agnipod 0.1.2

Official Python SDK for the AgniPod LLM API

AgniPod Python SDK

Official Python SDK for the AgniPod LLM API — chat completions, text generation, batch processing, and more.

Installation

pip install agnipod

Quick Start

1. Authenticate

# Option A: CLI login (recommended — saves to ~/.agnipod/credentials)
agnipod login

# On Windows (if 'agnipod' is not recognized):
python -m agnipod login

# Option B: Environment variable
export AGNIPOD_API_KEY="ak_live_your_key_here"        # Linux/macOS
set AGNIPOD_API_KEY=ak_live_your_key_here              # Windows CMD
$env:AGNIPOD_API_KEY="ak_live_your_key_here"           # Windows PowerShell

# Option C: Pass directly in code
from agnipod import AgniPod
client = AgniPod(api_key="ak_live_your_key_here")

2. Chat completion

from agnipod import chat

response = chat("What is artificial intelligence?", model="qwen3:8b")
print(response.content)

3. Text generation

from agnipod import generate

response = generate("Once upon a time", model="qwen3:8b")
print(response.content)

---

CLI Commands

After installing the package, the agnipod command is available:

agnipod login          # Save your API key locally (~/.agnipod/credentials)
agnipod logout         # Remove saved credentials
agnipod status         # Show current auth config & key sources
agnipod models         # List all available models
agnipod version        # Print SDK version
agnipod help           # Show all commands

Windows: If agnipod is not recognized, use python -m agnipod <command> instead.

Login

$ agnipod login
  Enter your API key (starts with ak_live_ or ak_test_)
  API Key: ****

  ✓ API key saved to /home/user/.agnipod/credentials

You can also pass the key directly: agnipod login ak_live_your_key

---

Full Client Usage

For production applications, use the AgniPod client class:

from agnipod import AgniPod

client = AgniPod()  # reads key from env var or ~/.agnipod/credentials

Chat Completions

response = client.chat.create(
    model="qwen3:8b",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain quantum computing in simple terms"},
    ],
    temperature=0.7,
    max_tokens=512,
)

print(response.content)
print(f"Tokens: {response.prompt_tokens} in / {response.completion_tokens} out")

Streaming

stream = client.chat.create(
    model="qwen3:8b",
    messages=[{"role": "user", "content": "Tell me a story"}],
    stream=True,
)

for chunk in stream:
    print(chunk, end="", flush=True)
print()

Text Generation

response = client.generate.create(
    model="qwen3:8b",
    prompt="Write a Python function to reverse a string",
    system_prompt="You are a senior Python developer.",
    max_tokens=256,
)
print(response.content)

Models

# List all available models
models = client.models.list()
for model in models:
    print(model.id, model.owned_by)

# Get a specific model
model = client.models.get("qwen3:8b")

Batch Processing

# Create a batch
result = client.batches.create(items=[
    {
        "custom_id": "q1",
        "model": "qwen3:8b",
        "messages": [{"role": "user", "content": "What is 2+2?"}],
    },
    {
        "custom_id": "q2",
        "model": "qwen3:8b",
        "messages": [{"role": "user", "content": "What is 3+3?"}],
    },
])
print(f"Batch {result.batch_id} created ({result.total_items} items)")

# Start processing
client.batches.start(result.batch_id)

# Wait for completion (polls automatically)
batch = client.batches.wait(result.batch_id, poll_interval=5)
print(f"Status: {batch.status}")

# Download results
client.batches.results(result.batch_id, save_to="results.jsonl")

Upload JSONL Batch

result = client.batches.create_jsonl(
    file_path="my_batch.jsonl",
    batch_name="Weekly analysis",
)
client.batches.start(result.batch_id)

Billing & Wallet

# Check balance
balance = client.billing.balance()
print(balance)  # $1.2345 USD

# Usage summary
usage = client.billing.usage(days=7)

# Transaction history
for txn in client.billing.transactions(limit=5):
    print(txn.transaction_type, txn.amount)

---

Error Handling

All errors inherit from AgniPodError for easy catching:

from agnipod import AgniPod, AgniPodError, AuthenticationError, InsufficientBalanceError

client = AgniPod()

try:
    response = client.chat.create(
        model="qwen3:8b",
        messages=[{"role": "user", "content": "Hello"}],
    )
except AuthenticationError:
    print("Invalid API key")
except InsufficientBalanceError:
    print("Add funds to your wallet")
except AgniPodError as e:
    print(f"API error: {e.message} (status={e.status_code})")

Exception Hierarchy

Exception	HTTP Code	When
AuthenticationError	401	Invalid / revoked API key
InsufficientBalanceError	402	Wallet balance too low
PermissionDeniedError	403	Wallet inactive
BadRequestError	400	Invalid request payload
NotFoundError	404	Model / resource not found
RateLimitError	429	Too many requests
ServiceUnavailableError	503	No GPU workers available
ServerError	5xx	Unexpected server error
APIConnectionError	—	Network unreachable
APITimeoutError	—	Request timed out
StreamError	—	Error during streaming
ValidationError	—	Client-side validation
ConfigurationError	—	Missing API key / bad config

---

Configuration

API key resolution order (first non-empty wins):

Priority	Method
1	AgniPod(api_key="ak_live_...")
2	AGNIPOD_API_KEY environment variable
3	~/.agnipod/credentials (via agnipod login)

client = AgniPod(
    api_key="ak_live_...",         # optional if env/creds set
    base_url="https://custom.api", # override API URL
    timeout=60,                    # request timeout (seconds)
    max_retries=3,                 # retries on transient errors
)

Context Manager

with AgniPod() as client:
    response = client.chat.create(
        model="qwen3:8b",
        messages=[{"role": "user", "content": "Hello"}],
    )
# session auto-closed

---

Requirements

• Python 3.9+
• requests >= 2.28.0