ChatJimmy Python SDK (chatjimmy)
A developer-friendly, production-grade Python SDK for ChatJimmy — an ultra-fast AI inference platform powered by Taalas Inc.'s custom direct-to-silicon compute hardware running Meta's Llama 3.1 8B.
The SDK offers two intuitive developer experiences:
- Simplified Interface: A lightweight, minimal API (
client.chat("Hello")). - OpenAI-Compatible Interface: An interface mimicking the official OpenAI Python package (
client.chat.completions.create(...)) with full support for streaming and asynchronous calls.
Table of Contents
- Features
- Installation
- Quick Start
- Model Discovery
- Multi-Turn Conversations
- Real-Time Streaming
- Asynchronous Client (
AsyncChatJimmy) - File Attachments (Up to 50KB)
- Hardware Performance Metrics
- Configuration & Networking
- Error Handling
- Testing
- Known Limitations
- License
Features
- Zero Authentication: No API keys, login credentials, or browser automation required.
- OpenAI-Compatible API: Seamlessly drop into existing workflows expecting OpenAI-style
client.chat.completions.create(...)andChatCompletionobjects. - Dual Synchronous & Asynchronous Support: Both
ChatJimmy(sync viarequests) andAsyncChatJimmy(async viahttpx). - High-Performance Streaming: Token-by-token and chunk-by-chunk streaming with zero internal sentinel token leaks.
- Hardware Telemetry: Direct access to Taalas silicon execution metrics (TTFT, decode tokens/sec, prefill tokens/sec, compute latency).
- Multi-Turn Context: Automatic conversation tracking and session management.
- Attachment Support: Upload and analyze text documents or code snippets up to 50KB.
- Resilient Networking: Connection pooling with automatic exponential backoff retries on transient network errors.
Installation
Install the package via pip:
pip install chatjimmy
Or install with development dependencies:
pip install chatjimmy[dev]
Quick Start
1. Simplified Interface
from chatjimmy import ChatJimmy
client = ChatJimmy()
response = client.chat("What is the capital of Japan?")
print(response)
# Output: The capital of Japan is Tokyo.
2. OpenAI-Compatible Interface
from chatjimmy import ChatJimmy
client = ChatJimmy()
completion = client.chat.completions.create(
model="llama3.1-8B",
messages=[
{"role": "system", "content": "You are a concise, helpful tutor."},
{"role": "user", "content": "Explain transformers in machine learning in one sentence."},
]
)
print(completion.choices[0].message.content)
print(f"Usage: {completion.usage.total_tokens} total tokens")
Model Discovery
Discover supported models using either the OpenAI-style resource or direct helper:
from chatjimmy import ChatJimmy
client = ChatJimmy()
# OpenAI-style model list
models = client.models.list()
for model in models:
print(f"Model: {model.id} (Owner: {model.owned_by})")
# Retrieve specific model
model_info = client.models.retrieve("llama3.1-8B")
print("Retrieved:", model_info.id)
Multi-Turn Conversations
By default, ChatJimmy tracks conversation history automatically:
from chatjimmy import ChatJimmy
client = ChatJimmy()
# First turn
r1 = client.chat("Hello! My name is Arthur Dent and my favorite number is 42.")
print("Assistant:", r1)
# Second turn (context is preserved)
r2 = client.chat("What is my name and favorite number?")
print("Assistant:", r2)
# Output remembers Arthur Dent and 42!
# Reset history at any time
client.reset_conversation()
Real-Time Streaming
OpenAI-Compatible Streaming
from chatjimmy import ChatJimmy
client = ChatJimmy()
stream = client.chat.completions.create(
model="llama3.1-8B",
messages=[{"role": "user", "content": "Write a short poem about silicon processors."}],
stream=True,
)
for chunk in stream:
delta = chunk.choices[0].delta.content
if delta:
print(delta, end="", flush=True)
# Access hardware stats from completed stream
if stream.stats:
print(f"\n\nSpeed: {stream.stats.decode_rate:.0f} tokens/second")
Simplified Token Streaming
from chatjimmy import ChatJimmy
client = ChatJimmy()
for token in client.chat_stream("Count from 1 to 5"):
print(token, end="", flush=True)
Asynchronous Client (AsyncChatJimmy)
For asynchronous applications (e.g., FastAPI, Discord bots, async workers), use AsyncChatJimmy.
Async Chat Completions
import asyncio
from chatjimmy import AsyncChatJimmy
async def main():
async with AsyncChatJimmy() as client:
# Simplified async call
resp = await client.chat("Hello from async!")
print(resp.text)
# OpenAI-style async completion
comp = await client.chat.completions.create(
model="llama3.1-8B",
messages=[{"role": "user", "content": "What is 12 * 12?"}],
)
print("Answer:", comp.choices[0].message.content)
asyncio.run(main())
Async Streaming
import asyncio
from chatjimmy import AsyncChatJimmy
async def main():
async with AsyncChatJimmy() as client:
stream = await client.chat.completions.create(
model="llama3.1-8B",
messages=[{"role": "user", "content": "Count from 1 to 5"}],
stream=True,
)
async for chunk in stream:
content = chunk.choices[0].delta.content
if content:
print(content, end="", flush=True)
asyncio.run(main())
Concurrent Requests with asyncio
import asyncio
from chatjimmy import AsyncChatJimmy
async def query_city(city: str):
async with AsyncChatJimmy() as client:
resp = await client.chat(f"What is the capital of {city}?")
return city, resp.text.strip()
async def main():
cities = ["Japan", "France", "Germany", "Italy"]
tasks = [query_city(c) for c in cities]
results = await asyncio.gather(*tasks)
for city, answer in results:
print(f"{city}: {answer}")
asyncio.run(main())
File Attachments (Up to 50KB)
ChatJimmy accepts file attachments containing source code or text:
from chatjimmy import ChatJimmy, Attachment
client = ChatJimmy()
# Attach a local file
response = client.chat(
"Review this script for performance bugs.",
attachment="script.py",
)
print(response)
# Attach raw text in-memory
attachment = Attachment.from_text("def add(a, b): return a + b", name="math.py")
response = client.chat("What does this code do?", attachment=attachment)
print(response)
Hardware Performance Metrics
ChatJimmy operates on custom hardware developed by Taalas Inc. Each response provides fine-grained hardware execution statistics:
response = client.chat("Explain quantum entanglement.")
if response.stats:
print(f"Time to First Token: {response.stats.ttft * 1000:.2f} ms")
print(f"Decode Generation Rate:{response.stats.decode_rate:.0f} tokens/s")
print(f"Prefill Rate: {response.stats.prefill_rate:.0f} tokens/s")
print(f"Total Compute Time: {response.stats.total_time * 1000:.2f} ms")
print(f"Total Tokens: {response.stats.total_tokens}")
print(f"Exit Reason: {response.stats.reason}")
| Metric Field | Description | Typical Value |
|---|---|---|
stats.ttft |
Time To First Token in seconds | ~0.0008s – 0.0015s (sub-2ms) |
stats.decode_rate |
Output token generation speed | 15,000 – 18,500 tokens/s |
stats.prefill_rate |
Context / prompt processing speed | 14,000 – 21,000 tokens/s |
stats.device_compute_time |
Pure on-chip execution time | < 0.05s for 800 tokens |
Configuration & Networking
from chatjimmy import ChatJimmy
client = ChatJimmy(
base_url="https://chatjimmy.ai", # Override base URL if proxied
model="llama3.1-8B", # Default model ID
top_k=8, # Top-K sampling (1 to 8)
system_prompt="Be concise.", # Default system instruction
timeout=45.0, # Timeout in seconds
max_retries=3, # Retries on 429, 502, 503, 504
headers={"X-Custom": "Header"}, # Extra HTTP headers
maintain_history=True, # Auto-manage multi-turn history
)
Error Handling
All SDK exceptions inherit from ChatJimmyError:
from chatjimmy import (
ChatJimmy,
ChatJimmyAPIError,
ChatJimmyConnectionError,
ChatJimmyTimeoutError,
ChatJimmyValidationError,
)
client = ChatJimmy(timeout=5.0)
try:
response = client.chat("Hello!")
except ChatJimmyTimeoutError:
print("The request or stream timed out.")
except ChatJimmyAPIError as e:
print(f"Backend HTTP error {e.status_code}: {e.error_message}")
except ChatJimmyConnectionError:
print("Network connectivity failed.")
except ChatJimmyValidationError as e:
print(f"Input validation error: {e}")
Testing
Run the test suite:
# Run all unit and live integration tests
python -m pytest -v
# Run only mocked unit tests
python -m pytest -v -m "not live"
# Run only live integration tests
python -m pytest -v -m "live"
Known Limitations
- Available Models: The platform is dedicated to
llama3.1-8B. Requests specifying other model IDs are mapped to the default model. - Attachment Size Limit: The platform enforces a strict 50KB limit on file attachments.
- Context Length: The backend enforces context limits; if reached, the generation terminates with reason
"max context limit <N> reached". - Service Availability: Because this library interfaces with a public web endpoint without an SLA, changes to
https://chatjimmy.aimay require SDK updates.
License
Distributed under the MIT License.
Metadata
Release files for chatjimmy 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| chatjimmy-0.1.0.tar.gz | 26.1 kB | Without provenance | Details for chatjimmy-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| chatjimmy-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for chatjimmy-0.1.0-py3-none-any.whl |
Total release size: 48.0 kB
Release files / chatjimmy-0.1.0.tar.gz
| Download URL | chatjimmy-0.1.0.tar.gz |
|---|---|
| Size | 26.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
965ae10a4ea838b19c27e2c787817a05b77a45b41aae6d2a2d8e4798394d7278
|
|
BLAKE2b-256 checksum How to use checksums |
63f3224c940ac3b40c5a12d19323dc4777b9938a94be0aac50eab31a68b308b0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.6
|
Release files / chatjimmy-0.1.0-py3-none-any.whl
| Download URL | chatjimmy-0.1.0-py3-none-any.whl |
|---|---|
| Size | 21.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3bd321e408e5fc1c8353e406465873c60461c0d9fc47b48bf6a3855865df750e
|
|
BLAKE2b-256 checksum How to use checksums |
0a10d54b1d27c82857277eae6f9a8259168d8f50d15a1f42457762f3cddcc816
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.6
|
Security for chatjimmy 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.