AI BEHAVIOR DESIGN, ENGINEERED FOR SIMPLICITY

GET STARTED

LEARN MORE

from thoughtlathe import ThoughtLathe

client = ThoughLathe()

completion = client.chat.completions.create(
	messages=[{"role": "user", "content": "hey there!" }],
	model="gemini-1.5-flash-002",
	lathe_config=dict(
		name="friendly-conversation",
		description="Engage the user in friendly conversation",
		optimization=dict(
			budget=0.1,
			deploy_mode="auto",
		)
	)
)

INTUITIVE AI BEHAVIOR DESIGN

Hyperparemeter search in the background to deliver performance improvements and cost reductions.

LEARN MORE

AI RELIABILITY ENGINEERING

Automatic unit tests and evaluators to ensure your system behaves reliably.

LEARN MORE

from thoughtlathe import ThoughtLathe

client = ThoughtLathe()
completion = await client.chat.completions.create(
   messages=[{"role": "user", "content": "hey there!"}],
   model="gemini-1.5-flash-002",
   lathe=dict(
       name="friendly-conversation",
       evaluators=dict(
           truthfulness="raise",  # raise exception on failure
           toxicity="retry"       # retry on failure
       )
   )
)

try:
   response = completion.choices[0].message
   print(f"Response: {response}")
   print(f"Evaluation results: {completion.choices[0].evaluations}")
except LLMEvaluationError as e:
   print(f"Failed {e.evaluator} evaluation: {e.reason}")
from thoughtlathe import ThoughtLathe

client = ThoughLathe()

completion = client.chat.completions.create(
	messages=[{"role": "user", "content": "hey there!" }],
	model="gemini-1.5-flash-002",
	lathe_config=dict(
		name="friendly-conversation",
		description="Engage the user in friendly conversation",
		optimization=dict(
			budget=0.1,
			deploy_mode="auto",
		)
	)
)

ZERO-FRICTION IMPLEMENTATION

Hyperparemeter search in the background to deliver performance improvements and cost reductions.

LEARN MORE