Skip to main content
DSPy is a framework for programming language models rather than relying on static prompts. It enables you to build modular AI systems with code instead of hand-crafted prompting, and it offers methods to automatically optimize these systems. Features
  • Programmatic approach to LLM interactions through Python
  • Modular components for building complex AI pipelines
  • Self-improvement algorithms that optimize prompts and weights
  • Support for various applications from simple classifiers to RAG systems and agent loops

​
Installing Libraries

pip install -U dspy -q
Set your Neosantara AI API key: 
export NAI_API_KEY=***

​
Example

Setup and connect DSPy to LLMs on Neosantara AI 
import dspy

#Configure dspy with a LLM from Neosantara AI
lm = dspy.LM('openai/nusantara-nase',
             api_key=os.environ.get("NAI_API_KEY"),
             api_base="https://api.neosantara.xyz/v1")

#now you can call the LLM directly as follows
lm("Say this is a test!", temperature=0.7)  # => ['Ini adalah sebuah tes.']
lm(messages=[{"role": "user", "content": "Say this is a test!"}])  # => ['Ini adalah sebuah tes.']
Now we can set up a DSPy module, like dspy.ReAct with a task-specific signature. For example, question -> answer: str tells the module to take a question and to produce a string answer below. 
#Configure dspy to use the LLM
import dspy

#Configure dspy with a LLM from Neosantara AI
lm = dspy.LM('openai/archipelago-70b',
             api_key=os.environ.get("NAI_API_KEY"),
             api_base="https://api.neosantara.xyz/v1")

#Configure dspy to use the LLM
dspy.configure(lm=lm)

## Gives the agent access to a wikipedia search tool
def search_wikipedia(query: str):
    # Use a more general search or the specific page if known to be reliable
    # For a general search, you might use a different tool or API
    # For this specific case, we stick to the provided URL but simplify the output
    results = dspy.ColBERTv2(url='https://id.m.wikipedia.org/wiki/Prabowo_Subianto')(query, k=5) # Get 5 result
    # Return the raw text content for the LLM to process
    return results[0]['text'] if results else "No relevant information found."


## setup ReAct module with question and string answer signature
# Changed the signature to expect a string answer
react = dspy.ReAct("question -> answer: str", tools=[search_wikipedia])

# Use the question that needs a text answer
pred = react(question="when the birthday prabowo?")

print(pred.answer)
The output will return an string for answer below:
output
 Prabowo Subianto lahir pada tanggal 17 Oktober 1951.