from memfuse.llm import OpenAI
from memfuse import MemFuse
import os
 
# Initialize MemFuse
memfuse_client = MemFuse()
 
memory = memfuse_client.init(user="alice")
 
# Initialize LLM client with memory
llm_client = OpenAI(
    api_key=os.getenv("OPENAI_API_KEY"),
    memory=memory
)
 
# Make a chat completion request
response = llm_client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "I'm planning a trip to Mars. What is the gravity there?"}]
)
 
print(f"Response: {response.choices[0].message.content}")
 
# Ask a follow-up question - MemFuse automatically recalls context
followup_response = llm_client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "What are some challenges of living on that planet?"}]
)
 
print(f"Follow-up: {followup_response.choices[0].message.content}")

from memfuse.llm import OpenAI
from memfuse import MemFuse

memfuse = MemFuse()
 
# All context parameters are optional - memories in MemFuse can exist with flexible context combinations
memory = memfuse.init(user="bob")
 
client = OpenAI(
    api_key=os.getenv("OPENAI_API_KEY"),
    memory=memory
)
 
# You can check the memory identifiers if needed
print(f"Using memory for conversation: {memory}")
# Output would be something like:
# Memory(
#   user='bob', agent='agent_default', session='bob-agent_default-ea7a0d5a',`
#   user_id='auto-generated-user-id', agent_id='auto-generated-agent-id', session_id='auto-generated-session-id'
# )

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "I'm working on a project about space exploration. Can you tell me something interesting about Mars?"}],
)
 
print(response.choices[0].message.content)
# Output: "Hi there! Mars has the largest volcano in the solar system, Olympus Mons, which stands at about 13.6 miles (22 kilometers) high—nearly three times the height of Mount Everest."

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "That's awesome! My project is about how humans could live on planet someday. Any thoughts on that?"}]
)
 
print(response.choices[0].message.content)
# Output: "Glad you liked that! For humans to live on Mars, we'd need to tackle a few big challenges—like the thin atmosphere, which is mostly carbon dioxide and offers little protection from solar radiation."

# Get AI to speak like a cowboy
cowboy_system_prompt = """
Respond in an authentic cowboy dialect. Use:
- Cowboy slang: "pardner," "fixin' to," "y'all," "reckon," "howdy"
- Drop g's from -ing words: "ridin'," "talkin'"
- Western expressions: "tough as a boot heel," "right as rain"
- Address user as "pardner" or "buckaroo"
- Reference ranch life: "the trail," "homestead," "saddlebags"
- Use contractions and informal grammar
- Start sentences with "Well shoot," "Listen here," or "I reckon"
 
Maintain accuracy while speaking like you grew up on the range. Don't explain you're using a cowboy persona.
"""
response = client.chat.completions.create(
    model="gpt-4o-mini", # Or use a different model
    messages=[
      {
        "role": "system", "content": cowboy_system_prompt
      },
      {
        "role": "user", "content": "I want to focus on food production since that seems like a major challenge. How could we grow enough food to sustain a colony on the planet?"
      }
    ]
)
 
print(response.choices[0].message.content)
# Output: "Well, pardner, let me tell ya 'bout feedin' folks on the red frontier! Them fancy hydroponics and aeroponics setups would be the trail boss of Mars farmin'! No dirt needed, and they use a measly 5% of the water compared to plowin' fields back on the homestead. You can stack 'em higher than a horse can kick, makin' the most of your limited bunkhouse space! Y'all could be roundin' up fresh grub every 24-30 days if conditions are right as rain. And listen here - tough-as-nails crops like taters, kale, and them soybeans are hardy enough to survive the Martian range and keep your cowboys and cowgirls properly fed. Say, partner, you got any particular crop seeds you're fixin' to pack in your saddlebags for this here project?"

from memfuse.llm import Anthropic
 
anthropic_client = Anthropic(
    api_key=os.getenv("ANTHROPIC_API_KEY"),
    memory=memory
)
 
anthropic_response = anthropic_client.messages.create(
    model="claude-3-5-haiku-latest",
    max_tokens=1024,
    messages=[
      {
        "role": "user",
        "content": [{"type": "text", "text": "Potatoes seem like a good start, just like in \"The Martian\"! What about water? How would colonists get enough water?"}]
      }
    ],
)
 
# Display response content
if anthropic_response.content and isinstance(anthropic_response.content, list):
    for content_item in anthropic_response.content:
        if hasattr(content_item, 'text'):
            print(content_item.text)
else:
    print("Anthropic response content not in expected format or empty.")
# Output: "The Martian had it right with potatoes! For water, we've got options—there's subsurface ice at the poles and even at mid-latitudes, which could be mined and melted. Mars also has water bound in soil minerals that could be extracted. The real game-changer would be closed-loop water recycling systems that reclaim nearly 100% of water from everything—sweat, urine, washing, you name it. Not glamorous but essential! Would you incorporate water recycling into your habitat designs?"

# Access the current context properties
user = memory.user
user_id = memory.user_id
agent = memory.agent
agent_id = memory.agent_id
session = memory.session
session_id = memory.session_id
 
print(f"User: {user} (ID: {user_id})")
print(f"Agent: {agent} (ID: {agent_id})")
print(f"Session: {session} (ID: {session_id})")