from litellm import completion
import os

API_BASE = os.environ.get("INCREDIBLE_API_BASE", "https://api.incredible.one/v1")
API_KEY = os.environ.get("INCREDIBLE_API_KEY")

if not API_KEY:
    raise RuntimeError("Set INCREDIBLE_API_KEY before running this example")

response = completion(
    model="openai/small-1",
    api_base=API_BASE,
    api_key=API_KEY,
    messages=[{"role": "user", "content": "Hello, world"}],
    temperature=0.2,
)
print(response)

from litellm import completion
import os

API_BASE = os.environ.get("INCREDIBLE_API_BASE", "https://api.incredible.one/v1")
API_KEY = os.environ.get("INCREDIBLE_API_KEY")

if not API_KEY:
    raise RuntimeError("Set INCREDIBLE_API_KEY before running this example")

stream = completion(
    model="openai/small-1",
    api_base=API_BASE,
    api_key=API_KEY,
    messages=[{"role": "user", "content": "Please stream a short greeting."}],
    temperature=0.2,
    stream=True,
)

chunk_index = 0
assembled_text = []
for chunk in stream:
    chunk_index += 1
    choices = chunk.get("choices") or []
    if not choices:
        continue
    delta = choices[0].get("delta") or {}
    content_piece = delta.get("content")
    if not content_piece:
        continue
    assembled_text.append(content_piece)
    print(f"Chunk {chunk_index}: {content_piece!r}")

print("\nFinal assembled text:\n" + "".join(assembled_text))

from litellm import completion
import os

API_BASE = os.environ.get("INCREDIBLE_API_BASE", "https://api.incredible.one/v1")
API_KEY = os.environ.get("INCREDIBLE_API_KEY")

if not API_KEY:
    raise RuntimeError("Set INCREDIBLE_API_KEY before running this example")

examples = [
    (
        "Friendly persona",
        [
            {
                "role": "system",
                "content": (
                    "You are an enthusiastic assistant that always keeps answers concise, friendly,"
                    " and formatted with exactly two bullet points using bolded headings, followed by"
                    " a one-sentence closing remark."
                ),
            },
            {"role": "user", "content": "Hello there!"},
            {
                "role": "assistant",
                "content": "Hi! It's great to meet you. How can I support your planning or analysis today?",
            },
            {
                "role": "user",
                "content": "Give me two bullet points about why streaming demos are useful.",
            },
        ],
        0.2,
    ),
    (
        "Terse response",
        [
            {
                "role": "system",
                "content": (
                    "You must answer every request in exactly one word."
                    " The word must be lowercase, contain no punctuation, and convey the best possible answer."
                    " If the request cannot be satisfied with a single word, reply with 'unknown'."
                ),
            },
            {
                "role": "user",
                "content": "Give me two bullet points about why streaming demos are useful.",
            },
        ],
        0.0,
    ),
    (
        "Assistant-context",
        [
            {
                "role": "system",
                "content": "You are a factual assistant who trusts previous assistant messages as ground truth.",
            },
            {
                "role": "assistant",
                "content": "Status update: The system just deployed version 1.2.7 successfully.",
            },
            {
                "role": "user",
                "content": "What version is currently live?",
            },
        ],
        0.2,
    ),
]

for label, messages, temperature in examples:
    response = completion(
        model="openai/small-1",
        api_base=API_BASE,
        api_key=API_KEY,
        messages=messages,
        temperature=temperature,
    )
    choice = response.choices[0]
    print(f"\n{label} role: {choice.message.role}")
    print(f"{label} content:\n{choice.message.content}")