Build with Luna — SDK Overview

# Local (default)
LLM_PROVIDER=ollama
OLLAMA_MODEL=qwen2.5:7b

# Groq — fastest hosted option
LLM_PROVIDER=groq
GROQ_API_KEY=gsk_...
GROQ_MODEL=llama-3.3-70b-versatile

# Anthropic Claude
LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-sonnet-4-6

# Any OpenAI-compatible endpoint (OpenRouter, LM Studio, llama.cpp)
LLM_PROVIDER=openai-compatible
OPENAI_BASE_URL=https://openrouter.ai/api/v1
OPENAI_API_KEY=sk-or-...
OPENAI_MODEL=meta-llama/llama-3.3-70b-instruct

LLM_PROVIDER=groq
GROQ_API_KEY=gsk_...

# Override just the coding agent to a local coder
CODING_PROVIDER=ollama
CODING_MODEL=qwen2.5-coder:7b

import httpx

with httpx.stream(
    "POST", "http://localhost:8899/api/chat/stream",
    json={"message": "Summarise my tasks for today", "conversation_id": None},
    timeout=60,
) as r:
    for line in r.iter_lines():
        if line.startswith("data: "):
            print(line[6:], end="", flush=True)

const response = await fetch('http://localhost:8899/api/chat/stream', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ message: 'What is on my calendar today?', conversation_id: null }),
});

const reader = response.body.getReader();
const decoder = new TextDecoder();
while (true) {
  const { done, value } = await reader.read();
  if (done) break;
  const chunk = decoder.decode(value);
  for (const line of chunk.split('\n')) {
    if (line.startsWith('data: ')) process.stdout.write(line.slice(6));
  }
}

with httpx.stream(
    "POST", "http://localhost:8899/api/coding/stream",
    json={
        "message": "Create a FastAPI endpoint that returns a hello world JSON",
        "workspace_root": "/my/project",
    },
    timeout=120,
) as r:
    for line in r.iter_lines():
        if line.startswith("data: "):
            import json
            event = json.loads(line[6:])
            print(event["type"], "→", event.get("content", "")[:80])

import asyncio, sys
sys.path.insert(0, "/path/to/Luna")

from backend.services.llm import ollama

async def main():
    async for token in ollama.stream_chat(
        messages=[{"role": "user", "content": "Explain async/await in Python"}],
        system_prompt="You are a concise technical writer.",
    ):
        print(token, end="", flush=True)

asyncio.run(main())

result = asyncio.run(
    ollama.complete("Extract the key topics from: " + my_text, temperature=0.2)
)
print(result)

from backend.services.memory_manager import MemoryManager
from backend.models.database import SessionLocal

db = SessionLocal()
mm = MemoryManager(db)

# Store a fact
asyncio.run(mm.store_fact(
    "User prefers dark mode",
    category="preference",
    importance=0.8,
))

# Retrieve semantically relevant facts
facts = asyncio.run(mm.retrieve_relevant("what UI preferences does the user have?"))
for f in facts:
    print(f.content, f.confidence)

db.close()

from backend.services.personality import PersonalityEngine
from backend.models.database import SessionLocal

db = SessionLocal()
engine = PersonalityEngine(db)

# Update mood based on the user's last message
engine.update_mood("I'm so excited about this!")

# Get a personality-aware system prompt
prompt = engine.build_personality_prompt(user_name="Alex")
print(prompt[:300])
db.close()