See it in action
DSPy: see a signature produce the author’s haiku
Build the first haiku program from the author’s complete LM setup and signature example. The printed poem is the tutorial’s illustrative output; its exact generation model is unspecified. The copied recipe preserves the setup model string and API-key placeholder.
Example input
import dspy
# Pass the key explicitly...
lm = dspy.LM("openai/gpt-5-nano", api_key="YOUR_OPENAI_API_KEY")
dspy.configure(lm=lm)
haiku_signature = "subject -> haiku"
haiku_generator = dspy.Predict(haiku_signature)
result = haiku_generator(subject="computer science")
print(result.haiku)Example output
Silent code unfolds
Logic threads through hidden paths
Bugs bloom, then resolveLicense
Complete original license and any source copyright retained. Source-authored examples are separate from site-reproduced results; no original scripts, installations, model/API requests, account creation, memory writes, graph queries or tests were executed. Complete pinned source blocks are assembled in documented order, with their separate file bindings retained in the review record. The SDK setup and printed illustrative poem come from separate source sections; no exact generation-model binding is claimed.
MIT License
Copyright (c) 2023 Stanford Future Data Systems
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
DSPy: specify setting, mood and season
The full source signature adds setting, mood and a Literal season before calling the haiku program for Bodega Bay. The author’s printed poem stays alongside the recipe. Setup and sample output come from separate documented sections, so the poem’s exact model is unspecified.
Example input
import dspy
# Pass the key explicitly...
lm = dspy.LM("openai/gpt-5-nano", api_key="YOUR_OPENAI_API_KEY")
dspy.configure(lm=lm)
from typing import Literal
Season = Literal[
"spring", "summer", "autumn", "winter",
]
class HaikuBot(dspy.Signature):
"""
Write a classical haiku given the provided inputs.
"""
location: str = dspy.InputField()
mood: str = dspy.InputField()
season: Season = dspy.InputField()
haiku: str = dspy.OutputField()
haiku_bot = dspy.Predict(HaikuBot)
result = haiku_bot(location="Bodega Bay", mood="mysterious", season="autumn")
print(result.haiku)Example output
Fog drifts over waves,
Crimson leaves swirl by the pier—
Night whispers secrets.License
Complete original license and any source copyright retained. Source-authored examples are separate from site-reproduced results; no original scripts, installations, model/API requests, account creation, memory writes, graph queries or tests were executed. Complete pinned source blocks are assembled in documented order, with their separate file bindings retained in the review record. The SDK setup and printed illustrative poem come from separate source sections; no exact generation-model binding is claimed.
MIT License
Copyright (c) 2023 Stanford Future Data Systems
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
What DSPy offers
- 01Define model tasks with typed signatures
- 02Compose reusable modules in Python
- 03Optimize prompts using examples and evaluation metrics
What makes it different
DSPy treats model calls as parts of a program that can be evaluated and optimized. Useful results depend on representative examples and the metric supplied by the developer.
Who should consider it
- Developers evaluating language-model pipelines
- Teams tuning repeatable AI tasks
What to verify
- Try one real task and compare its output with your requirements.
- Check what data the service stores, shares, or uses for training.
- Confirm current pricing, usage limits, and cancellation terms on the official site.
Official link checked on Oct 20, 2026. Feature claims have not been independently tested.