install ollama

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# Windows
# Download from https://ollama.com/download

# Pull a model before using it
ollama pull llama3.2
ollama pull mistral
ollama pull codellama

from upsonic import infer_model

# Make sure Ollama is running and model is pulled
model = infer_model("ollama/llama3.2")

from upsonic.models.openai import OpenAIChatModel, OpenAIChatModelSettings

settings = OpenAIChatModelSettings(
    max_tokens=2048,
    temperature=0.7
)

model = OpenAIChatModel(
    model_name="llama3.2",
    provider="ollama",
    settings=settings
)

from upsonic import Agent, Task, infer_model

model = infer_model("ollama/llama3.2")
agent = Agent(model=model)

task = Task("Explain quantum mechanics")
result = agent.do(task)

from upsonic import Agent, Task, infer_model

def calculate(expression: str) -> float:
    """Evaluate a mathematical expression."""
    return eval(expression)

model = infer_model("ollama/llama3.2")
agent = Agent(model=model, tools=[calculate])

task = Task("Calculate 25 * 36 + 100")
result = agent.do(task)

from upsonic import Agent, Task, infer_model

# Use codellama for code tasks
model = infer_model("ollama/codellama")
agent = Agent(model=model)

task = Task("Write a Python function to find prime numbers")
result = agent.do(task)

from upsonic import Agent, Task, infer_model

model = infer_model("ollama/llama3.2")
agent = Agent(model=model)

task = Task("Write a short story about space exploration")

async for chunk in agent.do_stream(task):
    print(chunk, end="", flush=True)

from upsonic import Agent, Task, infer_model
from upsonic.messages import ImageUrl

# Use a vision model like llava
model = infer_model("ollama/llava")
agent = Agent(model=model)

task = Task(
    description="What's in this image?",
    attachments=[
        ImageUrl(url="file:///path/to/image.jpg")
    ]
)

result = agent.do(task)

from upsonic import Agent, Task, infer_model

# First request loads model into memory
model = infer_model("ollama/llama3.2")
agent = Agent(model=model)

task1 = Task("Hello")
result1 = agent.do(task1)  # Model loads (slower)

# Subsequent requests are faster as model stays in memory
task2 = Task("How are you?")
result2 = agent.do(task2)  # Faster

# Keep model loaded
ollama run llama3.2

# Or use API to keep alive
# Set keep_alive in model file or API call

from upsonic.models.openai import OpenAIChatModel, OpenAIChatModelSettings

settings = OpenAIChatModelSettings(
    max_tokens=4096,
    temperature=0.7,
    top_p=0.9,
    seed=42,
    stop_sequences=["END"]
)

model = OpenAIChatModel(
    model_name="llama3.2",
    provider="ollama",
    settings=settings
)

# Check if GPU is being used
ollama ps

# Models will show GPU memory usage

# Pull latest version
ollama pull llama3.2

# Pull specific size
ollama pull llama3.1:70b

ollama list

ollama rm llama3.2

ollama pull llama3.2  # Re-pull to update

# Modelfile
FROM llama3.2

# Set temperature
PARAMETER temperature 0.8

# Set system prompt
SYSTEM You are a helpful coding assistant

# Set stop sequences
PARAMETER stop "END"

ollama create mycustommodel -f Modelfile

from upsonic import infer_model

model = infer_model("ollama/mycustommodel")

# Make sure Ollama is running
# Start Ollama service:
# macOS/Linux: ollama serve
# Or use the Ollama app

# Use smaller model
ollama pull llama3.2:1b

# Or use quantized version
ollama pull llama3.1:8b-q4_0

# Use GPU acceleration
# Install CUDA drivers for NVIDIA GPUs

# Or use smaller/quantized models
ollama pull llama3.2:3b-q4_0