)
The Notdiamond Python library provides convenient access to the Notdiamond REST API from any Python 3.9+ application. The library includes type definitions for all request params and response fields, and offers both synchronous and asynchronous clients powered by httpx.
Not Diamond specializes in Prompt Optimization - automatically optimizing your prompts to work optimally across different LLMs. Each language model has unique characteristics, instruction-following patterns, and preferred prompt formats. A prompt that works perfectly for GPT-5 might perform poorly on Claude or Gemini. Manually rewriting prompts for each model is time-consuming and requires deep expertise in each model's quirks.
The Solution: Not Diamond automatically optimizes your prompts with:
- Automatic optimization of both system and user prompts
- Built-in evaluation metrics
- Minimum 25 training examples recommended
- Processing time: typically 10–30 minutes
The REST API documentation can be found on docs.notdiamond.ai. The full API of this library can be found in api.md.
# install from PyPI
pip install notdiamondimport os
from notdiamond import NotDiamond
client = NotDiamond(
api_key=os.environ.get("NOT_DIAMOND_API_KEY"), # This is the default and can be omitted
)
# Step 1: Start a prompt optimization job with prototype mode
result = client.prompt_optimization.optimize(
fields=["question"],
system_prompt="You are a mathematical assistant that counts digits accurately.",
target_models=[
{
"model": "claude-sonnet-4-5-20250929",
"provider": "anthropic",
},
{
"model": "gemini-2.5-flash",
"provider": "google",
},
],
template="Question: {question}\nAnswer:",
train_goldens=[
{
"fields": {"question": "How many digits are in (23874045494*2789392485)?"},
"answer": "20",
},
{
"fields": {"question": "How many odd digits are in (999*777*555*333*111)?"},
"answer": "10",
},
{
"fields": {"question": "How often does the number '17' appear in the digits of (287558*17)?"},
"answer": "0",
},
{
"fields": {"question": "How many even digits are in (222*444*666*888)?"},
"answer": "16",
},
{
"fields": {"question": "How many 0s are in (1234567890*1357908642)?"},
"answer": "2",
},
],
test_goldens=[
{
"fields": {"question": "How many digits are in (9876543210*123456)?"},
"answer": "15",
},
{
"fields": {"question": "How many odd digits are in (135*579*246)?"},
"answer": "8",
},
{
"fields": {"question": "How often does the number '42' appear in the digits of (123456789*42)?"},
"answer": "1",
},
{
"fields": {"question": "How many even digits are in (1111*2222*3333)?"},
"answer": "10",
},
{
"fields": {"question": "How many 9s are in (999999*888888)?"},
"answer": "11",
},
],
evaluation_metric="LLMaaJ:Sem_Sim_1", # Or use custom evaluation
prototype_mode=True, # Enable faster prototype mode for quick experimentation
)
print(f"Optimization started: {result.optimization_run_id}")
# Step 2: Poll for completion (typically takes 10-30 minutes)
while True:
status = client.prompt_optimization.get_optimziation_status(result.optimization_run_id)
print(f"Status: {status.status}")
if status.status == "queued":
print(f"Queue position: {status.queue_position}")
if status.status in ["completed", "failed"]:
break
time.sleep(30) # Poll every 30 seconds
# Step 3: Get the optimized prompts
if status.status == "completed":
results = client.prompt_optimization.get_optimization_results(result.optimization_run_id)
print(f"\nOrigin model baseline: {results.origin_model.score:.2f}")
for target in results.target_models:
print(f"\n{'='*50}")
print(f"Model: {target.api_model_name}")
print(f"Optimized System Prompt:\n{target.system_prompt}")
print(f"Optimized Template:\n{target.user_message_template}")
print(f"Pre-optimization score: {target.pre_optimization_score:.2f}")
print(f"Post-optimization score: {target.post_optimization_score:.2f}")
print(f"Improvement: {((target.post_optimization_score / target.pre_optimization_score - 1) * 100):.1f}%")
print(f"Cost: ${target.cost:.4f}")For more details, see the Prompt Optimization documentation.
Select the best model automatically:
import os
from notdiamond import NotDiamond
client = NotDiamond(
api_key=os.environ.get("NOT_DIAMOND_API_KEY"), # This is the default and can be omitted
)
response = client.model_router.select_model(
llm_providers=[
{"model": "gpt-4o", "provider": "openai"},
{"model": "claude-sonnet-4-5-20250929", "provider": "anthropic"},
{"model": "gemini-2.5-flash", "provider": "google"},
],
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Explain quantum computing in simple terms"},
],
)
print(response.providers)For even better performance, you can train a custom router on your own dataset. This allows the router to learn the specific patterns and preferences of your use case:
from pathlib import Path
from notdiamond import NotDiamond
client = NotDiamond(
api_key=os.environ.get("NOT_DIAMOND_API_KEY"), # This is the default and can be omitted
)
client.custom_router.train_custom_router(
dataset_file=Path("/path/to/file"),
language="english",
llm_providers='[{"provider": "openai", "model": "gpt-4o"}, {"provider": "anthropic", "model": "claude-sonnet-4-5-20250929"}]',
maximize=True,
prompt_column="prompt",
)Nested request parameters are TypedDicts. Responses are Pydantic models which also provide helper methods for things like:
- Serializing back into JSON,
model.to_json() - Converting to a dictionary,
model.to_dict()
Typed requests and responses provide autocomplete and documentation within your editor. If you would like to see type errors in VS Code to help catch bugs earlier, set python.analysis.typeCheckingMode to basic.
When the library is unable to connect to the API (for example, due to network connection problems or a timeout), a subclass of notdiamond.APIConnectionError is raised.
When the API returns a non-success status code (that is, 4xx or 5xx
response), a subclass of notdiamond.APIStatusError is raised, containing status_code and response properties.
All errors inherit from notdiamond.APIError.
import notdiamond
from notdiamond import NotDiamond
client = NotDiamond()
try:
client.prompt_optimization.optimize(
fields=["question"],
system_prompt="You are a helpful assistant.",
target_models=[
{
"model": "claude-sonnet-4-5-20250929",
"provider": "anthropic",
},
{
"model": "gemini-2.5-flash",
"provider": "google",
},
],
template="Question: {question}\nAnswer:",
train_goldens=[
{"fields": {"question": "What is 2+2?"}, "answer": "4"},
# Add at least 25 examples...
],
test_goldens=[
{"fields": {"question": "What is 3*3?"}, "answer": "9"},
],
)
except notdiamond.APIConnectionError as e:
print("The server could not be reached")
print(e.__cause__) # an underlying Exception, likely raised within httpx.
except notdiamond.RateLimitError as e:
print("A 429 status code was received; we should back off a bit.")
except notdiamond.APIStatusError as e:
print("Another non-200-range status code was received")
print(e.status_code)
print(e.response)Error codes are as follows:
| Status Code | Error Type |
|---|---|
| 400 | BadRequestError |
| 401 | AuthenticationError |
| 403 | PermissionDeniedError |
| 404 | NotFoundError |
| 422 | UnprocessableEntityError |
| 429 | RateLimitError |
| >=500 | InternalServerError |
| N/A | APIConnectionError |
By default requests time out after 1 minute. You can configure this with a timeout option,
which accepts a float or an httpx.Timeout object:
from notdiamond import NotDiamond
# Configure the default for all requests:
client = NotDiamond(
# 20 seconds (default is 1 minute)
timeout=20.0,
)
# More granular control:
client = NotDiamond(
timeout=httpx.Timeout(60.0, read=5.0, write=10.0, connect=2.0),
)
# Override per-request (note: prompt optimization may take 10-30 minutes, so increase timeout accordingly):
client.with_options(timeout=120.0).prompt_optimization.get_optimziation_status(
optimization_run_id="your-optimization-run-id"
)On timeout, an APITimeoutError is thrown.
Note that requests that time out are retried twice by default.
These methods return an APIResponse object.
This library is typed for convenient access to the documented API.
If you need to access undocumented endpoints, params, or response properties, the library can still be used.
To make requests to undocumented endpoints, you can make requests using client.get, client.post, and other
http verbs. Options on the client will be respected (such as retries) when making this request.
import httpx
response = client.post(
"/foo",
cast_to=httpx.Response,
body={"my_param": True},
)
print(response.headers.get("x-foo"))If you want to explicitly send an extra param, you can do so with the extra_query, extra_body, and extra_headers request
options.
To access undocumented response properties, you can access the extra fields like response.unknown_prop. You
can also get all the extra fields on the Pydantic model as a dict with
response.model_extra.
You can directly override the httpx client to customize it for your use case, including:
- Support for proxies
- Custom transports
- Additional advanced functionality
import httpx
from notdiamond import NotDiamond, DefaultHttpxClient
client = NotDiamond(
# Or use the `NOTDIAMOND_BASE_URL` env var
base_url="http://my.test.server.example.com:8083",
http_client=DefaultHttpxClient(
proxy="http://my.test.proxy.example.com",
transport=httpx.HTTPTransport(local_address="0.0.0.0"),
),
)By default the library closes underlying HTTP connections whenever the client is garbage collected. You can manually close the client using the .close() method if desired, or with a context manager that closes when exiting.
from notdiamond import NotDiamond
with NotDiamond() as client:
# make requests here
...
# HTTP client is now closedThis package generally follows SemVer conventions, though certain backwards-incompatible changes may be released as minor versions:
- Changes that only affect static types, without breaking runtime behavior.
- Changes to library internals which are technically public but not intended or documented for external use. (Please open a GitHub issue to let us know if you are relying on such internals.)
- Changes that we do not expect to impact the vast majority of users in practice.
We take backwards-compatibility seriously and work hard to ensure you can rely on a smooth upgrade experience.
We are keen for your feedback; please open an issue with questions, bugs, or suggestions.
If you've upgraded to the latest version but aren't seeing any new features you were expecting then your python environment is likely still using an older version.
You can determine the version that is being used at runtime with:
import notdiamond
print(notdiamond.__version__)Python 3.9 or higher.